Poor Adherence to Antipsychotic Medication in People with Schizophrenia: Diffusion, Consequences and Contributing Factors

Abstract

This chapter tries to present the issue of poor adherence to treatment in people with schizofrenia, by treating the following topics: role of antipsychotics in the treatment of schizophrenia, compliance versus adherence, assessment of medication adherence, frequency of poor adherence to antipsychotics and consequences of poor adherence to antipsychotics present and medium-term perspectives.

Premise

Even panacea does not work when the patient intentionally or unintentionally fails to take it. Once it is accepted that “the fate of a drug therapy is with the patient” [1], the inevitable consequence is that the patient is the ultimate health care decision maker. These common sense observations get even more complicated by factual evidence that the “ideal of the patient as a passive obedient recipient of medical instructions” [2] is far from the real world. Human kind is indeed generally reluctant to take medicines in the absence of adequate support.

The negative predisposition toward medicines has very old roots. More than 2000 years ago, for example, the conflicting nature of the man–drug ticket was popular, so that Titus Lucretius Carus [3] used it in an allegory to explain the stratagem of using poetry to disseminate the Epicurean lesson; in the fourth book of the De Rerum Natura, the poet–philosopher referred to the expedient of the physician who sweetens the rim of the glass with honey to induce the sick boy to drink the bitter absinth.

The presence of a disease may be not enough to motivate patients to follow the physician’s prescriptions and dedicated incentives may be needed to induce correct use of medication, but it is also true that many proposed interventions are often unproductive in practice. For example, it was reported that, in a sample of healthy volunteers who were paid to take a daily single dose of aspirin and received explicit, written and verbal instruction to follow the prescription, only 35 % adherence was observed over a 2-week period [4]. Based on the various components of this discouraging sequela, it is far from surprising that poor adherence to medication regimens detracts relevant resources from the health care system and creates a heavy burden for the patients, the families and society. These dramatic consequences occur with all types of diseases and all types of medicines, even placebos. A meta-analysis of studies evaluating the association between adherence to drug therapy and mortality depicts the generality of the phenomenon [1]: compared with poorly adherent patients, those with good adherence presented lower mortality, irrespective of attribution to a beneficial drug or placebo. Considering the well-known tenet that placebo has little effect on health outcomes, it seems reasonable [1] that medication adherence is a proxy expression for a healthy adherer effect related to overall healthy behaviour.

Two examples demonstrate that major psychiatric disorders in general and psychoses in particular are privileged vehicles for poor healthy adherer effect, although with possible disorder-related specificities. The first is that compulsory treatments unrelated to quarantine constitute an almost exclusive prerogative of mental disorders. The second refers to the chronologic link that exists between the beginning of the psychopharmacologic era and the first explicit mention of medication non-compliance in a patient affected by a severe mental disorder. In a historical article from 1949 on the first series of 10 cases treated with lithium for psychotic excitement, John Cade [5] reported that a patient, a man with a “state of chronic manic excitement” recovered sufficiently on lithium to make return to “his old job” possible but developed a full-blown relapse leading to readmission to hospital once “he became more lackadaisical about his medicine and finally ceased taking it” even though he had received “instructions to take a maintenance dose of lithium carbonate, five grains twice a day”.

Role of Antipsychotics in the Treatment of Schizophrenia

Nowadays, it is fully recognised that the best treatment option for people with schizophrenia requires multidisciplinary interventions. Nevertheless, it remains indisputable that antipsychotic drugs are the mainstay of therapeutic intervention. International guidelines on the treatment of schizophrenia, systematic reviews and meta-analyses [6–18] unequivocally lead to the same conclusion: although they are far from ideal as to efficacy and tolerability, antipsychotics, especially second-generation antipsychotics, play a vital role not only in the attenuation and suppression of psychotic symptoms but also in the prevention of relapses and recurrences. The importance of continuous maintenance therapy with antipsychotics for a good long-term prognosis is also manifest after the first episode of schizophrenia. For example, first-episode patients have been reported to have relapses fivefold more frequently when antipsychotics are not taken as prescribed [19]. However, evidence-based conclusions probably do not present the best picture of pharmacotherapy on schizophrenia outcome. To get a more immediate idea on this issue, it may be for example remembered that individuals with psychosis experienced quite different results in the years immediately before and subsequent to the advent of chlorpromazine, the first of the neuroleptic medications [20, 21]. Previously, almost half of the patients admitted to psychiatric hospitals had to spend more than 10 years of their life in hospital [22]. When chlorpromazine became available, the number and length of the hospitalisations started to decrease enough to allow the transition, previously considered to be a largely utopian idea, from a model of care centred on the psychiatric hospital to a community-based approach. In addition, comparison of cohorts of patients evaluated in the same setting before and after the introduction of antipsychotic medications confirms how “the course of schizophrenia has become less malignant” concomitant with the advent of this class of agents, as indicated, in particular, by “the marked reduction in occurrence of catastrophic schizophrenia and virtual disappearance of catatonic schizophrenia over this timeframe” [23].

Although about half a century has elapsed since the first pioneering attempts at deinstitutionalisation, the valuable therapeutic potential of antipsychotic medications remains under-expressed for a significant number of patients. Successful pharmacotherapy for schizophrenia requires continuous consumption [24–28] but, in clinical practice, this requirement is far from being realised in a significant number of patients.

Schizophrenia has been reported [28] to rank second among the clinical conditions characterised by major difficulties in achieving levels of medication adherence sufficient enough to obtain a therapeutic effect, which attests to the presence of relevant obstacles to the use of antipsychotics.

Compliance Versus Adherence

The acknowledgement that patients frequently do not follow the prescriptions recommended by their doctors can be traced back to ancient times. However, the entry into medical terminology of the words compliance and adherence for this peculiar behaviour was relatively recent [29]. The authors of early pioneering reports on the effects of neuroleptics in the real world have systematically used circumlocutions such as “patients who do not take their drugs” [30], “patients who failed to take the drugs prescribed” [31] or patients who do not take the “courses of the drug prescribed as indicated by the doctor” [32]; the words compliance and adherence were not used to describe medication-taking behaviour.

Several definitions for compliance and adherence have been proposed over the years. However, all of them focus, with nuances, on “the extent to which a person’s behaviour coincides with medical or health advice” [33]. The label embraces a wide spectrum of conditions such as, for example, “failure to enter a treatment program, premature termination of therapy, and incomplete implementation of instructions, including prescriptions” [34] along a continuum of distinct behaviours related to the amount and timing of medicines actually taken.

Compliance and adherence could also be considered according to a functional perspective as measures of health outcome, for example “the number of doses not taken or taken incorrectly that jeopardize the therapeutic outcome” [35] or “the point below which the desired preventive or desired therapeutic result is unlikely to be achieved” [36].

However, the definitions based on the extent to which a patient deviates from prescriptions from the health care provider or on the clinical consequences of a defined medication-taking behaviour do not get to the heart of the difference between compliance and adherence. The difference refers principally on two distinct models of the patient–doctor relationship. In particular, treatment and medication compliance are the expressions of a process of care centred on an indisputable leading actor, the paternalistic doctor who prescribes or, better, orders the therapy for a supporting actor, the patient, without paying appreciable attention to an alliance. Treatment and medication adherence are the expressions of a process of care centred on two principals, the doctor and the patient who are actively engaged in genuinely shared decisions within their specific roles. Therefore, the gap between compliance and adherence is not trivial and not confined to mere semantics but has relevant implications for clinical practice. Despite this, many publications incorrectly use compliance and adherence as synonyms.

The prevalence of one term or the other has fluctuated over time. The result of a MEDLINE search on articles from 1976 to 2011 citing compliance and/or adherence in schizophrenia (Fig. 1) testifies to how the initial preference for compliance has been progressively changed to a gain in popularity for adherence; nowadays the terms are used almost equally. The risk is that the current inclination to qualify the reduced propensity of patients to take medicines as adherence depends principally on its more politically correct profile rather than real evolution of patient–doctor communication.

Fig. 1

Changes over the years of the MEDLINE compliance/adherence citation ratio in studies on schizophrenia

Given the differences characterising the two labels, research on medication-taking behaviour should state whether a study refers to compliance or adherence. In this regard, an acceptable, easy option could consist of systematic reports on the interventions used to promote active participation of the patients within the treatment project¹.

Assessment of Medication Adherence

Treatment adherence is a favourite topic in modern research on schizophrenia and related disorders. The growing interest of the scientific community in the topic is illustrated by the almost 1,700 citations in the first decade of the millennium extracted from the National Library of Medicine’s PubMed online search engine using the keywords “adherence” and “compliance” in combination with “schizophrenia” (Fig. 2).

Fig. 2

Annual MEDLINE citations of the keywords “compliance” and “adherence” in combination with schizophrenia: period 2000–2011

However, studies on treatment adherence are commonly vitiated by shortcomings that inherently interfere with the possibility of valid measurements [28, 37–42]. In particular, the available methods of assessment cover a very large number of approaches: for example, directly observed therapy, clinical response, patient’s diaries and dedicated interviews, reports concerning factors, attitudes and opinions that can influence the adherence process, indications of individual autonomy in taking medicines, specific queries directed to significant others, judgement of the treating physician or, more broadly, the provider, chart reviews, formal pill count, rates of refilled prescriptions, microelectronic monitoring measures such as MEMS, changes in physiologic markers, drug concentrations and analysis of tracer substances in body specimens.

Methods differ in many aspects. Some are direct, others indirect, and in both cases it is possible to distinguish between approaches that are objective or subjective, qualitative or quantitative and self-reported or informant reported [34, 37, 38, 42–49]. Furthermore, the many measuring methods allow for alternative instruments.

Despite the complex scenario, heterogeneity among distinct measures of medication adherence is generally not sufficiently discussed. This is wrong. The degree of concordance and correlation between independent methods of measurement is fairly scant. The literature on this issue appears unequivocal. For example, when three dedicated questionnaires, one clinician rated and two patient rated, were applied to 329 schizophrenia patients living in four European cities and participating in the QUATRO study [39], a concordant label of non-adherence was attributed to only 4 % of the cases; rates of individuals classified as non-adherent by any single questionnaire fluctuated widely: 54.9, 20.4 and 14.1 %, respectively. A measurement effect has also been documented in a group of patients with a first episode of psychosis tested with four methods [50]: the rate of adherence was estimated at 91 % by family members, 83 % by patients, 76 % by clinicians and 73 % by pill counting. Similarly, the application of a medical record-based MPR over 1 year and a five-point patient’s self-report scale to 1,579 US subjects with a diagnosis of schizophrenia [51] showed that only 8.8 % of the sample classified as non-adherent according to one measure was also non-adherent for the other measure. Furthermore, in a small group of patients with schizophrenia followed for 3 months after discharge from hospital [52], 77, 65 and 40 % of the total sample were considered poorly adherent according to plasma drug concentration, pill count, or self-assessment, respectively. Poor correlation between the same three measures was also detected in another independent 3-month study [53]: 9, 23 and 55 % of the sample were judged adherent when pill count, blood level data or patient’s self-reports were considered. Even poorer correlations between different methods for measuring adherence emerge when electronic monitoring was considered. For example, the application of self-reports, physician reports, pill counts, electronic monitoring and drug plasma concentration [54] showed not only that pill counts and MEMS were strongly correlated with each other and weakly correlated with self-reports and physician ratings but also that drug plasma levels were not correlated at all with any other measure of adherence. Similarly, a 3-month study limited to 25 subjects with schizophrenia reported rates of 48 and 0 % non-adherent patients according to a MEMS daily adherence below 70 % or a clinician rating score equal to or less than four [55]. In a subsequent report by the same group [56] on 61 patients with schizophrenia or schizophreniform disorder followed for up to 6 months, electronic monitoring was challenged against three visual analogue scales elaborated by the prescriber, the research assistant and the patient. MEMS registered 57 % of non-adherent patients, which was close to the 54 % reported by the research assistants but higher than the 7 % estimated by the prescribers and the 5 % indicated by the patients. These studies suggest that MEMS plausibly detects greater non-adherence rates than other methods. An 8-week study [57] of 51 Korean outpatients with schizophrenia treated with a single antipsychotic medication reported similar results: the rate of non-adherence according to MEMS was considerably higher (41 %) than the 26, 8, and 8 % for patient’s self-report, pill count and clinician rating scale, respectively. Notably, 38 % of the patients labelled adherent by the clinician were judged non-adherent by MEMS. Given the weak concordance between the different measures, it is far from surprising that results on the clinical consequences of poor medication adherence are appreciably influenced by the specific method of measuring used in the different studies [58].

However, there is some evidence of good correlation between different measures of medication adherence. For example, in an observational, prospective, large scale, German study [59], physician’s and patient’s ratings of compliance showed 93.2 % concordance. However, irrespective of the measure used, the presence of very high compliance rates, more than 80 %, makes the validity of the results questionable.

An obvious measure effect is evident when independent studies on adherence or compliance are reviewed. For example, in a comparative analysis of findings based on patient interviews, urine tests or clinician assessments, the estimates of compliance were 52, 60 and 72 %, respectively [40].

Thus, that different measures of adherence fail, de facto, to estimate the same trait seems very plausible. Even a superficial comparison between the main characteristics of individual instruments of measuring corroborates this suggestion.

For example, a formal impression of treatment adherence is in sharp contrast to the use of rigorously standardised methods of quantification. Measures that are over-inclusive and charged by relevant confounders (e.g. the drop-out rate) are conceptually far apart from other measures (e.g. the drug-attitude inventory) where the focus on a partial aspect of adherence may lose touch with a global approach to medication-taking behaviour.

Another issue of divergence between studies involves the adoption of a dimensional or categorical approach for measuring adherence. So far, the latter has been more popular, probably because it is easier and more immediate: a simple all-or-nothing partition of medication adherence according to a pre-ordered cut-off. However, the categorical approach has some weak points. For example, the cut-offs used in independent studies vary and the preference for one or another threshold seems largely arbitrary, in the absence of a solid base with a predefined clinical effect. It is also true that definite cut-offs to separate adherent patients from non-adherent patients could be clinically meaningful in the case of particular clinical characteristics (e.g. symptom severity) but not necessarily so when the focus is on another index (e.g. long-term prognosis). A supplementary criticism involves the issue of partial adherence. The “notion of noncompliance as complete, wilful cessation of all antipsychotic medications is not an accurate representation of actual medication-taking behaviour among outpatient populations with schizophrenia” [60] because many subjects should be classified as partially compliant rather than non-compliant. Furthermore, “partial compliance may take several forms, including taking an amount that is consistently less than recommended, irregular (‘on-and-off’) dosing behaviour, and having discrete gaps in antipsychotic therapy” [60]. This adds complexity to a vexing and already complex topic. The categorical approach counts all patients posited above a definite threshold as adherent but it does not consider that, for a non-negligible minority of individuals, deviation from the prescribed medication consists of taking medicines in excess. For example, in a large-scale study involving 3,968 veterans treated with antipsychotics in monotherapy, the rate of medication over supply was 7.6 % [61]. Therefore, in practice, the label adherent is applied to a mixed population of truly adherent and over-adherent individuals.

Thus, a dimensional approach seems inherently preferable, because the use of a continuous measure is more indicated when the phenomenon under scrutiny is multifactorial in origin and devoid of an unequivocally validated discriminatory threshold. Both these requirements are fully satisfied in the case of medication adherence. Nevertheless, the application of a purely dimensional design may be problematic in practice, because it commonly generates results that are hard to interpret and require wide sample populations. Partition into a number of predefined classes acting in continuity with each other could be a reasonable strategy to bypass these limitations.

Each method for measuring medication adherence has specific and appreciable pitfalls. For example, phenomena such as aversion to pharmacologic treatments and unjustified beliefs about convenience or the need to please doctors and significant others may frequently induce patients to voluntarily over report their adherence. This manipulation primarily damages those measures directly or indirectly based on the personal reports of the patient but may also undermine more objective approaches, such as direct pill counts, MEMS, photos, monitoring of drug concentration and analysis of a tracer substance. However, direct pill counts, MEMS and photos certify that the blister is empty, the bottle has been opened or the pills are in the palm of the hand but do not prove that the patient has really ingested the therapy as prescribed. Physiologic markers and levels of drugs or tracer substances investigate medication intake during the few days immediately before the assay but may offer distorted information about adherence behaviour outwith this specific time frame. Unfortunately, many patients are susceptible to the white-coat compliance phenomenon [29] immediately before a medical visit and medication-taking behaviour then declines in the interval between two scheduled visits [62]. Some adherence measures may contribute to bad estimates in the absence of deliberate manipulation. For example, patient’s difficulties in recalling detailed information about drug intake may inflate or deflate the rates of poor adherence when the source of information is based on direct reporting. The same misleading consequences may occur when inattentive relevant others are deputed to quantify the ability of the patient to follow regularly the established medication regimen.

Physicians may be not reliable in detecting poor medication adherence [63–66]. Psychiatrists, in particular, have been reported to be especially prone to underestimating the adherence of their clients [65, 67, 68]. The conclusion of a recent large-scale study [67] carried out in Germany and involving 5,729 patients with schizophrenia and 699 psychiatrists working in hospitals or in private practice is representative in this regard. When specifically interviewed, the treating physicians estimated 68 % of unintentional partial compliance during the last month and 69 % intentional lifetime partial compliance. Both these values exceed the mean and modal percentages generally observed when other adherence measures are used. If confirmed, the poor performance by psychiatrists to identify patients with correct medication-taking behaviour could be relatively atypical, because doctors in other specialties have generally been found to be more prone to overestimate compliance [69]. However, contrasting results also exist. For example, in a 1970 VA study [66] reporting “that therapists erred in 20 % of their prediction”, the wrong prediction resulted in 71 % of the cases “in the direction of believing that the patient was not taking his drug or drugs as prescribed when in fact he was”.

The frequencies of poor adherence obtained using direct assays of antipsychotics may be equivocal. For an identical dose, blood levels are subject to relevant inter-individual variability. Furthermore, the presence of poor and ultra-fast metabolizers may contribute to false positives and false negatives in the non-adherent population; because two corresponding cytochrome P450 genotype variants produce opposite effects on drug degradation [70, 71], a regular taker may be spuriously labelled as poorly adherent or non-adherent. The current trend for multi-racial societies requires that attention must be paid to the effects of cytochrome genotyping on medication adherence. Various ethnicities differ in the global distribution of the individual alleles involved in classifying patients’ according to metabolic status [72] and this implies a supplementary, ethnic-dependent risk of erratic estimates of medication adherence when direct assessments of drug plasma levels are performed.

Highly elaborate questionnaires and interviews, microelectronic monitoring, assays of the medicines in body fluids or, more broadly, the use of procedures that deviate from daily clinical routine may also exert distortive effects on habitual medication-taking behaviour, because they selectively channel the patient’s attention on the phenomenon [37, 42].

Methods of measuring such as MEMS, drug monitoring and assays of tracer substances in body fluids have some important limitations. In particular, they are expensive and, therefore, their use contrasts not only with the need to obtain data from large, representative samples of patients but also with current worldwide financial constraints. Furthermore, these approaches may be perceived by some patients as intrusive or dangerous to the point that formal dedicated informed consent is required.

Given the long list of inherent weaknesses that affect almost all measures of medication adherence and the lack of a universal standard of reference, it is easy to see why the choice of one method over another varies greatly among studies. In a recent review of 258 studies [37], the fact that no measure of medication adherence was reported to have the lion’s share (Fig. 3) is well representative of this reality.

Fig. 3

Most utilised methods for measuring adherence to antipsychotic medications (adapted, with permission from Velligan et al. (2006), ref n deg 37)

Frequency of Poor Adherence to Antipsychotics

Awareness that a relevant proportion of people with schizophrenia do not take medicines as prescribed was already manifest when neuroleptics entered the market and mental health community care was in its infancy. A 1962 study [32] of patients with schizophrenia discharged from eight London mental hospitals may be paradigmatic, because it reported that 44 % “of the courses of a drug prescribed were probably not taken as intended by the doctor”, with rates of drug courses classified as “taken but definitely not as ordered”, “taken but probably not as ordered”, “definitely not taken at all”, and “probably not taken at all” posited at 28, 8, 3, and 5 %, respectively. Almost contemporaneously, in a systematic study [31] of all patients with a diagnosis of schizophrenia or schizophreniform state discharged from the West House Division of the Royal Edinburgh Hospital between January 1959 and December 1960, it was emphasised that, “of the 124 patients who had been prescribed neuroleptics at some time during the follow-up period, 54 % probably took them as ordered, 46 % did not take them as ordered”.

These first claims have been corroborated in the decades since then. In a 1986 review [63] of 21 articles published between 1958 and 1984 and involving almost 3,000 patients, non-compliance with oral first-generation antipsychotics, defined as “any significant deviation from the prescribed medication”, was estimated to range between 10 and 76 %, with a median value of 41 %. About 10 years later, another review [64] of “fifteen subsequent studies using varying definitions of noncompliance and many mixing patients taking oral and depot medications reported a median 1-month to 2-year noncompliance rate of 55 %”, with a range from 24 to 88 %. In a review [40] of 24 articles published over a 20-year period and involving 26 groups of patients and 3,590 individuals, the compliance rate was found to fluctuate between a minimum of 24 % and a maximum of 90 %, with an overall mean of 58 %. However, this impressive rate of variability seems largely due to the common inclusion of studies based on disparate measures and definitions of compliance.

For example, at the beginning of the millennium, a review [47] of 39 articles published between 1981 and 2002 compared three different definitions of medication adherence. According to the broadest definition, which was applicable to all the studies, the non-adherence rate ranged between 4 and 72 %, with an unweighted mean and median value of 40.5 and 40 %, respectively. These figures remained substantially unchanged when the analysis was restricted to the 10 articles that adopted a more conservative definition of adherence, “regularly taking medications as prescribed”, and only trained personnel were used to assess medication-taking behaviour. When the selection of the articles involved an even stricter working criterion, “taking medications as prescribed at least 75 % of the time”, the unweighted mean non-adherence value increased to 47.3 %, with a median rate of 47 %. One year later, another systematic review [73] judged 86 studies suitable for re-analysis. Of the 23,796 patients, 5,790 had a diagnosis of schizophrenia, 6,372 had psychosis in general and 11,634 had severe mental illness. The overall weighted mean rate of patients who were found to be non-adherent to medication and/or selected appointments was 25.8 %; the proportion of individuals who did not take pharmacologic therapies as prescribed was 29.7 %, which was slightly but not significantly higher than the proportion of the group that missed the appointment (24.3 %).

In addition to the different measurement methods, other factors are also likely to play a major causative role in discrepancies between results. The setting and the context in which the treatment is carried out certainly belong to this list. Regularity in taking medicines may be influenced by the prescribed drug, referral as inpatient rather than outpatient, compulsory therapies, enrolment from clinical trials or the real world and access to services that are engaged in treatment adherence in different ways. Given the current widespread success of a model of psychiatric care founded on community interventions, the results of the earliest literature on medication adherence by people with schizophrenia seem hardly transferable to the practice of today. Problems of comparability also derive from the fact that, in the clinical practice of the last 15–20 years, second-generation antipsychotics have substantially replaced first-generation agents, which were at the centre of the initial reports on adherence by people with schizophrenia.

The duration of follow-up can also affect medication adherence because of the reported inverse relationship between the two phenomena [28, 55, 74, 75], a relationship that probably starts early after the start of the therapy. Compliance is also likely to vanish over time in patients treated with depot antipsychotics [76]. However, the link between the level of adherence and the length of exposition to medicines has been refuted in other reports. For example, a revision of the literature [73] has failed to show different rates of medication adherence between cross-sectional and prospective studies based on follow-up lasting less than 6 months, between 6 and 12 months or more than 12 months. A preference for the hypothesis of a decrease of adherence with time is also supported by some clinical considerations. In particular, it seems not only that the longer the duration of the illness, the higher the susceptibility for an exacerbation of symptoms leading to poor medication adherence but also that the longer the asymptomatic or quasi-asymptomatic period, the lower the patient’s perception of the need to continue therapy, with the consequence of an increased risk for relapse [34, 77, 78]. Despite the evidence, studies continue to refer to periods of variable length or do not explicitly report the duration of follow-up. Therefore, comparisons between studies are not possible.

A supplementary source of heterogeneity in the prevalence of medication adherence comes from the frequent recruitment of small or relatively small samples. The epidemiology of medication-taking behaviour is the ultimate result of a complex, multi-determined process and is at high risk for selection bias when the study population is not numerically representative.

Several pitfalls affect the construct of the various epidemiologic studies on adherence to antipsychotics to varying degrees, and therefore definite conclusions about the true prevalence of the phenomenon cannot be drawn. However, even when only reports that have used similar measures of medication adherence and challenged samples of more than 1,000 patients followed in a naturalistic setting are considered, an appreciable variability in adherence rates remains between independent studies. Few citations seem sufficiently paradigmatic in this regard. For example, in a population-based study [79] of 6,662 Quebec residents who had a diagnosis of schizophrenia and were treated as outpatients with second-generation antipsychotics, 67.5 % of the sample persisted with the same class of medication because they “filled at least one prescription … in the 45 days before the first anniversary of treatment initiation” and 78.6 % of the persistent group were deemed compliant, as indicated by 80 % or more continuous medication availability. Another study [80] involving 63,214 patients included in the VA National Psychosis Register and treated with oral antipsychotics due to schizophrenia or schizoaffective disorder reported that “approximately 40 % of those receiving one antipsychotic during the year and 38 % of those receiving two different antipsychotics had MPRs less than 0.8, indicating poor antipsychotic adherence”. A subanalysis of patients receiving only one antipsychotic medication has also documented similar percentages of poor adherence, 41.5 and 37.8 %, among the 23,072 patients prescribed a second-generation antipsychotic and the 25,931 patients treated with a conventional agent. Furthermore, “eleven percent of [the patients] receiving one antipsychotic and 19 % of those receiving two antipsychotics during the year received more days’ supply of medication than would be required to take their antipsychotics as prescribed”. A study [81] of Medicaid beneficiaries with schizophrenia treated in San Diego with oral first-generation or second-generation antipsychotics concluded that 24 % of the patients were non-adherent, 16 % were partially adherent, 41 % were adherent and 19 % were excess fillers. This result is in contrast to the report of an MPR-based 1-year study [60] on a cohort of 4,325 California Medicaid patients who were prescribed antipsychotics for the treatment of schizophrenia; less than 16 % of the sample presented an MPR below the 70 % threshold. Similarly, in a multi-site, prospective, naturalistic study [51] involving 1,579 schizophrenia patients extracted from the US-SCAP database and treated in usual care settings with any oral antipsychotic, the 1-year rate of individuals with an MPR of 80 % or less was only 10.2 %.

Persistence of appreciable differences in the results among studies characterised by relevant similarities in design may be seen as a proof of how patient settings and general context may act as confounders. A comparison [82] between patients with schizophrenia who were resident in two Canadian provinces, Quebec and Saskatchewan, documents this. Although the two populations were treated similarly in routine community practice, and for both groups the prescription was limited to risperidone, olanzapine or quetiapine, the index data were taken on the day of the first prescription of one of the three atypicals, and the MPR was used to measure medication-taking behaviour, patients from the two provinces diverged in the rate of individuals falling below the cut-off of an MPR of 80 %. Among the 40,854 and 3,291 patients resident in Quebec and Saskatchewan, a status of moderate to poor compliance was detected in 39 and 55 % of the two sample populations, respectively. The disproportionate size of the two samples could have reasonably contributed to the discrepancy in the results, even though the smaller cohort was probably more representative than the larger one because publicly funded health insurance in Saskatchewan covered a larger proportion of residents.

All the examples reported so far lead to an univocal conclusion: the amount of epidemiologic data on adherence to antipsychotics could increase further but expansion of the references is likely to have little impact on the possibility of stronger consensus on the rate of poor medication-taking behaviour.

The descriptive epidemiologic approach to adherence has promoted deeper understanding of some other relevant aspects of the phenomenon. The first aspect concerns the time frame between hospital discharge or the start of therapy and the emergence of poor adherence. Knowledge of this interval is not trivial in the planning of dedicated, incisive community care interventions. A large, observational, cohort study [83] of patients with schizophrenia from the Maine and New Hampshire Medicaid programmes demonstrated that almost half of the prescription gaps from 1 to 10 days occurred in the first 50 days after initiation of a second-generation antipsychotic, with an appreciable proportion of the gaps occurring in the first month. Furthermore, the rate of subjects who failed to take antipsychotics as prescribed within the first 7 to 10 days after transition from inpatient to outpatient status has been reported to range between 15 and 25 %, in relation to the measure of adherence used [84]. Early emergence of non-adherence to antipsychotic medication has also been observed in recent-onset patients, as indicated by the finding that a "moderate or greater nonadherence typically began approximately six months after clinical stabilization" [85]. The report, based on de-identified computerised pharmacy records from 1,157 US pharmacies, that patients “who had not filled a prescription for an antipsychotic during the 180-day period prior to the index date had a … ten-fold increase in the risk for medication discontinuation at the start of the therapy” [86] gives further indirect support to the idea that the time period around the start of the therapy is crucial for patients with schizophrenia to persist and adhere to their therapy. This very short interval is not surprising. In clinical routine, rehospitalisation soon after discharge is common among people with schizophrenia spectrum disorders. Furthermore, early occurrence of poor adherence is not confined to antipsychotics and schizophrenia but seems to represent a generalised event that is far from new. For example, almost half a century ago, it was reported that 3 % of 2019 prescription orders in general medicine were not filled within 10 days [87].

Another issue refers to the persistence over time of a defined adherence status. As emphasised earlier, non-adherence seems to become more diffuse as the disorder progresses. For example, follow-up of a group of 127 actively adherent patients with schizophrenia showed that 75 % remained adherent at 5.2 months and 50 % at 13.7 months, and, more broadly, that “the risk of becoming nonadherent was fairly even across the 22 month study period” [88]. This finding agrees with the observation [74] that, in patients with schizophrenia, the prevalence of discontinuation or interrupted use of antipsychotics was approximately 50 % after 1 year and 75 % after 2 years. Nevertheless, once established, non-adherence seems to have a discrete persistence, at least in the absence of dedicated interventions. The switch from a poor to a good medication adherence status has been reported to occur fairly quickly with the possibility of transition decreasing substantially with time [88]. Reports that MPR values significantly correlated with each other over a 3-year period [89] and that previous adherence was the best predictor of future adherence [51] supports the proposal that individual proneness to poor medication-taking behaviour is expressive of a relatively stable trait. In agreement with this, it was found that during the 4 years after an index discharge from hospital, “the majority of patients identified as medication noncompliant … continued being noncompliant… after subsequent readmissions” [90]. The stability of poor adherence cannot be dismissed as irrelevant, because it supports the uncommon use of repeated measures and just the pre-post comparison in trials on the efficacy of interventions aimed at improving medication adherence.

Another clinically relevant chronologic issue pertains to the stage of the disorder at which poor adherence to antipsychotics becomes manifest. Overall, the literature indicates high rates of poor adherence among patients with schizophreniform disorder, first-episode schizophrenia or, more broadly, early schizophrenia [19, 91–97]. An experience derived from the Suffolk County Mental Health Project [95] is definitely representative of this. During the year after discharge from a first admission, 63 % of the patients had one or more gaps, defined as “any discontinuation in the use of antipsychotic medication, whether initiated by the patient or by the physician”; gaps initiated by the patient, the physician or jointly occurred at rates of 73, 16 and 11 %, respectively. Similarly, in a group of 605 patients with first-episode psychosis, 33.7 % of the patients were classified as fully adherent, 47.4 % failed to take medications for at least one phase of 1 week, and 18.8 % persistently refused medication [97]. Some evidence also exists that non-adherence to medication regimens in the early course of schizophrenia could be even more frequent than in the chronic phases of the disorder. For example, in a systematic review of 83 studies [73], the weighted rate of non-adherence to medication regimens and scheduled appointments was 46.9 % in first-contact patients, 23.2 % in those already undergoing treatment, and 53.3 % in those with a history of low adherence. The presence among patients in the early stages of their psychosis of a very high and possible heightened susceptibility to poor antipsychotic medication-taking behaviour fits well with the Health Belief Model [64, 98], in which medication adherence is considered to be a dynamic process of “the patient’s beliefs about need for treatment and the benefits of treatment weighed against the negative aspects of treatment” [94]. Patients with schizophreniform disorder, first-episode schizophrenia or recent-onset schizophrenia are indeed “just beginning to come to terms with having a psychiatric disorder and have not been in treatment long enough to recognize the necessity of adhering to their medication regimen” [85].

Epidemiologic studies are inconclusive regarding the popular perception that people with schizophrenia are at special risk for poor medication adherence because of their psychopathology. The experimental evidence does not seem to substantiate this belief. For example, a subanalysis of the Canadian Community Health Survey database [99] centred on a sample of 6,201 individuals taking psychotropic drugs has indirectly indicated that different psychiatric conditions share a similar medication-taking behaviour; non-adherence to antipsychotics, sedative hypnotics, anxiolytics, mood stabilizers and antidepressants was found in 34.6, 34.7, 38.1, 44.9 and 45.9 %, respectively. However, the size of the group of patients taking antipsychotics, only 168 individuals, seems too small to ensure accurate estimates. In addition, the current literature on medication-taking behaviour suggests that there are large overlaps in the frequency of poor medication adherence among patients with schizophrenia and individuals with other medical or psychiatric conditions [40, 64]. However, a great deal of evidence on this issue comes from a patchwork of studies that, having been carried out independently, are at high risk for an unbalanced distribution of many sources of variation, particularly adherence measures. Furthermore, studies on people with schizophrenia carried out before the turn of the millennium have commonly used weaker measures of adherence than those used in research focused on subjects affected by a medical condition [40].

Similarity of compliance rates across therapeutic areas has been reported only when studies using electronic monitoring devices were reviewed [100]. Therefore, before concluding that the rates of poor adherence to prescribed medicines are similar across diseases and disorders, more direct evidence needs to be acquired. Head-to-head comparison between distinct clinical conditions is one of the most feasible strategies, even though this design is also charged with interpretative difficulties, such as the risk of spurious conclusions due to different diffusion of community care interventions among the various branches of medicine. In this regard, among a group of revolving door patients who met the RDC for schizophrenia, schizoaffective disorder, major depressive disorder or bipolar disorder, medication non-compliance was not found to be “associated more commonly with any of the four specific diagnostic categories” [101]. Intra-subject comparisons of adherence to different classes of medicines in people with comorbid disorders is a reasonable alternative approach that promises to offer superior methodological guarantees, but this experimental strategy has so far produced inconclusive evidence. Nevertheless, it seems of interest to cite a large-scale study of VA patients with schizophrenia who had diabetes and hypertension in comorbidity [102]; this study not only reported adjusted ORs of poor adherence “significantly higher for hypoglycemic and antihypertensive medications than for antipsychotic medication” but also significant associations “between the MPRs for each physical condition and patients’ antipsychotic medication”. This last finding suggests a generalised pattern of medication-taking behaviour that crosses different disease and drug classes. It seems plausible that medication adherence is also charged with supplementary illness-specific and/or drug-specific contributors because “information about antipsychotic adherence explained only 13 and 16 % of the variance in patients’ antihypertensive and hypoglycemic MPRs”[102]. In addition, a study of non-demented middle-aged and older VA outpatients with schizophrenia [103] confirmed that poor adherence was “equally problematic for both antipsychotic and nonpsychiatric medications” but failed to replicate the presence of the correlation between the 12-month cumulative mean gap ratio for antipsychotics and those for antihypertensives, antihyperlipidemics and antidiabetics.

Consequences of Poor Adherence to Antipsychotics

The interest in poor medication adherence resides not in the phenomenon itself but in its consequences. Different mechanisms may be involved in the transformation of non-adherence into sequelae of clinical relevance. Some mechanisms refer to reduced adherence, others to increased adherence.

In particular, in the presence of a status of subadherence, two main biological pathways may be advanced. One is more applicable to patients with occasional failure to take medicines as prescribed and assumes that, when the magnitude of the missed therapy is enough to settle receptor occupancy by antipsychotics at a level below the therapeutic threshold, this opens the doors to the re-emergence or recrudescence of psychosis [104, 105]. The other pathway is more relevant for patients with long-lasting poor adherence problems and involves the well-known assumption [106] that compensatory receptor supersensitivity induced by chronic administration of antipsychotics makes receptors prone to over react when an abrupt discontinuation of the therapy occurs. Under these circumstances, the withdrawal is likely to have psychotogen potential in itself, as indicated by the trigger of a rapid-onset supersensitivity psychosis. The old claim [107] about the need to escalate antipsychotic doses concomitant with an acute relapse seems compatible with a supersensitivity model of psychosis. Subtherapeutic receptor occupancy and supersensitivity psychosis have slim boundaries and are susceptible to reciprocal transition.

The presence of an over adherence condition has three main possible pathways. The first suggests that the practice, for any reason, of taking medicines more than prescribed may lack clinical interest because of a neutral impact on the course of schizophrenia. The second postulates that patients not completely satisfied with their therapies may decide to ingest a relative excess of antipsychotics in an attempt to self-medicate, which, when unsuccessful, may lead to treatment cessation, and thus to non-persistence. The third proposes that, when patients are over adherent, they inevitably become more vulnerable to the iatrogenic health effects of antipsychotics, and thus switch to under adherence, with a consequent increased risk for both a recrudescence of schizophrenia and a supersensitivity psychosis.

However, the recent report [108] of an association between medication adherence and superior increase in frontal lobe intracortical myelin volume after risperidone therapy also supports the hypothesis that clinical correlates of poor medication-taking behaviour depend at least in part on a failure to promote white matter development in individuals affected by a deficit of the normal myelinisation trajectory.

Whatever the underlying mechanisms involved, failure to follow antipsychotic therapy as prescribed activates a cascade of negative effects in people with schizophrenia (Fig. 4). This dramatic conclusion could be even worse considering that untreated schizophrenia has been reported to exert a neurotoxic effect per se [109].

Fig. 4

Cascade of consequences due to poor adherence to antipsychotics in people with schizophrenia

The strength and consistency of the effects played by poor medication adherence vary in relation to the specific consequence that is under consideration. Furthermore, most studies have a cross-sectional design that makes it impossible to specify which comes first when a chicken-and-egg situation related to a bidirectional interrelationship exists. Unfortunately, a number of associations between poor antipsychotic medication-taking behaviour and schizophrenia are at least potentially bidirectional, because poor adherence may worsen the disorder and definite features of the disorder may promote poor adherence. Furthermore, the two factors in the relationship may influence each other reciprocally. To conclude that poor medication adherence acts as a risk factor for a definite schizophrenia-related variable, longitudinal studies are certainly preferable because they could potentially settle the chronologic sequence of the two components of the association. However, longitudinal studies also frequently have the inherent limitation that interrelatedness between adherence and clinical variables does not necessarily imply an exclusive causal relationship.

Clinical Consequences

A great deal of evidence accumulated over the years emphasizes that the failure to follow antipsychotic therapy as prescribed interferes with symptom severity, short-term and long-term outcome and prognosis, degree of autonomy and functioning in daily life, the presence of comorbidities and the risk for violence, illegal acts, suicidal behaviour and, possibly, death in general (Fig. 5).

Fig. 5

Clinical and behavioural consequences of poor adherence to antipsychotics in people with schizophrenia

Symptom Severity

Despite some negative findings, comparisons with individuals who follow doctor’s prescriptions strongly support the conclusion that patients with schizophrenia who are poorly adherent continue to be afflicted by a more severe symptomatology [51, 57, 89, 93, 110–118]. Some specific domains of psychopathology, the positive cluster in particular and even some individual symptoms, such as conceptual disorganisation, lack of insight, poor attention and stereotyped thinking, are likely to be responsible for this unfavourable situation.

The impact of medication adherence on the clinical picture is likely to have appreciable relevance, because a regression analysis based on data from a 1-year naturalistic study [116] predicted an increase in PANSS total score of 3.1 points for each 20 % drop in treatment compliance.

Clinical Response

The influence of poor medication adherence on clinical response of patients with schizophrenia treated with antipsychotics have undergone experimental evaluation. A post hoc analysis of an 8-week trial [117] has shown how any additional day of non-adherence “reduced the likelihood of achieving response at study end by 6 %”.

Moving from clinical response to remission, a German prospective, randomised, observational, 2-year follow-up trial of 2,960 patients with schizophrenia [119] seems particularly representative. The study reported that, compared with patients who complied with antipsychotics, those who did not comply, 36.5 % of the total sample, had an OR of 0.73 for achieving symptomatic remission, defined as “receiving a CGI-Schizophrenia severity score of absent to mild in assessments of overall severity, and positive, negative, and cognitive subscores” [119]. In contrast, no relationship has been reported between non-compliance and functional remission, defined as a positive occupational/vocational status. The discrepancy between symptomatic and functional remission may be explained by the fact that the latter implies possibilities of access to opportunities largely independent of how much a patient follows the prescriptions of the treating physician. Medication adherence has also been reported to exert a negative influence on the chances for and time to remission of first-episode populations [120].

Another study [121] failed to demonstrate, over a 5-year period, an association between the percentage of time spent in taking antipsychotics and symptom remission. However, this last negative finding was extracted from a sample population in which “long-term medication adherence was very high, since subjects usually resumed medication following staff interventions or the return of symptoms” [121]. Therefore, the possibilities for generalisation of the results seem questionable.

Relapses

The association between poor medication adherence and psychotic relapses seems especially solid [58, 64, 85, 121–132]. In particular, in a group of first-episode patients with schizophrenia, schizophreniform disorder or schizoaffective disorder recruited in Hong Kong and followed for 3 years, subjects “taking less than 70 % of prescribed medication” had a 57 % cumulative relapse rate, a much higher value than the 36 % found in patients with good adherence [124]. When, in the same study, only patients with schizophrenia were considered, the influence of poor medication adherence on relapse risk remained substantially unchanged: at the end of the first, second and third year of follow-up, the cumulative relapse rates were 20, 31 and 37 for subjects with good medication adherence and 36, 64 and 64 % for those with poor adherence. In a logistic regression model applied to the same dataset, non-adherence was found to be an appreciable predictor of relapse, with an OR greater than seven. Another 5-year follow-up study of patients with a first-episode of schizophrenia or schizoaffective disorder who responded to initial therapy with antipsychotics [123] reported similar results; treatment discontinuation against the advice of the clinician implied an HR of 4.57 for a second relapse in the group of subjects who, after an initial relapse, were prescribed to continue antipsychotic medication for the remainder of the trial.

Although preferentially supported by studies of patients with first-episode schizophrenia, the negative influence of poor adherence to antipsychotic medication on relapse risk is a phenomenon that is commonly present at all stages of the disorder. For example, in a review [64] of seven independent studies, it has been reported that schizophrenia “patients rated as noncompliant have a 6-month to 2-year risk of relapse that is an average of 3.7 times greater than patients rated as compliant”. In addition, a logistic analysis based on data drawn from the US-SCAP study [126] has reported an OR of 1.79 for relapses in patients presenting a non-adherence status. Furthermore, it has been estimated that “at the 1 year point, approximately 68 % [of relapses] is owing to loss of neuroleptic efficacy and approximately 32 % is owing to neuroleptic noncompliance” [122].

A brief period of medication non-adherence is probably enough to induce a relapse. A study [85] of patients with recent-onset schizophrenia or schizoaffective disorder treated with risperidone reported an HR of 5.8 for relapses after mild non-adherence, defined as compliance of 50–75 % of the prescribed medication for at least two consecutive weeks during the follow-up period. Similarly, among patients with first-episode schizophrenia followed for 1-year after discharge from hospital, more subjects classified as non-compliant, because they “had used less medication than prescribed or completely skipped … medication for ten consecutive days” were represented in the relapsed group than in the non-relapsed group (70 and 25 %, respectively) [125]. Underlying the statement that “there is very little leeway for brief gaps in oral antipsychotic medication used or dosage reductions” [85], the last two studies highlight the need for assiduous and vigorous interventions aimed at improving adherence, even when only brief deviations from the prescribed antipsychotic regimen are suspected.

Given the unfavourable impact on the risk of relapses and their duration, poor adherence to antipsychotics could also lead to long-term secondary effects associated with recurrences such as the loss of responsiveness to these agents and the consequent needs to increase the doses concomitant with increased frequency and duration of psychotic breakdowns [133–135].

Early Discontinuation

Some evidence exists that medication adherence influences the probability of early discontinuation from therapeutic programmes. For example, in an 18-month study of 99 patients with schizophrenia followed by the University Psychiatric Services of Brescia Spedali Civili, it has been observed that, compared with adherent patients, those classified as moderately or poorly adherent had relative risks of premature discontinuation for any reason that were 2.5 and 5.6 times higher, respectively [136]. Furthermore, according to a secondary analysis of a 52-week, randomised, double-blind, flexible-dose trial centred on the effectiveness of olanzapine, quetiapine and risperidone in patients with a first-episode of schizophrenia, schizoaffective disorder or schizophreniform disorder, “each point improvement on the medication adherence rating scale resulted in almost a 30 % reduction in the hazard of treatment discontinuation” [137]. The impact of medication-taking behaviour on rates of discontinuation does not seem restricted to long-term therapies, because the association has also been fully documented in an 8-week trial [117]. However, compliance scores for patients with first-episode schizophrenia enrolled in the EUFEST did not differ in relation to the discontinued–nondiscontinued dichotomy [138].

Wellbeing and Functioning

Some findings suggest that the negative effects of poor adherence to antipsychotics may extend to wellbeing and functioning, social relations and activities of daily living in particular [59, 89, 119, 121, 139–141]. For example, non-compliance has been associated [119] with an OR of 0.73 for subjective wellbeing. Compared with non-adherent subjects, adherent patients have also been reported to present “significant and sustained improvements over the following 2 years in mental functioning, satisfaction with social life, satisfaction with basic needs, and general life satisfaction” [89]. Furthermore, patients who have improved compliance have been shown to improve their total score by 45.7 % for the SWN-k, a percentage definitely better than the 27.9 and 17.9 % found in subjects whose compliance was unchanged or worsened [59]. Overall, it seems reasonable to hypothesize that poor medication-taking behaviour contributes considerably to the WHO’s listing of schizophrenia as the seventh leading cause of DALYs worldwide [142].

However, it cannot be dismissed that “evidence for interrelatedness between the course of compliance and subjective wellbeing” does not explain which variable has a causal effect and does not exclude a reciprocal influence [59].

Substance Use Disorders

Non-adherence to antipsychotic medication has been associated with substance use and greater severity of alcohol-related problems [51, 89, 90, 101]. The observation [143] that, when compared with non-adherent non-depot initiators, non-adherent depot initiators are at higher risk for substance abuse before starting with depots gives further indirect weight to the association.

Despite the convergent support of a sufficient amount of evidence, the possibility of quantifying with precision the impact of medication-taking behaviour on the risk for substance use and related disorders remains arduous, because the two terms of the association influence each other reciprocally.

Suicidal Behaviour, Violence and Crimes

Demonstrations of links between poor adherence to antipsychotics and suicidal behaviour, violence and crime in people with schizophrenia are robust.

With regard to suicidal behaviour, a meta-analysis of 29 case–control or cohort studies concluded that poor adherence to antipsychotics more than triples the suicide risk [144], thus confirming a 1984 report [145]. In addition, a large Canadian retrospective study reported a reduced risk of attempted and successful suicides among patients from Quebec with good compliance [82]. However, the finding has not been replicated among the residents of another province, Saskatchewan. The discrepancy in the results between the two provinces is plausibly due to the relatively low base rate of suicidal behaviour, which requires large samples to detect reliable associations. This prerequisite was satisfied much better in the Quebec sample (41,754 patients) than in the Saskatchewan group (3,291 patients). The link between suicidal attempts and poor adherence to antipsychotics has also been considered in other studies. For example, in an analysis of drug-dispensing records carried out in the Netherlands and involving a sample of 603 exclusive users of olanzapine or risperidone who were “suspected to suffer from schizophrenia”, the adjusted relative risk for suicide attempts leading to hospitalisation “among patients with drug holidays was increased four-fold compared to patients without drug holidays” [146]. A higher risk among non-adherent patients also emerged in the SOHO study, which followed 6,731 outpatients with schizophrenia for 3 years [127]. Furthermore, in a sample of 36,195 California Medicaid subjects followed prospectively for 1 year due to ICD-9 schizophrenia [147], adherent patients presented lower rates of current and past suicide attempts compared with both non-adherent and partially adherent patients.

Although violence is only marginally associated with schizophrenia [148–150], poor adherence to antipsychotics was reasonably associated with increased proneness to this behaviour [51, 89, 151–158]. In particular, in a numerically representative sample of patients enrolled in the US-SCAP study and followed for a 3-year period [89], individuals with poor adherence to antipsychotics were reported to be more frequently violent, arrested or victims of crimes.

Furthermore, a non-adherent status over the first year has been found to lead to a 2.2 and 1.8 times increased risk for being arrested or victim of a crime in the successive 2 years, respectively. Data related to homicides seem to follow the same trend. According to a Swedish national case–control study of patients with schizophrenia or other psychoses [157], it was observed that, compared with compliant patients, non-compliant subjects had an almost quadruple risk for homicide within the first 6 months after hospital discharge. Reports referring to the impact of improved medication-taking behaviour on violence risk constitute a valid countercheck of the strength of this association. In a trial involving patients with schizophrenia or related disorders who received services in the North Carolina public sector mental health system and were followed for a 3-year period [152], “compliance with prescribed medication ‘most of the time’ or ‘all of the time’ was significantly associated with reduced violence” and olanzapine was found to be superior to risperidone for the control of violence, largely as a consequence of a more pronounced positive effect on adherence. Another study by the same group [153] reached similar conclusions; an inverse correlation between medication-taking behaviour and violence was observed, and novel antipsychotics were shown to control violence better than conventional neuroleptics. This superiority was attributable to the presence of a cumulative effect between medication and compliance that was present with novel agents but not conventional agents. The trial also highlighted that long-lasting compliance with second-generation antipsychotics made the risk for violent behaviour almost negligible. The clinical potential of this finding, if confirmed, is obvious. Reports [151, 159] that involuntary outpatient commitment and administration of depot antipsychotics can reduce violent behaviour in people with severe mental illness are a further indication of the central role played by medication-taking behaviour on the risk for violence.

In general, the association of poor medication adherence with violence, assaultive behaviour and other crimes appear trans-diagnostic in nature, because it has also been found in patients affected by severe mental disorders other than schizophrenia [35, 149, 154, 155, 158–161].

In addition, evidence derived from the ECA study [148] suggests that the impact of medication adherence on violence is especially pervasive in the presence of a comorbid substance abuse disorder, as indicated by the observation that a mood disorder, schizophrenia or substance abuse were each associated with 3. 5, 8.4 and 21.3 % of subjects having violent behaviour, whereas the co-diagnosis of a mood disorder or schizophrenia with substance abuse increased the percentage to 29.19 and 30.30 %, respectively.

Distinct, so far unidentified, characteristics associated with problematic medication-taking behaviour may increase the risk of imprisonment. In a selected population of poorly adherent subjects, patients judged worthy of initiating a long-acting first-generation antipsychotic were reported to have an almost 4-fold higher probability of being arrested or jailed in the preceding 6 months compared with non-depot initiators [143].

Mortality

Evidence of an association between non-adherence to antipsychotic medication and death for any reason also exists. In particular, in a retrospective large-scale study of more than 40,000 patients with schizophrenia, good compliance status was found to be associated with HRs for death equal to 0.65 and 0.58 in the short term and long term, respectively [82]. This result seems to be reconcilable with the healthy adherence effect hypothesis [1] and contributes to explain the common observation that people with schizophrenia have a reduced life expectancy not only due to unnatural causes but also due to natural causes [162–173].

False Non-Responders

When unrecognised, poor medication adherence aggravates the course of schizophrenia and promotes misdiagnoses and erroneous classification as non-responders among patients who do not respond because they fail to take medicines as prescribed. “The physician bases critical decisions in management and diagnosis on the patient’s response to a therapeutic regimen. If he observes no response to a usually effective regimen, he may decide to change the treatment or he may even question his diagnosis” [69]. Common consequences of these misleading processes are increased doses of the medication to near or over the maximum permitted, the introduction or implementation of polypharmacy therapies and anticipation of switching from one antipsychotic to another. Data on 7,864 patients with schizophrenia or bipolar disorder who entered a Southeastern Medicaid programme [174] demonstrate this association. Compared with compliant subjects, partially compliant patients were shown to be 64 % more likely to switch from one antipsychotic to another or to increase therapy. Switching and augmentation strategies can lead to appreciable hazards; therefore their unjustified use should be avoided and judged reprehensible.

Societal and Health Care System Consequences

The different negative clinical consequences sustained by poor medication adherence reverberate inevitably on the health care system, the patients and their families and wider society (Fig. 6).

Fig. 6

Impact of poor adherence to antipsychotics on health care and welfare system, research, family and wide society

Health Care Services

With regard to the consequences for health care services, a large preponderance of the literature supports the concept that poor adherence to antipsychotic drug regimens is an appreciable risk factor for frequent hospitalisation [51, 58, 60, 81–83, 89, 93, 111, 115, 127, 130, 131, 175–186]. A realistic snapshot of the dimension of the association may be offered by three small studies. In a group of revolving door patients with schizophrenia admitted to short-stay urban psychiatric units in New York, non-compliance was regarded as the most common suspected cause of relapse leading to new hospital admissions and was classified as the primary cause in 50 % of cases [178]. In addition, a study involving a diagnostically mixed group of patients reported a sharp contrast between the 6 % rate of non-compliant patients who incurred one hospitalisation and the 29, 30 and 34 % found among patients with 2–4, 5–10 or more than 10 hospitalisations, respectively [101]. In parallel, a case–control study [176] of seriously ill patients with schizophrenia followed by community mental health centres in Mississippi reported that medication non-compliance was associated with the highest adjusted OR for rehospitalisation, [8. 18].

However, full validation of the strong relationship that exists between adherence to antipsychotics and the risk for psychiatric hospitalisation comes from some large-scale studies based on continuum or quasi-continuum measurements of adherence. In particular, in a large sample of VA patients with schizophrenia or schizoaffective disorder who “had an outpatient prescription for an oral antipsychotic medication between October 1, 1998 and September 30, 1999”, the link between the two variables was well expressed by a U-shaped curve: a progressive decline in the rate of psychiatric admissions occurred as the patients’ MPR approached one and a successive increase in rehospitalisations emerged when patients presented excess medication fills [179]. In particular, subjects with poor adherence or excess medication fill were 2.4 and 3.0 “times as likely be admitted during the study year than patients with good adherence”. A quasi carbon copy analysis based on Medicaid beneficiaries with schizophrenia followed in San Diego County [81] confirmed that individuals classified as non-adherent, partially adherent, adherent or excessive fillers differed in the frequency of psychiatric hospitalisations: 34.9, 24.1, 13.5 and 24.8 %, respectively. Hospitalisations due to medical conditions were also affected by antipsychotic adherence, because “individuals who were nonadherent or excess fillers were about 70 % more likely to be hospitalised for medical care, and those who were partially adherent were about 30 % more likely to be hospitalised than those who were adherent” [81]. A 2-year prolongation of the period used for the inclusion of the patients [187] confirmed the association of adherence with the rates of both psychiatric and non-psychiatric hospital admissions. A Canadian study based on bipartition of patients into compliant and non-compliant individuals further consolidated the presence of an association between non-compliance and increased hospitalisation rates due to psychosis or medical conditions [82]. Specifically, patients with good adherence in the first year after the index prescription of an atypical antipsychotic had an all-cause hospitalisation adjusted HR of 0.65 in Quebec and 0.80 in Saskatchewan. The relationship between poor antipsychotic-taking behaviour and higher probability of hospitalisation in general [82] is congruent with reports of both increased any-cause mortality and relative excess of medical conditions that distinguish people with schizophrenia from the general population [162, 173, 188–194]. With regard to the mechanisms responsible for the link, it seems plausible to assume that adherence to antipsychotics may act as a marker of individual propensity to follow prescriptions from a physician in general and/or engage in correct healthy behaviour.

Poor medication adherence has also been reported in association with compulsory treatment and involuntary commitment [93, 115, 195–198]. A 2-year follow-up study of subjects with schizophrenia or other psychotic disorder consecutively admitted for the first time to the acute wards of Bordeaux psychiatric hospital reported a three times greater risk of compulsory readmissions in patients with poor medication adherence, defined as complete discontinuation of psychotropic medication against medical advice for at least 2 weeks over a 6-month interval [93]. Abnormally, high rates of compulsory treatments among poorly adherent patients with schizophrenia are likely to be at least partially mediated by the common association between this medication-taking behaviour and substance use, another well-established cause of compulsory treatments.

Overall, the clinical impact of adherence to antipsychotics on the risk of new hospitalisations is relevant; a regression analysis based on more than 4,000 patients with schizophrenia has shown how the odds for hospitalisation were lowered by a factor of 23 % in the presence of a 10 % improvement in MPR [60].

The effect of medication non-adherence on the risk of hospital admissions is likely direct, because a logistic regression analysis demonstrated that, even after controlling for others factors, “irregular medication use remained a significant and persistent predictor of rehospitalization”, as indicated by an adjusted OR of 1.99 [180].

Furthermore, the association between poor medication adherence and increased vulnerability to the revolving door phenomenon or, more broadly, rehospitalisation, seems to extend across different diagnostic categories [38, 101, 199–203].

With regard to the interval between the emergence of poor medication adherence and a new hospitalisation, it seems plausible to hypothesize a short time period [58, 60, 83, 90]. In particular, a cohort study of patients with schizophrenia in Maine and New Hampshire Medicaid who started therapy with second-generation antipsychotics supports this suggestion; the participants with a disruption in medication adherence had HRs of 1.54 and 1.77 for hospitalisation in the first 10 days of the gap in medication use, according to whether the admission was due to problems with general mental health or schizophrenia, respectively [83]. Gaps longer than 30 days have been shown to increase all-cause hospitalisations, that is, admissions related to medical or psychiatric conditions, with an HR of 1.57. Similarly, in a study of 4,325 California Medicaid patients with schizophrenia placed “on the more compliant end of the compliance continuum”, the group with a maximum medication gap defined “as small as one to ten days in a one-year period” had almost doubled odds of mental health hospitalisation compared with the population who did not have gaps in medication therapy [60]. Furthermore, an Australian study based on consecutive hospitalisations or interventions from a 24-h community-based crisis team confirmed that a relevant number of readmissions among medication non-compliant patients occurred within the first 2 months after discharge [90].

At least two orders of indirect evidence further corroborate the conclusion that even a few days of poor medication adherence are sufficient to promote a new hospital admission. The re-emergence of psychotic symptoms after discontinuation of treatment with antipsychotics for any reason has been reported to occur in almost 50 % of cases within the first 2 months [25, 204] and the immediate postdischarge period has been associated with heightened risk for non-compliance [84, 85, 205, 206]. If it is true that very brief gaps in medication adherence may be sufficient to cause a rehospitalisation, it is also true [60] that the more the maximum gap in medication adherence increases, the more rehospitalisations occur. Therefore, the impact of poor adherence to antipsychotic medication on the risk of new hospital admissions seems regulated by a typical dose–effect relationship.

However, the negative influence of poor adherence to antipsychotics on hospitalisations is not restricted to an increase in the number of readmissions. It also involves longer duration of hospital stay [90, 115, 175, 179, 180, 184, 197, 207]. Gaps in the length of stay in hospital between adherent and non-adherent patients cannot be generalised. The phenomenon is highly dependent on the characteristics of the care system involved. The influence of medication adherence on the number of days spent in hospital is direct, because poor adherence remains the second best predictor of the length of rehospitalisation after controlling for other adherence factors [180]. From a quantitative perspective, the impact of adherence on the length of hospitalisation is likely to be not trivial, because a series of multivariate regressions performed with Medi-Cal data on 35,815 patients with schizophrenia showed that “the fraction of inpatient days attributable to not receiving antipsychotic medications was 13.1 %” [207].

As in the case of compulsory hospitalisations, substance abuse is a supplementary factor in the association between non-compliant status and increased risk for repeated readmissions and more occupied bed days [90]. For example, in patients with schizophrenia, schizoaffective disorder or schizophreniform disorder living within the Central Sydney catchment area and followed for 4 years, non-compliant individuals who abused substances had a 2-month interval between hospital readmissions, a value not far from the 4 months found among non-compliant subjects who did not abuse substances but 4.5 and 14 times shorter than the intervals found for compliant patients who abused or did not abuse substances, respectively [90].

Adherent and non-adherent patients have also been reported to be different in relation to the use of other health care services. In particular, non-adherent individuals have been found to be higher users of emergency psychiatric services and acute wards [89, 90, 111], lower users of long-term rehabilitative beds [90] and have an increased rate of drop out from mental health clinics and day hospitals [111]. However, replication studies are needed for most of these issues.

Research

Poor adherence to antipsychotics may have unfavourable consequences on schizophrenia research. When present but not adequately controlled, scarce medication-taking behaviour facilitates spurious results in clinical trials; this may produce a delay in registration and regulatory procedures, an unjustified premature cessation of research and development of new agents, a decrease in the incidence of adverse events with eventual preclusion of early recognition of important warnings, and inaccurate delimitation of the therapeutic dose range. Furthermore, interpersonal variability in adherence is a confounder in studies on markers and correlates of efficacy and tolerability of antipsychotic medications and weaknesses the power of statistical analyses, forcing the recruitment of larger sample populations.

Despite this long list, much of the current research includes only a rough measure of adherence: the pill count. This decision is based largely on a false perspective: the confidence that trial participants must be substantially adherent to the established medication regimen, because they have given their informed consent to enter the study.

Family and Significant Others

Poor medication adherence promotes several negative consequences for the people who are close to the patient. Family members and significant others face an increased risk for violence and other illegal acts related to a psychotic recrudescence of the patient who has partially or totally discontinued antipsychotic medication. Individuals within the families of patients with schizophrenia who are poorly adherent are also frequently judged to be responsible for insufficient surveillance and, for this reason, they too are subjected to stigma. This blame for insufficient involvement is in sharp contrast with the evidence: the lives of the people close to poorly adherent patients are frequently beset by excess worry, distress and other negative emotional feelings, resulting in over involvement with supportive assistance and caregiving, and possible withdrawal from social contacts and employment difficulties.

Wider Society

The clinical and behavioural effects of poor antipsychotic medication-taking behaviour in people with schizophrenia inevitably reverberate on society. This occurs through two main ways: the reinforcement of stigma attached to the disorder and its pharmacologic treatment and the increased susceptibility of the lay public to become victims of aggressive assault and other crimes committed by patients when they develop poor adherence.

Economic Consequences

The dramatic burden played by poor adherence to antipsychotic medication on patients, families, health care and justice systems, the community and wider society inevitably leads to substantial tangible and intangible costs. Despite the universality of this picture, extrapolation of costs outside the original context seems difficult to generalize, especially for absolute values of expenditure. A number of considerations contribute to this conclusion. For example, different countries guarantee different standards of care and supply services that are not comparable. Consequently, direct costs change on a nationwide basis according to the specific profile of the particular health care system. National effects also clearly operate in welfare, social assistance and justice systems. These national policies affect the economic burden for the patients and their families. Furthermore, expenditure can vary for some items over time. Changes in the price of medicines in relation to new more expensive drugs and substitution of older brand products with cheaper generics, the ongoing trend to reinforce community interventions and short-term hospital stays, and the pressure to guarantee new welfare and assistance standards are examples of how costs directly or indirectly attributable to schizophrenia can vary within this ever changing situation. These limitations inherent to the costs of schizophrenia also apply to the economic consequences of poor adherence to antipsychotic medications.

The literature on the cost of schizophrenia in general and poor adherence to antipsychotics in particular is frequently charged by another limitation: much of the evidence is derived from models that are highly conditioned by the set of postulates, assumptions and definitions used by different studies. For example, a 1995 study [122] found that, at the 1-year point, the relative contribution of non-compliance to rehospitalisation of “neuroleptic responsive, multi-episode … schizophrenia inpatients … discharged back to outpatient treatment” was approximately 32 %. Thirteen years later, another study [207] carried out by one of the authors of the previous report concluded that the annual “fraction of acute care inpatient admissions attributable to not receiving antipsychotic medications was 12.3 %” in subjects “who had received at least two outpatient or one inpatient claim for schizophrenia”. Effects related to the time frame between the two studies may have contributed to the almost 20 % difference in the estimate of the rehospitalisation rate due to poor medication adherence. Nevertheless, it seems unrealistic to assume that the interval between the two studies was the sole reason for the discrepancy in the results.

Other distinguishing features related to the design of the study must be hypothesised as being at least equally and possibly more important: differences in the sources of data collection, the clinical profile of the sample of reference, the definition of medication adherence, adjustments for the patients’ background characteristics, the number and type of prerequisites and key assumptions to be satisfied, the levels of extrapolation required and the specific equations used should be taken into account. Key assumptions, in particular, are a crucial factor in determining the results of economic simulation modelling studies because of their reliance on expert opinion rather than real practice. For example, by changing the scenarios within the simulation model, paroxetine has been reported to be more, equally, or less cost-effective than imipramine [208]. There is no reason to assume that a similar variability in the results does not apply in the case of economic comparisons of different antipsychotics or the adherence/non-adherence dichotomy.

Within the limits specifically imposed by these critical points, a number of general principles may be taken from the current literature [81, 122, 126, 174, 180, 207, 209–214].

First, “a definitive relationship exists between compliance and the economic costs of schizophrenia. Lower rates of compliance lead to higher costs of treating schizophrenia” [209].

Second, many relevant costs of illness cannot be easily quantified. Examples include the costs related to caregiving services provided by families, comorbid medical conditions, specific training and dedicated research. Therefore, the direct and indirect costs of schizophrenia, especially the latter, are systematically underestimated.

Third, because poor adherence to antipsychotic medication is a widespread phenomenon among people with schizophrenia and the direct medical costs per capita associated with this behaviour are relevant, it is not surprising that schizophrenia consumes a substantial share of medical expenditures [210, 213]. Despite its low prevalence, schizophrenia requires a large fraction of the funds allocated for the management of all mental illnesses in many health care systems [211, 213, 215–217]. The typical onset of schizophrenia in early adulthood and its lifelong and deteriorating course justify this burden. Considering that poor medication adherence apparently occurs at similar rates among patients affected by different mental disorders, its seems hard to qualify the relative redundancy of the direct costs of schizophrenia as a mere end product of the difficulties in taking antipsychotics according to the prescription of the treating physician.

Fourth, among the direct costs, the disproportionate consumption of resources accompanying poor medication adherence refers not only to costs related to psychiatric care but also to medical expenditure in general and involves the range of inpatient and outpatient services offered to individuals affected by schizophrenia. Therefore, interventions aimed at reducing the impact of poor adherence on the direct costs of schizophrenia require a global approach. Otherwise, the risk that decreases in expenditure for one item is counterbalanced by increase in others. Within this costly scenario, few doubts exist that “inpatient care constitutes the greater portion of direct medical costs for persons with severe mental illness” [180]. To get an immediate idea of the impact of medication-taking behaviour on the use of inpatient services, it may be enough to cite that “patients who failed to adhere to their medication regimen were over one-and-a-half times as likely as patients who did adhere to it to report use of in-patient services” [213]. However, non-adherence has also been shown to be “one of the most significant factors in increasing external service costs, by a factor of almost three” [213].

Fifth, inpatient expenditure related to poor medication adherence commonly identifies rehospitalisation as the most costly contributor. This persists even though continuous efforts are being made to externalize psychiatric care as much as possible. That rehospitalisation is a key cost-related item of inpatient services is well documented. For example, a study of schizophrenia Medicaid beneficiaries followed by San Diego County Adult Mental Health Services has found that “the hospital expenditures of those who were nonadherent were more than three times higher than the hospital expenditures of those who were adherent. The costs of those who were partially adherent or who had excess medication fills were about two and one-half times higher than the costs of those who were adherent” [81]. In a study carried out in Wisconsin involving patients with a severe form of schizophrenia or schizoaffective disorder, the hospital costs over a 12-month period were almost doubled among irregular medication users compared with regular users, US$3,421 and 1,799, respectively [180]. Therefore, the estimate [122] that 37 % of rehospitalisation costs incurred over 2 years by patients with schizophrenia were attributable to non-compliance seems plausible or even optimistic because, based on an incorrect a priori assumption, only the first rehospitalisation was considered, despite the fact that, in 1 year, many patients require multiple rehospitalisations in real-world practice [218].

Sixth, poor adherence to antipsychotics by patients with schizophrenia increases the likelihood of the need for external services [89, 90, 111, 126, 213] to the point that it has been reported to be the most significant factor responsible for increased direct costs unrelated to a stay in hospital [213].

Seventh, as emphasised earlier, many of the items typically included under the heading of indirect costs are unequivocally worsened by a poor medication adherence status. “Day-to-day care and support (of people with schizophrenia) is left to a great extent to family and friends, even if the patient does not reside with them” [219]. Furthermore, poorly adherent patients with schizophrenia present worsened functioning in almost all areas of daily living, have reduced life expectancy, develop a relevant loss of productivity due to disability, absenteeism, dismissal or unemployment, require increased use of welfare and assistance and incur justice problems more frequently. Despite this strong body of evidence, the issue of solid expenditure attributable to indirect costs related to poor adherence to antipsychotics continues to be substantially unanswered, even though indirect costs are a major fraction of the total economic burden of schizophrenia [215, 216, 220]. There is no reason to assume that this trend will be reversed for the items related to poor medication adherence. It seems reasonable to assume that indirect costs related to insufficient medication-taking behaviour may at least match the direct costs. The incontrovertible evidence that schizophrenia “is most prevalent during highly productive periods” of the vital cycle [215] makes these considerations easy to understand, and indirect costs are also at special risk for underestimation.

Eighth, direct and indirect costs attributable to violence, a possible secondary effect of non-adherence to antipsychotics, deserve separate consideration. Health care expenditure for offenders is not marginal because prisons are full of patients who suffer from schizophrenia or other severe mental illnesses, and have a dual diagnosis of substance use disorder. This is an objective reality not necessarily related to the deinstitutionalisation process that could have fostered the “shift of persons with serious mental disorders from hospitals to criminal justice settings, a form of trans-institutionalization” [158]. Violent acts precipitated by non-adherence introduce patients into the justice system, and consequently a number of specific and expensive costs are incurred, such as police, civil and criminal courts, private and public defence lawyers, jail and prison. Furthermore, because “the potential for violence increases public fear, (and) prevents acceptance and inclusion of persons with psychiatric disabilities in society” [153], it follows that poor adherence to antipsychotics may lead to supplementary stigma. The economic consequences of this powerful reason for stigma remain in limbo as intangible costs.

Determinants and Moderators of Adherence to Antipsychotics

Non-adherence is often assumed a priori to be an irrational phenomenon [221]. Despite this premise, a comprehensive list of more than 200 variables as candidates related to medication adherence was drafted a quarter of century ago [222]. In addition, this highly diversified scenario has been progressively enriched over the years by supplementary items. Therefore, the term irrationality may extend beyond its literal meaning to explain the systematic, inevitable fiasco of attempts to reconcile medication-taking behaviour with single or few causal processes.

From a deterministic perspective, medication adherence must be regarded as a complex, multifactorial behaviour that expresses, at the phenotypic level, the result of a strong, dynamic flow of interrelations that occur systematically between numerous specific causal factors and several modifiable and unmodifiable superimposed moderators.

Both determinant and moderating factors are recognised as having multiple origins. Some are expressive of a preponderant involvement of the patient, the doctor, the medication or the system in general. Others are composite in nature in that they are simultaneously sustained by indissoluble interactions between variables of different origin. Furthermore, some variables apparently act as predictors of medication-taking behaviour, but they represent the sum of disparate contributors and do not have an appreciable direct effect on adherence.

Given these complexities, models devised to describe the mechanisms that govern the genesis of medication-taking behaviour (Fig. 7) are defective in origin since they represent an inevitable over simplification of what occurs in real life. Despite this, an analytical item-by-item discussion of different variables is useful for didactic purposes, provided by the reader always tries to reassemble the puzzle from a unitary perspective.

Fig. 7

Semplified, hypotetical model of contributors and moderators of adherence to antipsychotics in people with schizophrenia

System-Related Contributors

Many system factors are far from being neutral on medication adherence. Society and family constitute the areas of major interest, together with health care and welfare policies.

Wider Society

Culture and acculturation, religion, ethic norms and value orientations, general social climate, tolerance and acceptance for minorities and diversities, economic trends, quality of information supplied by the media and reinforcements from testimonials contribute worldwide to steer the prejudices of society. This also applies to the case of stigma against severe mental illnesses, schizophrenia in particular and the use of antipsychotics for their treatment [223–229].

Although expression of a manifest stigma has probably relented in recent years, the phenomenon of “not in my backyard” remains widespread. It is therefore easy to understand why patients with schizophrenia have indicated stigma as the principal [230] or one of the most common [196, 231–233] barriers to regular use of antipsychotics. However, in an Australian study [234], patients with schizophrenia were able to recognize stigma but failed to associate the phenomenon with difficulties in taking the medicines as prescribed.

“Psychiatric stigma may be especially severe in some non-Western communities because of the meagre expenditure on mental health care, limited access to medical information, unpopularity of the human rights discourse, prohibitive risk of disclosure of psychiatric treatment and the paucity of advocacy work” [235].

Because stigma is a dynamic condition that can be controlled with dedicated interventions, numerous international, national and local campaigns on destigmatisation have taken place in recent years. In general, the programmes have been based on the hypothesis that improved knowledge on the biological foundation of schizophrenia should promote assimilation of the disorder with any other brain disease and thus improve medication adherence. Despite some experimental support [228], the assumption that correct knowledge about schizophrenia can exert anti-stigma effects has been found to be reasonable but unrealistic in practice. Several follow-up studies [223–225, 236, 237] have documented that, after biologically oriented educational interventions, an increased number of people effectively embrace the concept of schizophrenia as a physical disease but they do not reduce personal attitudes to stigmatize the disorder and the use of antipsychotics. For a better understanding of this gap, it may be useful to consider some evidence from a number of surveys [224, 226, 227, 229, 238] centred on the relationship between public beliefs on the causes of schizophrenia and social distance from the people affected by it. In particular, most of these reports stressed that a low attitude to judging schizophrenia as a real brain disorder rather than a largely volitional behaviour is a weak key to reducing prejudices about schizophrenia and its treatment, because this belief represents only one of the multiple contributors to stigma. Lay public sectors are also likely to be more susceptible to the detrimental influence of other aspects of schizophrenia such as the dangers, use of substances, scarce productivity, unpredictability, impulsivity and perceived differences in communication modalities. In addition, discoveries on the biological correlates of schizophrenia, in particularly those related to molecular genetics and brain imaging, may sometimes be counterproductive because they erroneously lead to stigma. Translation from improved cultural disposition to concrete behaviour needs a long time.

Family and Significant Others

The influence of families and, more broadly, significant others on medication-taking behaviour of patients with schizophrenia is well acknowledged since the advent of neuroleptics. A 1963 report [31] emphasised, for example, that “patients whose drug taking was supervised by a member of the household were much more likely to take the drugs regularly”.

Several contributing factors may be involved in driving the relationship. The attitude of society towards schizophrenia and its treatment certainly play a crucial role. In general, the position of proximity to patients suggests that messages filtered by those close to the patient should be more penetrating. However, it must also be taken into account that the inevitable relationship with the affected person is associated with better acceptance of the disorder and the use of antipsychotic agents [229, 239]. This implies that the eventual stigmatising atmosphere of the community is more likely to be dampened rather than potentiated by significant others. Nevertheless, when individuals close to the patient become the object of societal preconceptions or are entrenched in coarsely obscurantist positions, they may be inclined to amplify or generate further stigma, thus contributing to the poor medication-taking behaviour of the patient.

Together with stigma and prejudices, distress, emotional distance, ambivalence and difficulties staying close to the affected person are common supplementary obstacles to reminding, supervising and persuading patients to take medications as prescribed. ([31, 81, 91, 111, 196, 231, 240–248]; for supplementary references, see [38, 47, 64, 98, 214]). Reports that direct supervision in residential treatment settings, living in supervised housing and consistent attendance at self-help meetings [249, 250] have a positive influence on medication adherence of people with schizophrenia may be seen as further indirect proof of the relevance that significant others assume in favour of correct medication-taking behaviour of patients living at home. The same applies to the demonstration that “multifamily group therapy specifically tailored to improve medication adherence is associated with improved outcome” [251].

Health Care and Welfare Policies

Funding and the rules of the system are established by health care policies. Health care policies pursued in different countries therefore have a strong effect on medication-taking behaviour. This influence is largely mediated by indirect effects related, in particular, to the minimum levels of guaranteed care, the intensity and quality of vocational training provided to members of the treatment team and the planning of educational campaigns addressed to patients, their families and the lay public. All these items affect relevant moderators of adherence behaviour, such as the competence of the personnel operating within the system, the patient–doctor relationship, shared decision making and stigma.

However, by imposing governance on accessibility to therapies, health care policies exert a direct driving force on a patient’s inclination to take medicines as prescribed. The pressing need for rigorous cost containment in response to escalation of the price of drugs and the concomitant funding shortfalls related to the worldwide economic crisis have promoted the adoption of a large series of ad hoc strategies that could potentially have a negative impact on medication-taking behaviour. Preferred drug lists, tighter formularies, priority use of generics, exclusion of some pharmacologic classes from reimbursement, medication algorithms, mandatory authorisation before reimbursement, limits to the number of prescriptions that may be filled without previous authorisation and forced cost-sharing belong to this list.

In general, restrictive policies that have resulted in cost savings have often been accompanied by unintended negative secondary effects, such as deterioration in medication adherence [252]. Because non-adherence is especially important for essential therapies and in the presence of chronic disabling illnesses [253, 254], schizophrenia and antipsychotics are ideal acid tests for challenging the direct impact of restrictive policies on medication adherence. Overall, the sequence that links cost-containment policies, particularly the need for a fee, with loss of medication adherence in patients with schizophrenia has received experimental support. For example, during the era of first-generation antipsychotics, there was a 15.4 % decrease in the use of this class of medications among New Hampshire Medicaid patients after the introduction of the limit for reimbursement to three prescriptions [255]. This decrease was even bigger, 21.2 %, for regular recipients, and after the discontinuation of the policy, the use of medication “rose to a level slightly above pre-cap rates”. The same study has also shown how, in New Jersey, a state that refused to lay down limits on drug reimbursement, patients did not change their level of medication use during the same period. Similar results have been reported more recently, after the advent of second-generation antipsychotics, in comparisons of data related to Mississippi, Indiana and Minnesota [256], three states characterised by remarkable differences in relation to cost-containment policies: Mississippi has imposed “a cap on the number of drugs reimbursable per enrollee at seven prescriptions per month with a prior authorisation requirement for monthly dispensations of more than five prescriptions, an increase in copayments for brand name medications from $1 to $3 per fill, a mandate to dispense generic medications when available, and a 34-day restriction on the days supply filled per prescription”; Indiana and Minnesota have adopted more lenient cost-containment strategies. The consequence for medication adherence has been that “patients in Mississippi were 4.9 % less compliant with antipsychotic treatments and experienced 20.5 % more 90 day antipsychotic treatment gaps” than patients in Indiana and Minnesota.

Patients harried by cost-containment interventions cut their therapies on the basis of disease-related and drug-related preferences. In a large-scale study of VA patients with schizophrenia [254], medication co-payment was found to be associated with an appreciable decline, nearly 25 %, in refills for psychotropic agents but not for other medicines.

Loss of medication adherence in response to the adoption of restrictive policies may be well reconciled with the hypothesis, typical of the Health Belief Model, that the need for a co-payment and difficulties with access to medicines may tip the balance in individuals with an already fragile medication-taking behaviour in favour of an excess of disadvantages over benefits. Furthermore, extra expenditure may sometimes be literally unaffordable, for example, in the case of most indigent patients. Reports indicating an association between poor adherence to antipsychotics and unemployment, financial obstacles, low occupational status, low parental class, homeless condition, lack of welfare or access to transfer payment add weight to this situation ([19, 79, 81, 93, 111, 147, 159, 187, 231, 257–260]; for supplementary references, see [64, 98, 214]). On the other hand, patients with schizophrenia have been reported [230] to perceive homelessness and lack of social support among the principal barriers to their adherence to antipsychotics.

Doctor-Related Contributors

Within the background defined by the levels of assistance and treatment contingently supplied by the health care system of reference, the specific competence and ability of the physician in the areas of psychopathology, general medicine, clinical psychopharmacology and maintenance of interpersonal relationships play an essential role in the patient’s inclination to follow medical advice.

Psychopathologic Expertise

Mistakes in psychiatric diagnosis and recognition of defined domains of psychopathology activate a cascade of events that reverberate negatively on medication adherence. Scarce psychopathologic expertise increases the probability of unsuccessful treatment and this, in turn, leads to unjustified use of polypharmacy and high medication doses, with the risk of an excess of adverse events and a drop in expectations towards psychopharmacotherapies. These factors enter a number of pathways leading to poor medication-taking behaviour.

Medical Expertise

People with schizophrenia have an excess of medical comorbidities and vulnerabilities linked, in particular, to unhealthy lifestyles, shared diathesis, psychopathologic traits facilitating denial or poor interest in physical health and care inequalities. In the absence of the necessary medical expertise, the treating physician risks exposing patients with schizophrenia to a disproportionate number of iatrogenic health effects directly related to the individual medical history, leading to probable induction of poorer medication-taking behaviour.

Psychopharmacologic Expertise

The expertise of the physician in clinical psychopharmacology interferes at multiple levels with adherence to antipsychotics. In particular, detailed knowledge of the efficacy, tolerability, pharmacokinetics and pharmacodynamics of the various antipsychotics is essential to tailor individualised pharmacotherapies to maximize the chances of improvement and minimize the risk of adverse events, two basic components for good adherence. Also the doctor’s competence in gathering information from the patient’s psychopharmacologic history is essential for individualised interventions.

Another important issue for medication-taking behaviour that requires the competence and attention of the physician refers to the use of polypharmacy, a phenomenon that is in use worldwide and is expanding [7, 80, 81, 261–278]. A large-scale study [274] on office-based psychiatry over the 10 years straddling the turn of the millennium has quantified the extent and evolution of this phenomenon. The trial showed that “visits with two or more medications increased from 42.6 % in 1996–1997 to 59.8 % in 2005–2006” and that during the same time period “visits with three or more medications increased from 16.9 % to 33.2 %”. The same study has also shown that a diagnosis of schizophrenia was associated with higher odds of receiving not only two or more antipsychotics but also combinations of agents of this class with antidepressants, mood stabilizers and sedative hypnotics. Increased pressure to search for greater and faster improvements, low levels of expertise in drug–drug interactions, the ability to recognize psychiatric comorbidities and renewed popularity of the dimensional approach to clinical psychopharmacology are some of the major factors supporting the use of polypharmacy in clinical routine. Polypharmacy is sometimes indicated. More commonly, however, patients are prescribed medication combinations in the absence not only of acceptable experimental support and specific endorsements but also despite common recommendations to use co-therapies as a last resort [272–274, 276–279] because, “while the evidence for added benefit of antipsychotic polypharmacy is limited, there is growing evidence regarding the increased adverse effects associated with such combinations” [274]. This call for caution is recommended based on meta-analytic evidence of the effects of polypharmacy on the efficacy and tolerability of antipsychotics [275, 280] and the probability of publication bias related to the confinement of efficacy data to studies carried out in China or other Asian countries [280]. Therefore, the statements [281] that “seldom can such polypharmacy be considered rational or scientific” and “on the contrary, it is indicative of a serious gap between a basic knowledge of drugs and their clinical use” seem fully relevant today.

Irrespective of whether polypharmacy is correct or not, it is a contributing factor for poor medication adherence, because it promotes dosing complexity and increases the incidence of adverse events. In confirmation of this, it may be enough to emphasize that in a representative sample of Florida Medicaid beneficiaries [210] “a status of limited to negligible adherence occurred at rates of 18.6, 28.4 and 37.8 % among patients treated with atypicals, typicals, or typical plus atypical, respectively”.

Relational Expertise

When the other skills and expertise of the physicians are equal, the ability to promote and maintain a trusting, collaborative patient–doctor relationship has an appreciable influence on adherence to antipsychotics, because this attitude represents a basic principle for therapeutic alliance and shared decisions.

Patient-Related Contributors

The group of mainly patient-related determinants and moderators of medication adherence include numerous factors, in particular premorbid functioning, wellbeing, symptoms of psychosis, co-presence of depression, cognitive functioning, insight, psychiatric and physical comorbidities and genetics. Some of the contributors have been extensively analysed, whereas others have seldom been challenged.

Premorbid Functioning and Wellbeing

Premorbid functioning and wellbeing belong to the group of candidate contributors to antipsychotics that have been the subject of sporadic interest so far.

With regard to premorbid functioning, in a report on 186 patients with a first episode of a schizophrenia spectrum disorder admitted to the Calgary Early Psychosis Programme, an association between good functioning and future good adherence has been reported [91]. Similar results have been found in a cohort of patients with a first-episode of psychosis admitted to the EPPIC of Melbourne [97]. This promising finding, however, is in contrast to some negative results [19, 93, 246]. Thus, it seems reasonable to assume, at least contingently, that the influence of premorbid functioning on adherence to antipsychotics is small, if any.

The issue of an eventual influence of personality and temperament on medication adherence has been almost ignored. However, current evidence is at least suggestive of some relationship. High sensation seeking, disinhibition and susceptibility to boredom have been associated with poor adherence in a mixed group of patients with psychosis or mood disorder [282]. Furthermore, in patients with schizophrenia, medication adherence has been reported to be inversely related to self-directness and novelty seeking, a concept close to sensation seeking [283]. A negative impact on medication-taking behaviour has also been reported for high agreeableness, another personality trait [284].

The weight exerted by wellbeing and functioning on the probability that patients will follow the prescriptions of the treating physician as indicated are more robust. A discrete number of reports emphasize that, with few exceptions, there is a positive association between wellbeing and functioning and medication adherence ([19, 59, 88, 91, 185, 249, 258, 285–288]; for supplementary references, see [64]). In particular, a subanalysis of the SOHO study [59] estimated that a 25 % improvement in SWN-k total and self-control scores was associated with higher ORs, 1.22 and 1.25, of being compliant. A longitudinal observation [88] on 162 patients in ambulatory treatment for psychotic disorders estimated an almost 5 % reduction in the risk of non-adherence for every one-point increase in WAI or GAF score.

Psychotic Symptoms

Undoubtedly, there is an association between adherence to antipsychotics and symptom severity in patients with schizophrenia. However, as repeatedly underlined, the relationship is bidirectional and the possibility of disentangling the moderating effect of medication adherence on the clinical picture and the other variables is often impossible because of the difficulties in identifying which variable comes first.

Despite this inherent limitation and the presence of conflicting results, the working hypothesis that the greater the severity of psychotic symptoms, the less the medication adherence seems well supported ([31, 51, 59, 89, 91–94, 96, 97, 111–113, 117, 125, 127, 181, 185, 186, 197, 205, 213, 243, 248, 250, 258–260, 284, 286, 287, 289–305]; for supplementary references, see [38, 47, 64, 98, 214]). The association, however, was based on different variables, such as global severity, severity of some definite psychopathologic domain, the positive in particular, or severity of defined symptoms, more frequently those pertaining to the area of delusional ideation.

Psychotic symptoms contribute to medication adherence in patients with both first-episode and chronic schizophrenia and have a modest power. For example, in a group of patients first admitted with schizophrenia or another psychosis, who were evaluated with the PANSS, “the likelihood to present with poor medication adherence was 1.6 greater for each one-point increase on the item delusion, and 1.5 times greater for each one-point increase in the item suspiciousness” [93].

Depressive Symptoms

The presence of depressive symptoms has been associated with poor medication adherence in such disparate disorders as cardiovascular disease, diabetes and hepatitis C [306–308]. People with schizophrenia are not immune to this general trend. A number of reports ([19, 94, 117, 137, 186, 287, 309–311]; for supplementary references, see [47]) that have explicitly tested the issue in patients with schizophrenia have concluded, with some exceptions, that there is an inverse relationship between depressive symptoms and adherence to antipsychotics. Furthermore, in a large-scale, 3-year, prospective, naturalistic follow-up of patients with schizophrenia included in the US-SCAP [51], previous treatment with antidepressants ranked fourth among the most powerful predictors of non-adherence: this result gives further indirect support to this negative correlation, because the prescription of antidepressants may be considered a rough proxy of a depressive condition. An opposite conclusion is suggested by the observation [290] from a small sample of first-episode patients that patients with schizoaffective disorders had higher rates of compliance compared with individuals affected by schizophrenia.

Cognitive Functioning

Some studies have explicitly tested whether cognitive function has an influence on adherence to antipsychotic medication in patients with schizophrenia ([19, 51, 67, 137, 232, 241, 289, 295, 312–315]; for supplementary references, see [38, 47, 64, 98]). Although studies vary as to the measures used, referral to global or partial aspects of cognitive functioning and some conclusions, the emerging overall picture is that cognitive dysfunction merits a mention among the contributors to poor adherence. Three large-scale trials seem particularly representative. In a German compliance survey [67] of 5,729 patients with schizophrenia, the treating psychiatrists mentioned the need for some external help to remind patients to take medicines and the presence of problems related to cognitive deterioration among the most frequent patient-related contributors of poor adherence, with frequencies of 61.8 and 54.9 %, respectively. A second trial [233] involving 2,824 Korean patients with schizophrenia treated by 131 psychiatrists concluded that 83 % of the sample “needed help from someone to remind … to take medication daily”. Similarly, the third trial, a prospective observational US study of 1,579 patients with schizophrenia treated in usual care settings [51], placed patient-reported cognitive impairment first within a list of 39 of the leading predictors of medication adherence. The observation [19] that patients with first-episode schizophrenia who stopped antipsychotics against medical advice “had poorer estimated premorbid cognitive ability than patients who consistently took medication” may be interpreted as a further indication of the negative impact of impaired cognition on medication-taking behaviour.

However, evidence of an association between higher cognitive performance and poor medication adherence has also been reported [137]. This finding is mirrored by the observation [312] that “increasing neurocognitive impairment was associated with more positive self-reports of medication adherence”. The association between neurocognitive impairment and a valid adherence status is probably spurious resulting from the propensity of patients with cognitive deficits to over report their adherence; when the judgement of a significant other was used to define individual medication-taking behaviour, the apparent beneficial influence of cognitive impairment on medication adherence disappeared [312].

Insight

Despite the persistence of some controversy about the composite nature of the phenomenon, insight does include a psychotic factor when it expresses a distortion of the reality process. Insight goes beyond this distinctive dimension when it depends on other contributing factors, for example, neurocognitive functioning, cultural and educational background and confrontation with testimonials. For this reason, the relationship between medication adherence and insight merits further analysis.

The available literature is substantial and supports, with some exceptions, the conclusion that insight or at least some of its major components are to be viewed as directly associated with medication adherence ([91, 92, 94, 97, 111, 113, 114, 137, 185, 186, 196, 231–233, 241, 244–246, 248, 259, 284, 286, 289–292, 294–296, 302, 312, 313, 316–327]; for supplementary references, see [38, 47, 64, 98, 214]). In a national survey of psychiatrists engaged in the management of schizophrenia, the indication that “denial of illness was the most commonly cited primary reason for antipsychotic nonadherence” [328] gives further weight to the relevance of insight in medication adherence.

The strength of the impact of insight on individual levels of medication-taking behaviour is not well established due to the relatively small size of the samples recruited in most of the studies and the presence of specific confounders. For example, some patients present too much insight and this feature is not an indicator of good medication adherence but, rather, suggests the opposite [330]. Another inherent confounder refers to the common entry of perceived need for treatment among the items that measure medication adherence; this choice implies that “examining the impact of such a component on adherence may be seen as tautological” [92]. However, the difference between asserted and actual adherence is likely to reduce the distortive consequences of this methodological bias [330–332].

The positive link between insight and adherence does not seem to be confined to some of the specific measures of insight used in different studies and may be indifferently applied to patients with both first episode and chronic schizophrenia.

The observation [317] that improvement in insight goes hand in hand with improvement in compliance should be viewed as further validation of the interrelatedness between the two phenomena and a strong indication in favour of the use of insight-oriented interventions to improve adherence to antipsychotics in patients with schizophrenia who manifest poor insight and fail to take therapies as prescribed. In addition, the demonstration of an inverse association between adherence to antipsychotics and denial coping may be considered as a rough example of the detrimental influence played by poor insight on medication-taking behaviour because “impaired insight and the employment of denial coping strategies have been found to be related but distinct characteristics” [300].

Psychiatric Comorbidities

Substance use disorders are a major topic in the current literature on the association between adherence to antipsychotics and psychiatric comorbidities ([51, 79, 81, 88, 90–94, 96, 97, 111, 113, 127, 137, 182, 185, 187, 205, 207, 210, 213, 244, 258, 284, 285, 293, 297, 333–335]; for supplementary references, see [38, 47, 64, 98, 214]). Overall, convergent evidence indicates that substances have an appreciable negative influence on the extent to which patients with schizophrenia follow the pharmacologic prescriptions of their physicians. However, as emphasised earlier, the association is bidirectional and thus at risk for a vicious circle of reciprocal reinforcements.

Because the few studies that have failed to document an association have mostly considered substance use disorders according to a lifetime or historical perspective [92, 93, 117, 259, 333], it seems reasonable to hypothesize that the link with poor medication adherence depends on a number of non-mutually exclusive pathways related to the recent use of substances. Within this perspective, impairment of judgement, exacerbation of psychotic symptoms, decline of motivation to pursue long-term goals, devaluation of the beneficial effect of antipsychotics, paradoxical fear of complications related to the substance–antipsychotic mixture, increased risk for conflicts with supportive figures, erosion of social support and distress secondary to the emergence of adverse events typical of exposure to substances are among the most obvious contributors. An appreciable role must also be attributed to increased risk for dosing complexity due to the need for polypharmacy. Demonstration [336] that psychoactive substances enhance extrapyramidal symptoms supports a supplementary alternative hypothesis based on the induction of increased emergence of these and, possibly, other adverse events typical of exposure to antipsychotics.

Whether the negative impact of comorbid substance use disorders on the medication-taking behaviour of people with schizophrenia is at least the representative of a generalised trans-diagnostic association between poor adherence to antipsychotics and a comorbid status remains largely unanswered because of scarce and contrasting evidence. For example, a study [337] of Medical outpatients with schizophrenia “who initiated monotherapy with a conventional or atypical antipsychotic medication” has reported that an additional diagnosis of bipolar disorder was “associated with a significantly lower rate of persistence” with the treatment. At the same time, however, a past psychiatric disorder other than substance use disorder has been associated with slightly less risk of the patients “to become medication refusers as opposed to nonadherent patients”[97].

Medical Comorbidities

People with schizophrenia are at increased risk not only for psychiatric but also for medical comorbidities [162, 173, 188–192, 194]. Patients with schizophrenia may not pay due attention to medical problems because they ascribe their physical symptoms to the mental disorder, generalize illness denial typical of schizophrenia to concomitant medical conditions, are scarcely involved in personal physical problems because of the overwhelming pressure of psychotic symptoms or, more simply, present deficits of cognitive flexibility and memory that prevent them from following medical advice. In addition, persistent unhealthy lifestyles that predispose them to medical conditions are commonly observed in concomitance with schizophrenia [338–340]. Unhealthy lifestyles and defined medical conditions found in individuals with schizophrenia may also depend on specific familial, possibly genetic, predisposing factors [338, 341–345]. Furthermore, the families, urged by the psychosis of the affected relative, may give insufficient attention to somatic problems, avoid intervening on unhealthy lifestyles and refuse to participate actively in the processes of care. Physicians and the health care system in general also contribute to the poor physical health of patients with schizophrenia. For example, it is common that psychiatrists have inadequate expertise to treat medical problems and general practitioners and specialists in other branches of medicines are insufficiently trained to work with patients affected by a severe mental illness. In addition, services that are not sufficiently tailored to the needs of people with schizophrenia may create barriers to patients receiving a good standard of medical care.

Multiple pathways contribute to increase the risk that patients with schizophrenia are not only undiagnosed or untreated for their medical comorbidities but also receive inequality of care [192, 346–355].

Irrespective of the participation of the various contributors, medical comorbidities interfere negatively with adherence to antipsychotic medications. This inverse relationship depends largely on the almost inevitable use of polypharmacotherapies. Polypharmacotherapy makes the medication regimens more complex, may facilitate drug–drug interactions that interfere with the pharmacokinetic and/or pharmacodynamic profile of the antipsychotic, and increases the chances of more frequent and more severe adverse events.

The therapies prescribed for comorbid medical comorbidities have also been reported to be at a higher risk for poor adherence compared with antipsychotics [102]. If confirmed, this could lead to a vicious circle whereby poor adherence worsens the prognosis of the medical comorbidity, the poor prognosis of the medical comorbidity forces more aggressive interventions, and the more aggressive medical interventions exert a further negative impact on adherence to medications in general and antipsychotics in particular.

Genetics

Many lines of evidence from descriptive and molecular genetics make the assumption that a number of factors that shape medication adherence have a relevant contribution from genes no less refutable. The psychopathologic domains of schizophrenia, drug metabolism and effectiveness of antipsychotics have sufficient in-depth genetic expertise [70, 72, 356–362]. Despite the strength of the construct, no study has so far tried to control for genetic variance in the influence of those contributors of medication adherence that have a genetic component. Therefore, any discussion on this topic remains purely unsubstantiated.

Furthermore, at a purely theoretic level, it cannot be excluded that genetics or, better, epigenetics affect medication-taking behaviour not only for some disease-specific or drug-specific contributors but also individual proneness to follow medical and health advice in general.

Drug-Related Contributors

Among the principal drug-related factors that define the individual levels of medication adherence, the differences between antipsychotics, effectiveness, dosing complexity and dosing regimens have been topics of major interest.

Differences Between Antipsychotics

After the advent of second-generation antipsychotics, it was hoped that they would improve the prognosis of schizophrenia and promote a more valid medication adherence due to their alleged improved efficacy/tolerability profiles compared with conventional antipsychotics [363]. Overall, controlled clinical trials, meta-analyses, guidelines and extensive use in practice have confirmed that many of the new antipsychotics are first-line options for the management of people with schizophrenia. Nevertheless, it is also true that second-generation antipsychotics are far from being ideal and that differences between individual agents do occur [6, 7, 9, 14–17, 364].

Despite the sound premise, the contribution of second-generation antipsychotics to the process of adherence improvement has been less striking than supposed ([61, 80, 81, 88, 94, 110, 111, 137, 153, 210, 248, 249, 285, 295, 337, 365–378]; for supplementary references, see [38, 64]). Three main lines of evidence support this conclusion. First, a non-marginal minority of comparisons between conventional and atypical agents has not corroborated the initial working hypothesis. Second, even when documented, the advantage of novel antipsychotics compared with first-generation agents has been shown to be small to moderate. Third, high numbers of patients with schizophrenia treated with second-generation antipsychotics continue to deviate from the prescriptions of their physicians. This scenario is not surprising, because the hypothesis that the entry of atypicals onto the market could be the keystone to overcome the problem of medication adherence in schizophrenia appears overly naive. The multifactorial origin of this treatment behaviour excludes a priori the possibility of an overwhelming contribution played by a single factor. Furthermore, comparisons of medication adherence involving patients treated naturalistically with different agents are at risk for selection effects related to an initial preference for one class of antipsychotics over another. In particular, in health care systems that encourage initial use of cheaper medications, patients who stay with first-generation antipsychotics should be inherently at higher likelihood for good adherence than individuals switched to an atypical agent. In addition, whenever first-line use of novel antipsychotics is promoted, the indication to switch to a conventional agent will presumably involve patient populations with an excess of cases with poor adherence. Selection effects may also be working in comparisons of medication adherence that involve clozapine, given its uniquely restricted indication. One of the largest studies on adherence to antipsychotics carried out to date [80] supports the relevance of selection effects in the daily routine. Compared with the group of 25,931 individuals exposed to a conventional antipsychotic in monotherapy, the sample of 23,702 patients treated with an atypical agent included only a 3.7 % excess of cases with an MPR value less than the 0.8 cut-off used to identify poorly adherent individuals. Despite the demonstration of a preferential link between first-generation antipsychotics and good medication adherence, patients on clozapine had higher MPR values compared with patients treated with conventional or other atypical agents [80]. Furthermore, among patients switched from a typical to an atypical agent, the percentage with poor adherence dropped from 46 to 40 %. In contrast, among patients switched from a new to an old antipsychotic medication, the percentage of poorly adherent individuals increased from 49 to 64 %. In turn, in a large sample of patients with schizophrenia starting on a second-generation antipsychotic, patients who had previously been exposed to a conventional agent persisted and complied better with the atypicals [79]; this indirectly supports the idea that novel antipsychotics are perceived to be better than conventional agents in relation to the issue of medication adherence.

The possibility of a differential effect played by individual antipsychotics on medication adherence has also been challenged through comparisons between agents of the same class. However, head-to-head comparisons between first-generation antipsychotics are scant ([66, 197, 379]; for supplementary references, see [64]). A 1971 report [379] from the Walter Reed General Hospital at Washington found that “more outpatients on thioridazine, 55 %, than on chlorpromazine, 15 %, were not taking minimal amounts of medication”, a rather surprising finding, in light of the demonstrated lower incidence of side effects “with thioridazine and the widely held notion that patients are more likely to take a drug having fewer side effects”. A report from a VA mental hygiene clinic [66] speaks in favour of some drug-specific effect played by first-generation antipsychotics on medication-taking behaviour: apart from 3–10 % of refusers who would not take medications at all, patients on chlorpromazine, thioridazine, trifluoperazine and perphenazine “were judged to be taking significantly less than the amount prescribed” in 20, 35, 10 and 12 % of cases, respectively. However, evidence also exists that, among acute inpatients, refusers and nonrefusers do not differ regarding the type “of described neuroleptic medications” [197].

Intraclass comparisons between second-generation antipsychotics are more common. The results, however, are conflicting [61, 79–81, 88, 110, 137, 152, 305, 365–368, 373, 375–378, 380–385]. Reports indicating that one medication is better than another in terms of good adherence rates are almost systematically at variance with others that fail to demonstrate any difference or, even, draw opposite conclusions. This may be explained by assuming an active role for a number of confounders. Two seem to merit particular mention. One refers to the common presence of a study sponsor. The other comes from the observation that in the decisional process in the real world, a pre-ordered ranking of preferences is typically adopted for the selection of the first-choice medication and this inevitably generates selection effects.

Although the various independent studies lack univocal results and frequently present appreciable points of weakness, it seems reasonable to conclude that second-generation antipsychotics should be weakly preferred to conventional agents in order to improve medication adherence. Within this specific perspective, however, none of the novel antipsychotics may be definitively considered more convenient than another, with the reasonable exception of clozapine in resistant patients.

Efficacy and Tolerability

Among the two main components of effectiveness, efficacy and tolerability, the former has so far received only marginal interest in studies that aim to challenge the putative contributors of adherence to antipsychotics. As intuitively predicted, the emerging evidence [59, 94, 137, 139, 231, 285, 386] speaks in favour of the fact that a poor response and no perceived benefits call for non-adherence and, on the contrary, a good response calls for adherence.

Attempts to establish a relationship between adverse events induced by antipsychotics and medication-taking behaviour are more common, although the decision to give preference to adverse events is somewhat surprising considering that the recognised benefits of medications have been reported to have more influence on adherence than adverse events [246]. In confirmation of this, a post hoc pooled analysis [387] of 866 patients with schizophrenia or related disorder who discontinued treatment in four randomised clinical trials has reported that “the most common reason for early discontinuation was poor response/psychiatric symptom worsening, which was three times the rate of patient discontinuation due to medication intolerability”.

Despite the presence of a discrete number of negative or even opposite results, retrospective, cross-sectional and prospective studies suggest overall that medication-taking behaviour is unfavourably affected by the patient’s present or past negative experiences in relation to the tolerability of antipsychotic medications ([19, 30, 31, 51, 59, 92, 94, 110, 111, 113, 139, 181, 197, 205, 231, 234, 241, 244, 246, 258, 285, 286, 289, 294–296, 302, 309, 310, 324,325 368, 374, 388–394]; for supplementary references, see [38, 47, 64, 98, 214]). The relatively lower benefit of compliance therapy observed among patients with extrapyramidal symptoms support this conclusion [317, 324, 392].

Harmful side effects such as akathisia, tremor and other extrapyramidal disturbances have negative effects on medication-taking behaviour. However, the association mainly involves those events that, irrespective of their objective severity, induce levels of subjective distress that are intrusive, scarcely acceptable or have a negative impact on quality of life and self-esteem. Among these are sedation, weight gain, sexual dysfunction, galactorrhea and some anticholinergic disturbances. Subjective distress due to side effects fluctuates over time [395] and is influenced by factors that, aside from the mere pharmacologic perspective, involve personal expectations, past experiences, trust in alternative therapeutic options, opinions of significant others, competence of the clinician in explaining and managing adverse events and fear of stigmatisation because some adverse event may explicitly indicate the use of an antipsychotic [395, 396]. The influence of non-pharmacologic contributors explains why subjective distress has sometimes been reported to be unrelated to objective records of adverse events and is especially difficult for clinicians to recognize and quantify [395]. Furthermore, the burden of distress due to side effects seems to be estimated differently by doctors and patients [397] and this constitutes a further relevant complication for precise definition of the involvement of distress in the genesis of poor medication-taking behaviour.

The subjective report seems to be “a much more relevant predictor of non-compliance than objective measures” [395] and this statement fits well with the observation that many patients cite side effects among the primary reasons for non-adherence or stopping the therapy [197, 230, 388].

From a critical perspective, it must be also stressed that a split approach to the effects played by the efficacy and the tolerability of antipsychotics on medication-taking behaviour is incorrect to some degree, because people taking medicines are not inclined to see side effects and symptom improvement as separate issues [398].

Treatment Complexity

The influence of treatment complexity on patient adherence to prescribed medications has been a subject of intensive investigation over the years in studies focused on dosing frequency and prescription of polypharmacy for a variety of illnesses and pharmacologic classes, schizophrenia and antipsychotics included ([29, 40, 42, 55, 79, 80, 100, 103, 110, 112, 240, 241, 249, 281, 297, 334, 374, 399–404]; for supplementary references, see [47, 64, 98, 214]). Apart from some results that swim against the tide, the overall emerging key conclusion is that the general principle of simplicity results in good adherence: indeed, the rule the lower the frequency of dosing, the higher the medication adherence seems to be sufficiently supported. Nevertheless, a non-negligible number of patients treated once a day, in clinical practice, continue “to have imperfect dosing” [42]. However, this finding is not surprising because, as already emphasised, medication-taking behaviour has a multifactorial origin and consequently single variables may explain only a small proportion of variance.

The influence of dosing complexity on adherence is affected by medication type and the common presence of other second-order associations. A VA study [334] involving more than 35,000 patients with schizophrenia or schizoaffective disorder who were prescribed one oral antipsychotic medication and filled at least one outpatient pharmacy prescription is paradigmatic of the intricate nature of the link between dosing frequency and medication adherence. In the group of 1,639 patients treated with a once-daily dose initially and then incurred an increase in their total dose, those who also increased the dosing frequency developed a relative adherence loss compared with individuals who remained on once-daily dosing after the increase. The 1,370 patients who had a decrease in dosing frequency from more than once-daily to once-daily had superior MPRs compared with patients without a decrease in dosing frequency. Among the 32,612 patients who remained on a consistent frequency, no difference in MPR emerged between individuals originally prescribed a once-daily dose and those treated more than once daily, even though in the multivariate analysis “more than once-daily dosing frequency was weakly associated with poor adherence” [334].

Whether the number of daily doses is also inversely associated with the appropriateness of taking medicines within the prescribed time frame [100] remains an open question.

The fact that patients with schizophrenia have ranked difficulties with the regimen in ninth position among the major barriers to taking antipsychotics as prescribed [230] further supports that dosing complexity has practical relevance in defining the individual level of medication adherence.

Clearly, although all the literature has so far focused on dosing frequency, other indicators of complexity may also affect medication-taking behaviour. The requirement to take drugs before food or other restrictive dosing regimens are some of the most obvious examples that should be investigated.

Antipsychotic Dose

The current literature on the putative effects of the antipsychotic dose on medication-taking behaviour is far from being conclusive ([79, 80, 111, 117, 197, 288, 291, 295, 309, 320, 365, 405]; for supplementary references, see [47, 64]). Reports indicating an association between poor adherence and high doses are at odds with others that have failed to demonstrate any appreciable dose effect or have documented that poorly adherent individuals are at lower risk for having ever received a high antipsychotic dose.

These discrepancies may be reasonably justified considering that the influence of medication dose on adherence is largely sustained by second-order associations and spurious contamination. It is true, for example, that a low dose of an antipsychotic agent may favour adherence, reducing the potential for adverse events, but it is also true that this pro-adherence effect frequently occurs at the detriment of efficacy, another adherence promoter. In clinical routine, the prescription of high doses of antipsychotics is also commonly coupled with an increase in dosing frequency and primarily involves patients with a severe and/or a scarcely responsive disorder; more than once-daily dose, high symptom severity and resistance to antipsychotics have autonomous contributions to poor medication adherence. Furthermore, patients with unrecognised non-adherence are almost systematically exposed to high doses of antipsychotics, because they are erroneously classified as non-responders.

Irrespective of the nature of the link between antipsychotic dose and medication-taking behaviour, clinicians should systematically regulate their practice based on the fact that even small deviations from the prescribed drug regimen may have a strong risk for a psychotic exacerbation [85]. However, “whether the current clinical practice of prescribing the lowest amount of psychiatric medication to ameliorate symptoms while keeping side effects at the minimum might be contributing to a situation in which there is very little allowance for even partial nonadherence” remains an unresolved issue that merits systematic monitoring.

Composite Contributors

Patient preferences, therapeutic alliance, and attitude towards medication come under the heading of composite determinants and moderators of adherence to antipsychotic medication.

Patient Preferences

In recent years, the decisional process leading to the selection of therapy has progressively shifted from a model strictly directed by the doctor to a shared approach in which the patient’s preferences and point of view have an appreciable role. This evolution has also been found to be important with regard to medication adherence. A recent experience [406] based on the use of psychiatric advice directives that “allow individuals with severe mental illness to document preferences for future treatment if they lose decisional capacity during a psychiatric crisis” is paradigmatic in this regard: “receiving at least one requested medication at the 12-month follow-up predicted greater adherence at 12 months”, with an OR equal to 7.8.

The involvement of patient preferences on medication-taking behaviour is mediated by a set of direct and indirect effects. For example, the active participation of the patient is an essential constituent of therapeutic alliance and attitude towards medications, two well-established contributors to medication adherence. Furthermore, the active participation of the patient has a direct influence on the extent to which a patient’s behaviour coincides with the prescription of the treating physician. Preference for a specific drug formulation is paradigmatic in this regard. Early evidence of a link between the formulation of a medicine and adherence can be traced back to the early stages of clinical psychopharmacology. For example, in 1967 a pioneering report [291] demonstrated how urine tested by the Forrest reagent changed from negative to positive in patients with schizophrenia who were switched from chlorpromazine in tablet form to an identical dosage of the liquid formulation. This initial finding has been partially confirmed in more recent years. Compared with standard tablets, the orodispersible tablets of olanzapine have been found, although not systematically, to be associated with improved medication adherence and/or efficacy [407–411]. However, the possible beneficial influence of the orodispersible tablets on medication-taking behaviour cannot be considered merely as a direct consequence of the preference attributed by patients to this specific formulation. The association has also been observed [410] in patients unaware of the specific formulation received because they were enrolled in a double-blind, double-dummy study. An alternative mechanism should therefore be hypothesised. Of particular interest is the emergence of a limiting weight gain effect mediated by sublingual absorption [410] because this might, in turn, facilitate medication adherence, at least in people prone to the weight gain phenomenon.

The influence of the drug formulation on medication adherence is not confined to comparisons between drops and tablets or orodispersible and conventional tablets; long-acting injectable antipsychotics are also of interest. The depots are likely to provide the best testimonials of an association between drug formulation and medication-taking behaviour, because they confer a reasonable benefit over oral agents on the global outcome [13, 68, 412–418]. There are two main advantages: long-acting antipsychotics offer a time-saving and money-saving solution to the problem of inveterate intentional refusal to take the prescribed therapies and make unintentional deviations from medical advice by the patient transparent. The pro-adherence effect of the depots can be reconciled with patient preference using this formulation. The literature suggests that, on the whole, individuals with schizophrenia accept long-acting injectable antipsychotics well, although studies are far from being univocal in their conclusions [408, 415, 419–423]. Reports in favour of depots over oral agents are contrasted by others that go in the opposite direction or fail to detect appreciable differences. These discrepancies are largely due to reporting bias, selection bias and other peculiar confounders. For example, in a sample of depot-naive patients with schizophrenia, only 23.4 % were reported to accept long-acting injectable antipsychotics, a rate definitely inferior to the 45.3 and 73.3 % observed in groups of individuals previously or currently treated with depots, respectively [408].

Therefore, there are good reasons for hypothesising that exposition to a defined formulation guides individual preference to an appreciable degree, with a stronger driving effect when the formulation, oral or depot, is taken contingently [408, 422]. This link is probably the expression of a generalised trend; it has also been documented in a comparison of orodispersible and conventional tablets of olanzapine [411].

The preference for an oral or a long-acting injectable antipsychotic also depends on the specific agent under review. The results of six studies included in a systematic review [419] are paradigmatic in this regard: five studies compared oral and long-acting first-generation antipsychotics and reported higher rates of preference for the depots, whereas the study in which the oral formulation emerged as the winner involved a comparison between risperidone tablets and long-acting conventional depots. Considering the differences that occur between atypical antipsychotics [14, 15] and the increasing number of second-generation depots available, the efficacy and safety profiles of each agent will have a greater influence on patient preference for a specific drug formulation in the near future.

A further moderator of individual preference for a specific drug formulation is the level of voluntarism that sustains patients in taking their medicines. The presence or perception of force or coercion is a common feature in patients exposed to antipsychotics in general [196, 231, 247, 286, 297, 424, 425]. In a group of outpatients with schizophrenia or schizoaffective disorder undergoing voluntary maintenance therapy, only 12.5 % denied any concern about coercion [425]; this seems enough to depict how common the phenomenon is. The observation that patients hospitalised “on an involuntary basis reported more decisions that they would have made differently than patients being treated voluntarily” explains why these patients are extremely vulnerable to poor medication adherence [426]. The association between coercion and poor medication-taking behaviour is relative, as exemplified by the report [427] of a lack of differences in medication adherence among patients admitted voluntarily and involuntarily with a first episode of psychosis. The formulation of the drug has some specific relevance in relation to the level of forced medication. From this specific point of view, the disadvantage of depots seems intuitive considering that, unlike the oral formulations, they are “given rather than taken” [425]. Experimental evidence supports this statement indicating that, compared with individuals treated with an oral formulation, patients treated with long-acting injectable antipsychotics have higher scores related to perceived coercion and negative pressure [425]. This finding is not surprising. Many physicians restrict the use of long-acting injectable antipsychotics to patients who refuse or are reluctant to take medicines as prescribed and both patients and doctors frequently state that the depots are not appealing because they promote stigma. The observation [408] that two-thirds of patients who experienced a depot prescription as a compulsory measure quickly suspended the treatment demonstrates the unfavourable impact of lack of voluntarism in taking medicines. A report [392] that patients admitted voluntarily responded better to compliance therapy gives further indirect support to this conclusion. This evidence strengthens the need for ad hoc interventions to give patients a genuine role in the decisional process that leads to the choice of a long-acting injectable antipsychotic. To do this, however, we first have to change the culture of many psychiatrists, because they often have a negative or a non-positive attitude to the depots [428]. This explains why, assuming comparable rates of relapse after treatment with oral or depot antipsychotics, 81 % of doctors reported that they would recommend the oral formulation [429] Therefore, it is not surprising that, in a questionnaire completed by 246 psychiatrists attending the eighth World Congress of Biological Psychiatry, “less than 36 % of participants’ patients have ever been offered antipsychotic depot treatment” [430]. Similarly, in a study in the canton of Zurich [420], 75 % of the psychiatrists interviewed reported that “they informed their patients about different formulations including the options of depot treatment”, but 68 % of them also acknowledged “that patients are not sufficiently informed about different methods of administering antipsychotic drugs”, a conclusion confirmed by the high proportion, 67 %, of patients who said that “they did not receive information about depot antipsychotics from their psychiatrist”. This difference between the opinions of patients and doctors is especially discouraging from a clinical perspective considering that in a group of patients with schizophrenia who were uncertain about whether to continue the depot or not, 87 % decided to persist with therapy after a dedicated psychoeducational intervention [431]. Supplementary evidence of the propensity of the psychiatrists to overestimate their engagement in the search for active involvement of patients in the decision-making process comes from a German study that estimated the participation of patients at 70 and 83 % according to the judgement of patients and doctors, respectively [426]. Working on patient preferences may therefore be regarded as a fruitful strategy to improve adherence, provided this direct intervention does not take the place of others aimed at consolidating therapeutic alliance and attitudes to medication. These three factors have certain relevant points in common. Nevertheless, they do not seem to be merely synonymous, because current drug formulations have been reported to be associated with medication preference but not to predict attitudes to antipsychotics [422].

Therapeutic Alliance

Therapeutic alliance recognises two undisputed protagonists: the doctor and the patient. The doctor contributes professional competence, interest in medicine-based shared decisions, communicativeness and empathy. The patient participates with personal existential background in general, illness and treatment history and contingent expectations. However, the environment plays a supporting relevant role because it interferes with the efforts of the doctor and patient to build the alliance.

The relevance of the link between therapeutic alliance and adherence of patients with schizophrenia is far from being a recent discovery. Almost half a century ago [379], for example, it was reported that 39 % of outpatients treated by a physician “who viewed medications as less than essential” defaulted in medication regimens, whereas patients cured by a physician “who viewed medications as essential in the outpatient treatment of chronic schizophrenia” defaulted in therapies with a 25% rate. In successive years, the conclusion that a valid and trusting patient–doctor relationship promotes the perception among patients that there is a need for continuity of care with prescribed medications has been vigorously supported. Antipsychotic drugs are included in this general rule ([88, 111, 186, 196, 221, 231, 248, 288, 296, 303, 309, 432, 433]; for supplementary references, see [38, 47, 64, 98, 214]). A number of reinforcements directed to the patient may turn a valid therapeutic alliance into good medication adherence. These include optimism about the usefulness of the prescribed medicines, correct expectations about the risks and benefits of the therapy, interest in understanding the illness, meaningful involvement in the therapeutic project and realistic perception of the therapist, together with responsibility and self-control for the treatment. “The fact that patients usually show better compliance and better treatment response rates in clinical trials than in real-life settings” [434] may be viewed as a demonstration of the relevance of this long list of factors on individual medication-taking behaviour.

Irrespective of the underlying moderators involved, the impact of therapeutic alliance on medication adherence is relevant. For example, a longitudinal study [88] has shown that every one-point increase in WAI score was accompanied by approximately a 5 % reduction in the risk of becoming non-adherent. The fact that patients with schizophrenia included lack of trust in the provider (i.e. a proxy for poor therapeutic alliance) in a list of the top ten barriers to medication adherence [230] may be considered a further indication that a valid alliance is a relevant prerequisite for taking medicines as indicated by the prescriber. Therapeutic alliance also has an indirect role on medication-taking behaviour. The ability of the doctor to work with the patient is crucial for linking individuals discharged from hospital to outpatient specialty facilities deputed to ensure long-term assistance and medication adherence. In this regard, the report [435] that 65 % of patients failed to attend scheduled or rescheduled initial outpatient appointments is a dramatic warning, even though the need for continuity of care has recently been emphasised. In a comprehensive review [47] that assessed numerous risk factors for poor adherence in people with schizophrenia, inadequate hospital discharge planning was included in the list of the variables most consistently associated with patient deviation from medical prescriptions.

Attitude to Medication

The composite origin of attitude to medication is well supported. A study in which structural equation modelling was applied to a wide number of variables concerning 228 patients admitted consecutively with a diagnosis of schizophrenia or schizoaffective disorder [221] illustrates this. Attitude to medication was found to be “predicted by insight, relationship with staff (especially the physician-prescriber), and the patient’s admission experience” [221], whereas a poor relationship with the prescriber, experience of coercion during admission and low insight were the main predictors of a negative attitude [221]. The same study also reported that attitude to drugs was more marginally influenced by medication knowledge and symptoms but substantially untouched by adverse events. The assessment of adverse events using the LUNSERS could have influenced the observation that iatrogenic health effects failed to have appreciable effects on adherence; the scale examines the objective presence of an adverse event rather than the severity of the associated subjective distress. Furthermore, as an expression of subjective interpretation of changes related to the pharmacologic therapy [436], attitude to medication includes other contributing factors such as symptom severity, functioning, personality traits, previous experience with antipsychotics, societal and familial attitude to psychoses and their treatment and possibility of access to health care services.

The issue of a link between medication adherence and attitude to antipsychotics is far from new. For example, in a 1964 report [247] on attitude to drugs of 30 inpatients with schizophrenia known to be acceptors or refusers of neuroleptics, the group of extreme refusers rated medicines less favourably. The negative influence played by low attitude to medication on adherence has been substantially confirmed over the years, although with some discrepant findings ([92, 113, 181, 185, 197, 221, 231, 240, 241, 246, 247, 249, 289, 294, 299, 303, 319, 320, 325, 327, 374, 388, 391, 436–439]; for supplementary references, see [38, 47, 64, 98, 214]). The reported association [324] between attitude to treatment and efficacy of compliance therapy indirectly corroborates this trend.

What remains is a largely unanswered issue: whether the driving role played by a negative attitude to antipsychotics on poor medication-taking behaviour involves not only actual but also hypothesised attitudes. One study [391] that explicitly tested the influence of hypothesised attitudes using the willingness of the patients to give the same “antipsychotic to their children, if they would have the same illness” suggests this last possibility could be realistic.

Predictors

The many heavy burdens imposed by poor adherence to antipsychotics and the existence of interventions that can successfully combat it justify the claim that individuals at risk for non-adherence must be identified promptly. Evidence of unexpected negative outcomes could constitute a first useful warning sign but this is a very rough index and has an appreciable risk for false positives. More powerful determinants and moderators of medication-taking behaviour would be more appropriate. However, extensive application of this approach may be restricted by the need for data collection requiring competences exceeding those typical of the daily clinical routine. Valid and easily accessible markers devoid of any causal influence on medication adherence may be a useful tactic to surmount many of these limitations.

Age and ethnicity of the patient, together with previous medication adherence are included among non-causal markers useful for identifying patients according to medication adherence.

Age

The influence of the age of the patient on adherence to antipsychotics has been tested with mixed results ([51, 59, 61, 79–81, 91, 113, 147, 159, 187, 196, 197, 210, 234, 241, 258, 285, 287, 291, 294, 298, 320, 324, 325, 334, 337, 385, 440]; for supplementary references, see [38, 47, 64]). Overall, when present, the association with age has almost systematically supported the conclusion that younger patients are at an increased risk for poor adherence. The result is not completely surprising. In agreement with the Health Belief Model [98], younger patients are “less likely to appreciate the severity of their illness or the need for medication” [80] because they lack direct personal experience with schizophrenia and its treatment, and generally have few opportunities to interact with a chronic patient who may act as a testimonial. Common evidence that people with a first-episode schizophrenia are particularly sensitive to metabolic and extrapyramidal adverse events [6] supports the hypothesis that iatrogenic health effects contribute to the development of the association between young age and poor medication adherence.

The observation, derived from the Calgary Early Psychosis Programme [91], that, among patients with first-episode schizophrenia, those presenting poor adherence were relatively younger suggests that the Health Belief Model is probably not enough to fully explain why younger individuals fail more frequently to follow medical advice.

Whatever the causal mechanisms involved, the fact that younger patients have more difficulties in taking prescribed medicines may be used successfully in clinical routine to identify first-choice candidates for interventions to monitor and promote good medication adherence. This is further justified when one considers that younger patients with problems of adherence have been reported to correct this specific behaviour quickly [88].

The presence of a preferential link between young age and poor medication status implies that the probability of being adherent may increase with age. This progression plausibly occurs until full adulthood but does not apply to the elderly. A long list of elements have led to the hypothesis that a second peak of increased risk for poor medication-taking behaviour occurs in old age. For example, elderly patients with schizophrenia inevitably incur age-related pharmacokinetic and pharmacodynamic changes [441]. Furthermore, elderly patients are more subject to cognitive and memory deficits, frequently receive insufficient social support, are more likely to be incapacitated in physical dexterity due to disparate medical comorbidities, and present an increased risk of perceiving medicines as unnecessary [202]. All these factors contribute to poor adherence as a result of unintentional oversight in taking medicines, more complex polypharmacotherapies and sensitivity to the unwanted effects of medications. This relevant theoretic framework, however, has not been accompanied by sufficient experimental support so far. More precise information on this issue is necessary. Two supplementary considerations are important: the increase in life expectancy over the last century [442] that has multiplied many-fold the number of people aged 65 years and older; the common observation that, in clinical practice, many elderly patients with schizophrenia or paraphrenia are treated with low doses of antipsychotics [443], with the consequence that they may be extremely vulnerable to even marginal deviations from the prescribed regimen.

Ethnicity

Some effect of ethnicity on medication-taking behaviour has been frequently but not systematically reported in the literature ([51, 61, 80, 81, 89, 111, 137, 147, 159, 187, 207, 210, 241, 258, 285, 298, 334, 367]; for supplementary references, see [47, 64]). The link has emerged more commonly in US studies involving multi-ethnic samples of patients with schizophrenia. In general, African Americans, Mexican–Americans, Latinos, Chinese and Asians have been found to be at higher risk for poor adherence to antipsychotics than non-Latino whites.

Despite the many positive findings reported, the conclusion that ethnicity represents “an imperfect flag for other factors that underlie adherence” [80] seems to be absolutely right. In the link between medication-taking behaviour and ethnicity, second-order associations play a major role. Unfortunately, the list of second-order associations that could mediate the apparent effect of ethnicity is substantial. Susceptibility to adverse events, culture, acculturation, language proficiency, level of familial and social support, income, trust in alternative medicine, religion, integration, immigration status, racial tolerance of the leading community, rights of minorities, inequalities in the standard of care, barriers to a valid patient–doctor and patient–provider relationship and opportunities to fully benefit from educational campaigns on schizophrenia and its treatment are some of the most obvious headings. In addition, patient–doctor interethnic and intercultural differences may make establishment of a valid rapport more difficult and facilitate the risk of a misdiagnosis [444]. In both cases, the probability of poor medication adherence increases as a result of an invalid therapeutic alliance and prolonged exposition to the unpleasant effects of therapies devoid of appreciable clinical benefits. Furthermore, under-prescription of second-generation antipsychotic may be mediated by ethnicity. For example, in a sample of 36,575 community-dwelling Medicaid beneficiaries affected by a severe mental illness, the “odds (of Blacks) of receiving a newer atypical antipsychotic were less than half of White and Other minority beneficiaries” [440]. This scenario is even more complex, because the representativeness of individual ethnic groups and opportunities for integration are susceptible to relevant variations over time and between countries. The effect of ethnicity on medication-taking behaviour, however, does not seem confined to sociocultural factors. Ethnic stratification was reported to exert a pharmaco-specific moderating effect not only on the risk for metabolic syndrome [390] but also on the association between treatment response and defined genetic variants [445].

Given this complexity, it is easy to understand that achieving valid and generalizable conclusions about the role of the different contributors grouped under the ethnicity label is at least problematic. Third-order associations are also likely to operate. For example, in a large-scale study that tested adherence to antipsychotics among non-Latin whites, Latinos and Asians with schizophrenia and considered the English proficiency of the two minority groups, “Latinos with limited English proficiency were more likely than English-proficient Latinos to be medication adherent and less likely to be excess fillers”, whereas “Asians with limited English proficiency were less likely than English-proficient Asians to be adherent, more likely to be nonadherent, and less likely to be excess fillers” [187].

Regardless of the underlying mechanisms involved, the evidence is that associations with medication adherence that are valid for one ethnic group may not be completely applicable to another, and this variability is a relevant complication for psychiatric services that care for ethnically and linguistically diverse adults with schizophrenia.

Previous Medication Adherence

Previous adherence to antipsychotics is not a favourite topic among studies examining predictors of medication-taking behaviour. Nevertheless, when tested, this variable has systematically been shown to have a valuable predictive power ([19, 51, 80, 88, 93, 111, 127, 241, 245, 301, 323, 388, 446]; for supplementary references, see [38, 47, 214]).

Findings from 1,579 patients with schizophrenia treated with any oral antipsychotic who were enrolled in the US-SCAP study are particularly informative [51]. This prospective observational study focused on “39 previously reported potential risk factors of nonadherence with antipsychotic medication” related to the patient, the environment and the treatment demonstrated that the single best predictor of future adherence was patients’ previous adherence (OR 4.14) and 78.5 % of subjects were accurately classified. The accuracy level was driven primarily by a high positive predictive value (86.6 %) and a moderate negative predictive value (43.9 %).

Overall, that a personal history of previous adherence to antipsychotics is a predictor of future adherence is understandable. Although transition from adherence to non-adherence is not an exceptional event and the opposite may occur, especially when dedicated interventions are predisposed, antipsychotic-taking behaviour may be regarded as a relatively persistent phenomenon.

With regard to the principal mechanisms sustaining the predictive value of previous medication-taking behaviour, it seems reasonable to argue that factual experiences with psychopharmacotherapies rather than pre-existing attitudes exert a preferential involvement. In a small, clinically heterogeneous group of never-treated patients with a first episode of psychosis, the initial personal prediction of compliance has been shown to have low reliability based on compliance observed over 3 months [294]. Furthermore, the demonstration that “past antipsychotic use rather than receipt of a specific medication was most associated with future antipsychotic adherence” suggests that “what patients know and do as a result of prior medication or illnesses experience … is a more robust predictor of future adherence than the receipt of a specific medication” [378].

Present and Medium-Term Perspectives

In 1992, a review on medication adherence in general [447] concluded that “more than 20 years of research in the area of compliance has produced very little consistent information on the factors which can be correlated with non-compliant behaviour”. Twenty years of intensive, worldwide, dedicated research have elapsed since then and we have certainly learned a lot about the dimension, the consequences and the contributors of poor adherence to antipsychotics in people with schizophrenia. Nevertheless, much more remains to be learned. Furthermore, relevant limits continue to affect these studies and this level of evidence explains well why the phenomenon of medication-taking behaviour continues to be “far better documented than understood” [64]. Therefore, urgent requirements have emerged not only for a critical analysis of the evidence accumulated so far but also for the planning of new research.

One lesson that has been learned refers to the persistent validity of the old sentence that “the answer to the question ‘is this patient taking his drug?’ will depend on the criterion used” [30]. The innumerable measures of medication adherence used so far have generated results that are only partially comparable. Furthermore, and this is even more relevant, no measure is faultless. Therefore, the ideal gold standard of reference is not at our disposal and the perfect measure that may be easily, systematically, and cheaply used probably does not exist.

On the other hand, a doctor’s ability to identify patients presenting poor adherence to antipsychotics is certainly better than indicated by the sentence, referring almost half a century ago to the intake of antacids, that “the physician ability to distinguish those patients who adhered well from those who adhered relatively poorly was no better than might be expected by chance” [69]. Nevertheless, when they have to identify which clients fail to follow their prescriptions, the clinicians of today continue, in the absence of dedicated interventions, to have difficulties so that even the worst measure would be more reliable than the personal feeling of the doctor [69]. Consequently, we are forced to use at best the imperfect measures of adherence currently at our disposal, searching in the meantime for consensus guidelines to make the results of independent experiences comparable.

While waiting for consensus, it may be useful to follow, for example, the proposal to always report “an estimate of the mean percentage of medication taken as prescribed during the follow-up period” and “include at least two measures of adherence” [37]. This approach, however, may become burdensome not only for practical purposes but also for research.

Clinicians and researchers should also keep in mind that measures that are strongly indicated for research may be less indicated for clinical practice and vice versa. In research on medication adherence, for example, it is essential that the adopted measure does not require the active voluntary cooperation of the patient. In the absence of this requirement, it is likely that individuals who are reluctant to strictly follow the advice of the physician will refuse to participate or deliberately develop surreptitious deviations from usual adherence. These two decisions imply that the products of the research, that is the results, cannot be generalised. Within this scenario, refilling of prescriptions reported in pharmacy databases seems to be one of the best options, at least in countries that guarantee universal drug coverage or adopt other close pharmacy systems. Despite the presence of some weaknesses, this approach has advantages over other indirect measures because it does not imply any active voluntary cooperation of the patient, the costs are reasonable, it is not time consuming, permits collection of data from numerically representative populations, and may be used in longitudinal studies aimed at evaluating eventual changes in treatment adherence. Measures indicated for clinical practice must also satisfy some of the prerequisites that are especially useful in research, such as convenience, low cost, not time consuming and can reveal eventual modifications of medication-taking behaviour. However, measures for clinical practice do not require the exclusion of active participation of the patient. In this specific setting, active participation could even be recognised as an added value because in the daily routine, the treating physician is also engaged with the need to reduce poor adherence and reinforce good adherence. For these purposes, active involvement of the patient has a positive influence due to a continuous sensibilising effect. A recent proposal [448] to use a cellular phone with a camera to photograph medicines in the patient’s palm just before taking them, which would time stamp when the photo was taken, is a new and promising strategy. Although photos do not ultimately prove that the patient ingested the therapy, this method, like MEMS, can document adherence to prescription timing and perform repeated evaluations and has the appreciable advantage that it is cost-effective, easy to use and based on widely available equipment.

Another lesson extrapolated from an overview of the current literature is that clinicians should desist from consuming too much of their time in debates about the precise percentage of people with schizophrenia who are poorly adherent to antipsychotics. Low concordance rates between independent measures and the diffuse risk of multiple selection biases address this issue coherently. Furthermore, it is already known that the proportion of individuals who do not take antipsychotics as prescribed is large enough to merit systematic study.

The promotion of head-to-head studies that use the same measures to test for adherence typical of distinct clinical conditions should be strongly encouraged, because this substantially neglected approach is relevant in order to reach valid conclusions about the trans-diagnostic rather than disease-specific or mixed nature of the processes underlying medication-taking behaviour.

In addition, the consequences of poor adherence to antipsychotics have been extensively explored over the years in almost all its principal aspects. Unrecognised poor adherence and costs, especially indirect costs, imputable to a failure of patients to follow medical advice correctly are the main partial exceptions. Because more precise knowledge on these two issues is relevant to understand the global impact of poor adherence, studies on these specific topics would be welcome.

Another area that calls for renewed experimental interest comes from the old observation that “trials based on unsupervised oral medication have probably been built on very unsure foundations” [449]. As emphasised earlier, clinical trials commonly use a rough measure of adherence, the simple countability of returned pills, despite the fact that failure to take medications meticulously adds variance to the results and thus weakens the statistical power. This fragile approach needs to be reversed. Stratification of the results according to adherence levels is a prerequisite for reaching more definite answers to some issues that are essential for regulatory purposes, in particular the superiority of a new drug compared with marketed comparators, the precise identification of therapeutic dose ranges and the anticipation of first-line candidates for special monitoring in post-marketing pharmacosurveillance [450, 451].

A point that probably does not merit any further motivation pertains to the search for new determinants and modulators of medication adherence. The issue has been extensively tested in a large number of studies and the list of proven contributors is ample enough to justify the assumption that absolute omissions seem unlikely or, if present, are candidates to explain only a marginal quota of variance. This does not mean that current knowledge on single factors involved in the adherence process is exhaustive. Premorbid functioning, personality traits, temperament and genetics are sufficiently representative of this. However, the central contribution that the subjective feelings of the patients play on medication-taking behaviour requires closer reassessment. The progressive shift from compliance to adherence as the preferred label, the common referral to the Health Belief Model, the pressing emphasis in favour of shared decisions, and, more broadly, the claims that the preferences of the consumers are valued, coherently address the issue. Nevertheless, studies on contributors to medication adherence continue to be anchored to a medically centred approach that leaves the opinions and expectations of the patients at the periphery to the point that they may be interpreted as repressive, as if the failure to follow the advice of the doctor merits blame. Excluding any blaming intent, this approach is wrong because it minimizes a priori the possibility of improving adherence based on patient preferences, shared decisions and therapeutic alliance. A strong call for taking the point of view of the patients seriously comes from the demonstration that doctors and patients frequently conceive adherence and the various pros and cons of the therapy in a conflicting way, to the point that what is important for one may be irrelevant for another [397, 452–456]. For example, the indication posited by patients of a clear hierarchy of preferences in which “the strongest priorities were placed on reducing confusion and increasing energy” [455] is in sharp contrast to the scarce attention that doctors generally pay to these two items. The impact of this and similar mismatches may be overwhelming for both research and clinical practice.

In the future, even with full knowledge of all the contributors to medication adherence, attempts to transfer into clinical daily routine an easily accessible, ready to use identikit of those patients with schizophrenia who are especially predisposed to poor antipsychotic-taking behaviour seem doomed. A number of considerations lead to this pessimistic conclusion. First, the number of susceptibility and plasticity factors that govern medication-taking behaviour implies very small to small levels of variance explained by each single contributor. Furthermore, the coexistence of protective and vulnerability factors, the presence of complex interplays between independent, addictive and interactive effects, relevant participation of indirect associations and variable expression of individual contributors among people with schizophrenia strongly limit the possibility of classifying individuals with poor or good adherence. Therefore, clinicians may at best target only some more or less reliable indicators of adherence. Previous medication-taking behaviour, eventual comorbidities with substances use disorder, young age, a history of repeated rehospitalisations and belonging to a minority are the most valuable factors. Although these accessible variables are imperfect indicators of adherence, they are better than nothing, because they may help clinicians in their efforts to direct the resources at their disposal to identify and manage medication adherence. Restriction of social and health care funds worldwide add value to this approach.

The statistical support necessary for better knowledge of poor adherence to antipsychotic plausibly requires innovation. Multivariate statistical models have so far provided essential support in improving global understanding of the mechanisms that sustain adherence to antipsychotics. Nevertheless, classic multifactorial statistics on group effects are only partially useful to predict individual proneness to poor medication-taking behaviour and identify the first-choice interventions for improvement. To this aim, statistical procedures centred on the single patient, for example artificial neural network analysis, are a useful approach. Unfortunately, these models have rarely been applied to the medication adherence phenomenon so far.

A number of appreciable unmet needs therefore persist in the field of research on medication-taking behaviour of people with schizophrenia. However, the most pressing demand involves what we can expect from systematic application of interventions in the daily routine aimed at improving adherence to antipsychotics. This question is not trivial. The prognosis of schizophrenia continues to be unfavourable in a significant proportion of patients despite the definite progress since the advent of second-generation antipsychotics. The search for new and better antipsychotics must therefore be pursued. Nevertheless, the observation that, compared with agents already in the market, “a hypothetical new medication that is 50 % more efficacious would decrease 1-year postdischarge rehospitalisation rates by 18 %” [122] does not leave room for excessive optimism, because agents with such clinical potential are not just around the corner. Thus, there are good reasons for strong investment on strategies to improve medication adherence; the hope for better standards of care granted by the antipsychotics of the future will remain a largely empty promise in the absence of adequate control of medication-taking behaviour. This conclusion is in agreement with the estimate that “improving compliance by 50 % would decrease 1-year rehospitalisation rates by 12 %” [122], i.e. a rate close to the estimate for drugs of the future, which are outside our current expectations. Fortunately, many of the contributors to poor medication adherence are modifiable as a result of several intervention. Some do not require expertise outside the standard professional education. Others need the involvement of ad hoc trained personnel.

Interventions that are within the reach of any mental health team include communication respectful of the patient, information adequate for formulation of valid consent and at low risk of error by patients, preferential use of drugs at the lowest efficacious dose, judicious reduction of polypharmacotherapies to the minimum required and systematic revision of prescriptions, so that therapies that are no longer necessary are ceased. Although all these pro-adherence strategies must be planned throughout the process of care, they should be intensified concomitant with discharge from hospital or, more broadly, during recovery from an acute episode of psychosis. As emphasised earlier, this period is at special risk for the emergence of poor medication adherence and incomprehension in the doctor–patient communication. Therefore, the proposal [457] of a “transition coach” who governs the passage from hospital to community care could be rewarding and cost-saving for psychiatric patients.

Another easy and direct strategy that may lead to improved medication adherence in patients who are strongly reluctant to follow the prescriptions of the treating physician involves a more systematic use of the long-acting injectable antipsychotics [416]. The use of this specific formulation in clinical routine is limited in many countries due to persistence of several barriers, in particular, the reluctance of doctors to prescribe long-acting antipsychotics, the scarce information for patients and families about these agents and the patients’ perception of being under unusual coercion by the treating physician. The time is now right for a turnaround. The recent commercialisation of different long-acting second-generation antipsychotics and the approach of others to the market provide an excellent opportunity for removal of these barriers because individualised use of the depots is now possible.

In recent years, an increasing number of non-pharmacologic interventions that require special training have also gained ground to combat the susceptibility of people with schizophrenia to poor adherence to antipsychotics [458, 459]. Despite the vast literature, the cost effectiveness of the different approaches, the identification of obstacles to their extensive use in clinical routine and the selection of the eventual front runners among the various interventions await more definitive answers. To this aim, supplementary head-to-head comparisons are particularly indicated.

Stigma against severe mental disorders and psychopharmacotherapies is among the areas of intervention that merit priority funds. Despite the message that schizophrenia has solid biological origins, and thus is to be viewed as a disorder like any other, and is a leading factor in the widespread war on stigma, large sectors of society continue to have stigmatising attitudes. Therefore, the anti-stigma campaigns of the future need to be radically changed to provide alternative messages with major emotional resonance. However, current straitened circumstances force us to identify target priority groups for anti-stigma interventions given that improvements in this area imply discrete time latencies before their beneficial effects may become tangible. High school students are a good first-choice population. Young people are especially receptive to anti-stigma messages, because they have been reported [226, 228] to be at a minor distance from people with schizophrenia. Furthermore, with successful intervention, students disseminate their position against stigma for many years, thus diluting the prejudices of the lay public further in the long term. Teachers, professional figures operating within the health care system, politicians and people close to patients with schizophrenia are also reasonable figures of elective reference. Due to their pedagogical expertise, teachers are central to students’ education and may become autonomous promoters of supplementary anti-stigma groups. The importance of professional figures within the health care system resides in the fact that they are in charge of reinforcing the message that antipsychotics are mandatory in the therapy of schizophrenia. In addition, the anti-stigma commitment of politicians is essential because they control the funds necessary for anti-stigma campaigns. The people close to patients need dedicated interventions because they too may experience stigma, sometime stigmatize the affected member of the group, and are always in the forefront for abating poor medication adherence.

An ultimate need that is persistently unmet is the inability to convert the many pro-adherence interventions into an opportunity for truly individualised countermeasures. This relevant advancement presupposes systematic access to statistical support that can define the weight of the various contributors of medication-taking behaviour in each individual patient and to insert these estimates into a model that takes into account the margin of improvement granted by the different interventions for any specific area. None of these conditions are even vaguely applied in current clinical practice.

Closing Remarks

Accumulated evidence promises new possibilities for the recognition and management of poor adherence to antipsychotics in people with schizophrenia. Nevertheless, unless the gap between what we know about the issue and what we use in our clinical work is drastically reduced, this progress risks remaining largely a dead issue. Consequently, the old maxim “every patient is a potential defaulter. Compliance can never be assumed” [401] will continue to express a nihilistic message that it is impossible to discriminate good and poor medication-taking behaviour rather than a simple, descriptive message that non-adherence may be always imminent.