Introduction

People with HIV infection are confronted with many challenges in achieving and consistently maintaining adherence to antiretroviral therapy (ART). Studies show that suboptimal adherence to dose instructions (i.e. below 95% adherence) is strongly associated with diminished viral suppression (Bartlett, 2002; Paterson et al., 2000), increased risk of viral mutation, and the emergence of viral resistance (Poppa et al., 2004). Suboptimal adherence to ART can lead to the development of resistant viral strains that can be transmitted to others (Poppa et al., 2004). Research has identified a number of medication and psychosocial factors associated with suboptimal adherence to dose.

Medication factors such as dosing frequency (Paterson et al., 2000; Stone et al., 2001) and higher daily pill number (Bartlett, DeMasi, & Quinn, 2001; Eldred, Wu, Chaisson, & Moore, 1998) are associated with lower adherence to dose. Patients who receive a protease inhibitor (PI)-based medication regimen are more likely to report lower levels of adherence to dose than those receiving nonnucleoside reverse transcriptase inhibitors (NNRTI) regimens (Altice, Mostashari, & Friedland, 2001; Kleeberger et al., 2001; Mannheimer, Friedland, Matts, Child, & Chesney, 2002). Other studies have shown that adverse side effects are strongly associated with suboptimal adherence to dose instructions (Ammassari et al., 2001; Bartlett, 2002; Catz, Kelly, Bogart, Benotsch, & McAuliffe, 2000; Duran et al., 2001; Gao, Nau, Rosenbluth, Scott, & Woodward, 2000; Gifford et al., 2000; Molassiotis et al., 2002; Murphy, Marelich, Hoffman, & Steers, 2004). Moreover, patients with little experience of HIV disease symptoms are likely to have lower adherence to dose (Gao et al., 2000).

Among the psychosocial factors related to suboptimal adherence to dose are levels of psychological distress and social support. Depressive symptoms (Catz et al., 2000; Chesney et al., 2000; Gordillo, del Amo, Soriano, & Gonzalez-Lahoz, 1999; Kleeberger et al., 2001; Murri et al., 2000; Paterson et al., 2000; Simoni, Frick, Lockhart, & Liebovitz, 2002), hopelessness (Sing et al., 1996), anxiety (Ingersoll, 2004; Molassiotis et al., 2002), HIV-related posttraumatic stress disorder symptoms (Delahanty, Bogart, & Figler, 2004) and lack of support (Altice et al., 2001; Catz et al., 2000; Duran et al., 2001; Gifford et al., 2000; Godin, Coté, Naccache, Lambert, & Trottier, 2005; Gordillo et al., 1999; Murphy et al., 2004; Murri et al., 2000; Simoni et al., 2002) have been shown to correlate to suboptimal adherence to dose instructions. Results are inconclusive as to the relationship between demographic variables such as gender, age, education level, and adherence to dose (Ammassari et al., 2002).

However, optimal ART efficacy, i.e. undetectable viral load (HIV-1 RNA < 50 copies/ml; Poppa et al., 2004), requires not only adherence to dose instructions (i.e. taking the number of pills prescribed), but also proper adherence scheduling of doses (i.e. follow instructions regarding time interval between doses) and to dietary instructions that accompany some antiretroviral medications (i.e. taking medication on empty stomach or in combination with food; Nilsson Schönnesson, Diamond, Ross, Williams, & Bratt, 2006; Nilsson Schönnesson, Ross, & Williams, 2004). In contrast to adherence to dose instructions, there is no definition of “how much” adherence to schedule and dietary instructions is needed to achieve optimal ART efficacy. Adherence to schedule and dietary instructions are usually regarded as factors influencing adherence to dose, but there is evidence that adherence to dose, schedule and dietary instructions are distinct components of medication adherence (Murphy et al., 2004; Nieuwkerk et al., 2001; Nilsson Schönnesson et al., 2004, 2006; Stone et al., 2001). While there seems to be consensus that adherence to schedule and dietary instructions are not as critical as adherence to dose, a recent longitudinal study (Liu et al., 2006) indicates that long-term viral suppression requires consistent and high percent dose adherence accompanied by optimal interdose intervals. To our knowledge, no one has examined the relationship between the above reported medication-related and psychosocial-related risk factors and suboptimal adherence to schedule and dietary instruction. Thus, the purpose of this exploratory study was to assess the degree of suboptimal ART adherence to dose, schedule, and dietary instructions and to examine the effects of demographic characteristics, medication-related factors (clinical characteristics, medication regimen), and psychosocial related factors (psychological distress, social support) on adherence across the instructions. From a theoretical perspective, these factors correspond to the three person-related domains of the biopsychosocial combined-extended Health Beliefs Model (Begley, 2004), although we do not attempt to test the theoretical model here. The model offers a general conceptual integration that incorporates the essence of the biomedical, communications, and extended health beliefs (cognitive processes and emotional barriers) models and thus encourages an examination of variables from intra-personal, inter-personal, and extra-personal domains. The extra-personal domain is a patient’s demographic characteristics, clinical characteristics, and medication regimen. The intra-personal domain is a patient’s psychological and cognitive factors (although not measured here) factors. The inter-personal domain consists of factors found in the patient’s social environment.

Method

Participants and Procedures

During a 6-month period (November 2000 thru April 2001) patients visiting two HIV clinics in Stockholm, Sweden for their routine 3-month immunological and virological check ups were invited to participate in the study. Clinic nurses approached patients who were receiving antiretroviral therapy for at least 6 months, were not using illicit substances, were not suffering from a diagnosed psychiatric disorder or dementia, and were fluent in Swedish. Patients who agreed to participate in the study were fully informed and asked to sign an informed consent form. Seventy-four percent of the patients approached (203/274) agreed to participate.

Once informed consent was obtained, patients were given a self-administered questionnaire that could be completed in the clinic or taken home and completed at a later time. It should be noted that at the time of the questionnaire completion the patient had no information about the test results of CD4 cell counts and HIV1-RNA (viral load) that were taken at the same visit as the questionnaire was handed out. Patients who completed the questionnaire at the clinic sealed it in an envelope and returned it to the clinic nurse, who forwarded it to the principal investigator. Patients completing the questionnaire outside the clinic mailed it in a pre-stamped envelope to the principal investigator. No reimbursement was given. Clinical data were obtained by a review of patients’ medical records. Data from the medical records were collected by research nurses at the two HIV clinics.

Measures

All data, except clinical data (AIDS-diagnosis, length of time on ART, HIV-1 RNA, and CD4 cell counts), were collected using a self-administered questionnaire. All measures were recoded into bivariate measures.

Extra-personal Domain

Demographic characteristics measured were gender, age, and education level. The mean was used as a cut point for age. Education was dichotomized into “university” versus “non-university education”.

Clinical characteristic measured was HIV disease stage (AIDS vs non-AIDS based on AIDS-defining illnesses), CD4 cell count and HIV-1 RNA level at the time of data were collected. HIV-1 RNA < 50 copies/ml was used as cut point for viral load. CD4 cell counts of less than 200 was used as cut point.

Variables related to medication regimen were self-reported and were confirmed with the patients’ medical records. The degree of agreement was 96%. No changes were made where there were differences between self-report and medical records. Patients were asked if they were taking any of the following drugs: nucleoside reverse transcriptase inhibitors, NRTI, (Zidovudine, Didanosine, Zalcitabine, Stavudine, Lamivudine, Zidovudine + Lamivudine), nonnucleoside reverse transcriptase inhibitors, NNRTI, (Efavirenz, Nevirapine), and protease inhibitors, PI, (Indinavir, Nelfinavir, Ritonavir, Amprenavir, Saquinavir, Lopinavir/Ritonavir). Based upon the medications reported, patients were put into one of five drug class combinations: NRTI/PI, NNRTI/PI, NRTI/NNRTI, and NRTI/NNRTI/PI and NRTI alone. The measure PI-based medication regimen was recoded to reflect presence or absence. Number of pills prescribed daily was measured by asking the patients to fill out for each drug of their medication regimen the number of pills to be taken each time and how many times a day. The mean was used as a cut point for number of pills prescribed daily.

Adverse medication side effects were measured by asking the patient “Do you currently experience any side effects from your HIV drugs?” (yes/no).

Intra-personal Domain

Three psychological constructs were used to measure psychological distress. The timeframe of all standardized psychological measures was seven days prior to the interview. The Anxiety subscale of the Brief Symptom Inventory (Derogatis & Melisatanos, 1983; α = 0.84 in this sample, mean 0.64, median 0.33, SD 0.74) was used to measure anxiety symptoms. Items were measured using a five-point Likert scale that ranged from 0 = not at all to 4 = extremely. The anxiety score was computed as the mean of at least five of the six non-missing items (n = 192). The normative mean (0.35) of anxiety symptoms of a non-patient sample was used as the cut off score (Derogatis & Melisatanos, 1983).

Hopelessness was measured using the Beck Hopelessness Scale (BHS; Beck, Weissman, Lester, & Trezler, 1974; α = 0.87 in this sample, mean 4.85, median 3.00, SD 4.09). Items were scored true, 1, or false, 0, and summed across items. In the case of missing response for two or more items the BHS score was regarded as missing data (n = 21). Level of hopelessness was recoded using a cut off point indicating a level of no to mild hopelessness versus moderate to severe hopelessness, the latter meeting the clinical level of hopelessness.

Post-traumatic stress disorder (PTSD) symptoms related to HIV diagnosis was measured using the Impact of Event Scale (IES; Horowitz, Wilner, & Alvarez, 1979). The scale measures current (within the last week) symptoms of intrusive thoughts and avoidant thoughts and behaviors of PTSD, although this scale is not designed for the identification of cases for PTSD diagnoses (Sundin & Horowitz, 2002). We specifically worded the questionnaire so that it related to HIV infection as the traumatic event that was examined in the present study (e.g., “did you think about HIV when you did not want to?”). Studies show that even years later, HIV diagnosis can result in significant PTSD symptomatology (Delahanty et al., 2004, Kelly et al., 1998). The scale contains 15 questions, rated on a four-point scale with 0 indicating not at all, 1 indicting rarely, 3 indicating sometimes, and 5 indicating often, giving a total score from 0 to 75 points with higher scores indicating more severe stress reactions. Although a diagnosis of PTSD cannot be based entirely on an IES score, scores above 35 are considered to be highly predictive of PTSD (Sundin & Horowitz, 2002). Items were summed to provide a total score, with higher scores indicating more severe PTSD symptoms (α = 0.92 in the present study, mean 23.02, median 19.00, SD 18.05, score range 0–68). In the case of missing response for one or more items the IES score was regarded as missing data (n = 4). Based on the original cut-off points, patients with scores of 35 or above were defined as having HIV-related PTSD symptoms.

Inter-personal Domain

Patient-provider relationship was measured by summing the responses to three statements: “My doctor understands me;” “I trust my doctor;” and “My doctor spends enough time with me” (α = 0.86, n = 192, mean 2.65, median 3.00, SD 0.57). Responses for each item were recorded using a scale ranging from 0 = don’t agree at all to 3 = agree completely. Measure of patient-provider relationship was dichotomized into “fully agree” (“agree completely”) and “less than fully agree” (“agree”, “partly agree”, and “don’t agree at all”).

Emotional social support was measured by the statement: “I get emotional support from my closest network in taking my HIV medication” (n = 159, mean 2.26, median 3.00, SD 0.90). Responses were recorded using a scale ranging from 0 = don’t agree at all to 3 = agree completely. Measure of emotional social support was dichotomized into “fully agree” (“agree completely”) and “less than fully agree” (“agree”, “partly agree”, and “don’t agree at all”).

Partner status was dichotomized into “partner” versus “single”.

Dependent Variables

Adherence to ART dose, schedule, and dietary instructions in the four days before the interview were measured using a modified version of the AACTG Adherence Instruments (Nilsson Schönnesson et al., 2004). Adherence was measured for each medication a patient was taking.

Adherence to dose instructions was measured by asking patients to report how many times they missed to take any of their HIV medication in the four days before responding to the survey. As the number of patients who reported that they had missed any dose of prescribed medication was low (n = 23, 12%), adherence to dose was scored as a dichotomous variable. A patient was scored adherent to dose instructions if they reported taking all reported medications ≥95% of the time. A patient was scored as suboptimal adherent to dose instructions if they reported taking all reported medications ≤94% of the time.

Adherence to scheduling instructions was measured by the question, “Most anti-HIV medication need to be taken on a schedule, such as “2 times a day” or “3 times a day” or “every 8 hours.” Have you altered the schedule of any of your HIV drugs over the last four days?” If the response was yes, patients were asked to report how often they had altered the schedule using a 5-point Likert scale, ranging from never to always. To arrive at a composite scheduling score, responses were summed across all medications and divided by the total number of medications. Four categories of adherence to schedule instructions were initially calculated: 100% adherence to schedule (63%), high adherence to schedule (18%), low adherence to schedule (12%), and non-adherence to schedule instructions (7%). Because the distribution of the responses on adherence to schedule was skewed, the measure was dichotomized to adherence to schedule instructions (100% adherence to schedule) and suboptimal adherence to schedule instructions (high, low, and non-adherence).

Adherence to dietary instructions was measured by first asking, “Do any of your anti-HIV medications have special food instructions, such as “take with food” or “on an empty stomach” or “with plenty of fluids”?” If the response was yes, patients were asked to fill out which medications were prescribed with dietary instructions and to rate how often they had not followed dietary instructions over the last four days using a 5-point Likert scale, ranging from never to always. Because of a low number of responses in the two lowest categories, often and always were collapsed with the response “sometimes”. Patients were then assigned to one of three categories: 100% adherence to dietary instructions (42%), high adherence to dietary instructions (42%), and low adherence to dietary instructions (16%). This measure was dichotomized to adherence to dietary instructions (100% adherence to dietary instructions) versus suboptimal adherence to dietary instructions (high and low adherence).

Data Analyses

Ten patients did not return a completed questionnaire. Analyses were conducted on data from the remaining 193 participants (95% return rate). Over two thirds (69%) filled out the questionnaire at home and returned it within a week. There were no significant differences on any of the three adherence measures between those who returned the questionnaire by post and those who completed it in the clinic.

To examine the associations between variables in each group of the person-related domains and the outcome variable of adherence to the three types of instructions, stepwise multivariate logistic regression equations were calculated. Dosing frequency was excluded due to its low variance. Multi-collinearity among the predictor variables was examined using the variance inflation factor. The effect of multi-collinearity was found to be fairly low (1.22–2.64).

We analyzed the predictor variables within the three person-domains in bivariate fashion evaluating each independent variable, one at a time, for its association with the outcome variables, on the basis of an odds ratio and its 95% confidence interval. Only variables showing a significant association with adherence to the three types of instructions, respectively, as indicated by having an odds ratio 95% confidence interval that excluded 1.0, were included into a multivariate logistic model.

The associations between adherence and the three types of instructions were examined using Spearman’s correlation.

Results

Descriptive Results

Extra-personal Domain

All patients in the study were Caucasian Swedes. 75% were males, 25% females, and 61% of patients identified as bi-homosexual. Mean age was 43 years (median 42, SD 8.93, age range 24–72 years), and 60% of the total sample were married or partnered in a relationship like marriage. Over a half had a university degree or professional training. Compared to the distribution of HIV sexual transmission routes among Swedes with HIV infection, male-to-male transmission and women are somewhat over-represented in this sample.

All patients were infected with HIV through sexual transmission and most had tested positive for HIV more than 10 years before they were enrolled in the study (length of time since HIV diagnosis ranged from 6 months to 17 years, SD = 4.58). About a fifth of patients (18%) were AIDS-diagnosed. Eighty-nine percent of patients had undetectable viral load (<50 copies/ml plasma, mean 307, median 50, SD 2285, range 50–31100) and 93% had CD4 cell counts higher than 200 (mean 532, median 498, SD 253, range 80–1430) at time of data collection.

At the time patients completed the questionnaire the vast majority was on a twice-daily medication regimen. The patients had on average been on ART for 47 months (range 5–96 months, median = 51, SD = 11.96) and were taking an average of 2.8 drugs at the time they completed the questionnaire (range 1–6, median = 3, SD = 0.83). Forty-six percent of the patients were taking three drugs simultaneously, 37% were taking two, 13% were taking four, and 3% were taking five or six. Only 1% reported monotherapy. The average number of pills patients were taking per day was 10.1 pills (range 2–26, median = 9, SD = 4.81). The most common drug class combinations were NRTI/PI regimens (n = 102, 53%) followed by NNRTI/NRTI (n = 53, 28%), and NRTI/NNRTI/PI (n = 12, 6%). Twelve percent of patients (n = 23) were prescribed a NRTI drug alone. Three were prescribed a NNRTI/PI combination. Sixty percent of patients were on a PI-based medication regimen. Over two thirds (n = 133) reported medication adverse side effects.

Intra-personal Domain

At the time of data collection close to one half (48%) of the patients reported anxiety symptoms above the norm. Twenty-three percent of the patients scored moderate to severe hopelessness and over one fourth (28%) reported HIV-related PTSD symptoms.

Inter-personal Domain

Close to two thirds of patients (61%) experienced a very trusting relationship with their physician. About one half (52%) was very satisfied with the emotional social support they received from their social network.

Adherence

Twenty-three patients (12%) reported suboptimal adherence to dose and seventy-two (37%) reported suboptimal adherence to schedule instructions in the four days prior to filling out the questionnaire. One hundred and twenty-eight patients were prescribed antiretroviral medications that had dietary instructions. Twenty-five patients were unaware of dietary requirements for their medication. Among the 103 patients who were aware of dietary instructions, over one half (n = 60, 58%), reported suboptimal adherence to dietary instructions. Weak, but significant associations were found between adherence to dose and schedule instructions (r = 0.25, P < 0.01) and between adherence to schedule and dietary instructions (r = 0.22, P < .05). Adherence to dose and dietary instructions were not associated.

Effects of Extra-, Intra-, and Inter-personal Factors on Suboptimal ART Adherence to Three Types of Instructions

As shown in Table 1, no factors in the inter-personal domain were significantly related to suboptimal adherence across any of the instruction types. More than ten pills prescribed daily (extra-personal domain), anxiety symptoms and hopelessness (intra-personal domain) were significantly related to suboptimal adherence to dose instructions (Table 1).

Table 1 Bivariate odds ratios of extra-, intra-, and inter-personal factors related to suboptimal ART adherence to dose, schedule and dietary instructions
Full size table

Two variables in the extra-personal domain were related to suboptimal adherence to schedule instructions, including PI-based medication regimen and more than ten pills prescribed daily. Older age and high HIV-related PTSD symptoms (intra-personal domain) were associated to adherence to schedule (Table 1).

Only one variable, less than 10 pills prescribed daily, was related to suboptimal adherence to dietary instructions (Table 1).

In the multivariate logistic analyses, the best predictive model for suboptimal adherence to dose instructions included only anxiety symptoms. The odds of those who had anxiety symptoms above the norm to be suboptimally adherent was five times greater than the odds for those with anxiety symptoms below the norm (OR = 5.507, 95% CI = 1.787–16.968).

Three factors were significantly related to suboptimal adherence to schedule instructions. Patients prescribed more than ten pills daily had almost four times greater odds of being suboptimally adherent (OR = 3.656, 95% CI = 1.918–6.969). Older age (OR = 0.508, 95% CI = 0.269–0.958) and HIV-related PTSD symptoms (OR = 0.316, 95% CI = 0.146–0.683) were found to decrease the odds of being suboptimally adherent to schedule instructions.

Discussion

Our data show that the percentage of suboptimal adherence varied considerably across type of medication instructions, from 12 to 58%. While the low number of patients reporting suboptimal adherence to dose instructions was similar to rates reported in other studies (Cederfjäll, 2002; Jensen-Fangel et al., 2002; Kleeberger et al., 2001; Molassiotis et al., 2002; Walsh, Horne, & Dalton, 2001), suboptimal adherence rates to schedule and dietary instructions were higher in our sample than those reported by Murphy et al. (2004). Patients quite often express that their medication controls their lives and the difficulty of integrating medications into their daily activities as a major reason of suboptimal adherence (Ammassari et al., 2001; Chesney et al., 2000; Gifford et al., 2000). Within such a context, it is quite possible that suboptimal adherence to schedule and dietary instructions is a coping strategy to regain a sense of control. Patients may perceive that suboptimal adherence to schedule and dietary instructions is less likely to jeopardize the effectiveness of ART than missing doses.

In contrast to Simoni et al. (2002), our data indicated a weak association among adherence to the three types of instructions. It provides support for the idea that adherence is comprised of multiple, discrete behaviors that may not result from common antecedents (Ingersoll, 2004).

A lack of dietary instructions information, or the patient’s perception that they do not have this information, may be a manifestation of regimen misunderstanding or regimen “reinterpretation”. This in turn could be a reflection of a failure of the health care staff to provide sufficient or understandable patient education. The finding confirms the importance of educational interventions at the appropriate level to enhance patients’ knowledge and understanding about their medication not only at the initiation of the medication but repeatedly during the course of treatment.

HIV-1 RNA level was not found to be associated with adherence in this study, a finding similar to others (Simoni et al., 2002). The high proportion of patients in this sample with an undetectable viral load suggests high past adherence. It is also important to note that reported adherence during a 4 day period prior to the measured HIV-1 RNA level does not influence the HIV-1 RNA level measured on the study day, since it takes at least 7–14 days for HIV-1 RNA level to react to suboptimal medication.

Multivariate analysis of factors associated with suboptimal adherence to dose instructions indicated that only one significant risk factor emerged, namely anxiety symptoms. This finding is consistent with other studies (Ingersoll, 2004; Molassioti et al., 2002). When anxiety symptoms begin to interfere with the patient’s daily life, the symptoms may take precedence over and present a barrier to adherence to dose instructions.

As to adherence to schedule instructions, age was a protective factor of suboptimal adherence to schedule, whereas patients who were prescribed more than ten pills daily were significantly more likely to report suboptimal adherence to schedule instructions. Unlike Delahanty et al. (2004), HIV-related PTSD symptoms emerged in this study as a protective factor of suboptimal adherence to schedule instructions. It may be the case that some level of PTSD symptoms may function as a motivator for adhering to schedule instructions due to self-protective alertness about following medication regimen. On the other hand, HIV-related PTSD symptoms may be the result of being adherent to one’s medication. Clinical experiences demonstrate that some patients are extremely anxious about failing to adhere to their medication regimen. The fear is associated with developing drug resistance and/or HIV-related symptoms/diseases. The focus on and preoccupation with HIV and medication may in turn evoke and/or reinforce HIV-related PTSD symptoms. Thus, it would appear as if psychological symptoms lie on a spectrum from disabling distress to health-protective concern. However, staff should always be alert to a patient’s emotional state, as the boundary between health-protective concerns and disabling distress is small.

Consistent with past studies on adherence to dose instructions (Ammassari et al., 2002), gender, education levels, and AIDS diagnosis were not significant predictors of adherence to any of the three types of instructions. In the present study, experiencing medication side effects was not a significant predictor of suboptimal adherence, supporting Godin et al.’s findings (2005). However, it is possible that it would have reached significance if we had measured specific side effects and their perceived intensity. On the other hand, it could be the case that it is not adverse side effects in themselves that play a role in suboptimal adherence but rather their psychological consequences; that side effects associate with psychological distress and psychological distress in turn associates with suboptimal adherence. However, this is a speculation that deserves further study.

Unlike other studies (Catz et al., 2000; Eldred et al., 1998; Gordillo et al., 1999; Simoni et al., 2002; Singh et al., 1996), no inter-personal domain predictors of suboptimal adherence to the three types of instructions were identified. It is possible that inter-personal factors have an indirect link to adherence. The associations between inter-personal and intra-personal domain factors may be strong and psychological distress may be a more powerful correlate of suboptimal adherence to specific type of instructions than inter-personal factors, but this warrants further research. Another explanation might be that our social support measures were poor.

Findings from this study are limited by a number of factors. The exclusionary criteria (ongoing illegal drug use, psychiatric diagnosis, non-fluency in Swedish) and the use of a convenience sample limit the generalizability of findings to other populations. Generalizability of the findings may also be limited by selection error. This kind of study may attract those patients who do not have any major problems with or concerns about adherence. Thus patients having adherence problems may be underrepresented. Additionally, compared to the distribution of HIV sexual transmission routes and gender among Swedes with HIV infection, gay men and women were over-represented in the sample. Considering these sample limitations, our sample cannot be considered fully representative of people living with HIV and ART in Sweden. The self-reported measures of adherence may be subject to social desirability (Turner, 2002) leading to overestimation of adherence. However, as adherence data were not collected by the primary physician or nurse or in a monitored clinical trial setting, it is likely that social desirability bias was reduced. It could be argued that the past four days for assessing adherence may not provide an accurate picture of actual adherence over time. However, research clearly indicates that recall periods longer than four days are subject to higher error. The presented data are cross-sectional and it is not possible to draw definite conclusions about the causality between external- and intra-personal factors and suboptimal adherence.

Much of the work examining suboptimal ART adherence of any sort has been conducted in the US. The United States almost alone of western civilizations has no comprehensive universal health care system, which does impose an additional burden on people with HIV infection who have no or inadequate insurance or who are unemployed. The fact that the right to adequate health care and treatment is denied to many in the United States will add a layer of uncertainty and distress to the process of HIV infection, which does not occur in Sweden. We suggest that these differences between the countries may have affected the results of the current study.

This study extends past research to show that predictors of suboptimal adherence vary across type of medication instructions being measured. Suboptimal adherence to dose instructions was predicted by anxiety symptoms, whereas suboptimal adherence to schedule instructions was predicted by heavy pill burden. Age and HIV-related PTSD symptoms were on the other hand protectors against suboptimal adherence to schedule instructions. Our data also imply that suboptimal adherence is not only a behavioral pattern, but may be a symptom of psychological distress or a coping strategy to adjust antiretroviral therapy to one’s daily living. Some level of HIV-related PTSD symptoms may, on the other hand, function as a motivator for adherence to schedule instructions. In addition to further assessment of the associations between external-, intra- and inter-personal domain factors and adherence and/or suboptimal adherence to the three types of instructions, examining interrelationships among factors that affect adherence to antiretroviral therapy is warranted. The effects of factors on adherence may not always be direct but may act indirectly through other factors. To improve adherence intervention we need a better understanding of the nature of these interrelationships.