Introduction

Psychiatric emergency services, on the front line of the on-going trend toward outpatient care, are confronted with repeat contact by a subset of patients who use/require considerable resources. Frequent repeaters (FR) have been characterized as being prone to conflict with the emergency department staff and a cause of demoralization due to their demanding yet help-rejecting attitude defying coherent treatment planning [1, 4, 13].

The burden of FR on psychiatric emergency room (PER) staff is clearly demonstrated by the proportion of contacts, between 20 and 30% in many PER, although they represent less than 10% of patients.

Frequent use of PER has received some attention in scientific literature. About 20 studies have dealt with this topic but the variety of contexts or samples (Academic hospitals, Veterans' clinic, PER in General hospitals or the level of urbanization of catchment areas), the generally small size of the FR group, index periods, definitions and methodologies, result in a mixed portrait. Moreover, if the focus on FR is justified for all the above-mentioned reasons, studies taking into account occasional repetition are scarce, and none has attempted systematic comparisons of FR, occasional repeaters (OR) and non-repeaters with a multivariate design.

Sullivan et al. [18] built groups according to the number of repeat contacts, but focused on FR to show progressive increase or decrease in the prevalence of diagnoses or socio-demographic characteristics; the number of contacts was used as a dependent variable in linear regression. Hansel and Elliott [8] clearly identified OR, but used a univariate design centered on diagnoses and disposition. Because of the specificity of the organization (VA hospital—thus almost exclusively male patients), no gender difference could be illustrated. Moreover, the limited number of characteristics used to differentiate between the three groups and the absence of logistic regression analysis failed to provide a clear picture of the OR group.

Definition of FR is empirical depending on the example studied. Saarento et al. [14] and Sullivan et al. [18] defined FR as being those patients belonging to the upper 10th percentile of the contacts. Most other studies however considered four or more contacts during the index period as the cut-off point. A recent case-control study compared n=74 patients with six or more visits to patients with five or less contacts [3]. Strict comparisons between studies are difficult because of the varying lengths of the periods considered—ranging from under 1 year, in only one of the studies, up to 22 months in another.

Factors responsible for frequent repetition of psychiatric emergency visits having emerged in scientific literature do cluster in socio-demographic, diagnostic and service utilization characteristics. FR have a greater risk of being male [8, 10, 17, 18], young or middle-aged [7, 17, 18], unmarried [12, 18], non-white [18], living alone [14]. They have been illustrated as lacking social support [2, 4, 6, 11, 14], they present a general social disability [9], they are unemployed [5, 18] or homeless [3, 5, 11].

Diagnostic characteristics of FR include chronic illness [6, 9], more serious diagnosis [14–16], psychotic disorder [5, 6, 15, 16, 18], schizophrenia [8, 18], personality disorder [6, 8, 11, 18], prominence of anxiety [6], substance abuse [6, 8, 17]. Dhossche and Ghani [5], on the contrary, have shown an inverse relationship with substance abuse, with the exception of schizophrenics.

Service use factors were previous psychiatric treatment [4, 6, 12, 17], history of voluntary and involuntary hospitalizations [11], current treatment-psychotherapy [6] and self-referral [6, 12, 17].

Disposition factors were contradictory with Hansel and Elliott [8] showing FR as being less likely to be referred to an outpatient clinic for follow-up. For Ellison et al. [6], FR were more often referred for outpatient treatment.

These predictors are scattered in various studies, and a comprehensive analysis with a multivariate approach is lacking. To our knowledge, OR predictors have not yet been addressed in scientific literature.

The objectives of the present study are (1) to provide a naturalistic evaluation of patients repeating admissions in a psychiatric emergency ward (distinguishing between OR and FR), (2) to identify patients' characteristics that predict repeated use of a PER and (3) to propose adapted treatment models.

Materials and methods

Brugmann Hospital is a public hospital located in the north of the Belgian Brussels, Capital Region (950,000 residents), with about 22,000 patient contacts per year. The psychiatric ward (120 beds) is organized with the following specialized treatment units: emergency crisis unit which admits patients for a limited period of 72 h (6 beds), alcohol dependence, personality disorder, affective disorders, psychotic disorders, sleep disorders and day-care unit. The emergency unit has an annual case-load of about 3,000 contacts per year. Temporary hospitalization for evaluation (12 beds) is also provided.

All patients were identified by name and birth date for unambiguous coding of each repeated contact during the index period. This study defines FR patients as being those patients who repeat contact four times or more during an index period of 16 months. This corresponds to the upper 15th percentile of contacts.

An extensive psychiatric emergency contact form was systematically used by a psychiatrist and a psychiatric nurse to screen all contacts. In an emergency ward, it appears unrealistic to expect a thorough use of DSM-IV. This study relies on the clinical judgment modeled on the axis 1 of the DSM-IV. Principal and secondary diagnoses corresponding to DSM-IV (axis 1) were initially limited to 12, but notification of other diagnosis was allowed. Most recurrent among these were later coded separately. Use of a primary and secondary set of categories is useful in the light of a tendency to shift some diagnoses to a secondary level, as with substance misuse (e.g. alcohol or opiate dependence), while other dimensions are chosen on an initial diagnostic level. Diagnoses such as major depression, alcohol dependence or non-active and active psychosis coded as principal or secondary were also coded in a single category, thus specifying the presence of any of these diagnoses. Schizophrenia was subdivided into active (positive symptoms) and non-active psychosis (without positive symptoms). The role of such a sub-division (although not mentioned in the DSM-IV) emerges from clinical assessment and contributes to the results of this study.

Stressors include housing, financial, study, accident, grief and all types of conflict (with parents, children, partner, another member of the family or outside the family).

Data were processed using SPSS v.11. Cross tabulations with Mantel–Haenszel statistics or ANOVA tests were used in univariate procedures and Forward Conditional Logistic Regression for multivariate analysis.

Results

In this study, frequent repeaters (FR) are defined as patients with 4 contacts or more during our index period of 16 months, and occasional repeaters (OR) as those with 2 to 3 contacts. If administrative considerations tend to focus on a 1-year period, the present study has indeed extended it to 16 months—January 1998 to April 1999—to allow a retrospective collection of a sample of n=100 FR (4% of all n=2,470 patients). They contributed to n=544 contacts (15.5% of all 3,511 contacts during the period). Consequently, OR represented n=479 patients (15.3%) and had n=1,074 contacts (30.6%). Total repeaters thus represent 46.6% of all contacts and 23.4% of patients.

Univariate analysis

Table 1 compares the three groups of patients under study using simple contingency tables and details socio-demographic characteristics: Men represent 54% of all patients and 70% of FR. Compared to all other patients and to OR, FR are significantly male. Mean age of patients is 37.5 years (±13.7). FR (33.6±9.2) are younger than all others but not younger than OR.

Table 1 Socio-demographic characteristics of repeater groups (in %)
Full size table

Less than one quarter of patients have a professional income. Differences between OR and non-repeaters are significant. Most patients depend on welfare with an important group (17.9%) receiving public health insurance allowance (a status obtained in Belgium after at least 1 year of sick leave). Among OR and FR, sick leave reaches 24%. This would tend to reveal a « sick role » endorsed by repeaters. Even in a total patient population facing professional hardships, FR are clearly the most socially disabled. The total absence of income presents a reverse trend with 10.5% of non-repeaters, 7.1% of OR and only 4% of FR. This is probably due to the presence of non-working women or young people still living at home in this category. A majority of 56% of women and 50% of under 26-year-olds are indeed present in this group of patients with no income.

The five principal diagnoses of patients at first contact in decreasing order are as follows: alcohol dependence (19.6%), adjustment disorder (15.9%), active psychosis (positive symptoms) or schizophrenia (14.4%), major depression (13.9%), non-active psychosis (without positive symptoms: 6.2%). The only significant principal diagnosis differentiating repeater groups was non-active psychosis. Combined first and second level diagnoses differentiated repeater groups under the heading of any psychosis (23.1% of patients, 21.5% of non-repeaters, 28% of OR and 30% of FR with significance for comparisons of OR and non-repeaters and also of repeaters and non-repeaters).

Half of the patients have a specific stressor. A conflict (within family or not) was the main stressor present for 25.7%: 13.6% had a conflict with their partner, parents (5%), child (2.5%), another family member (5.4%) or outside the family (1.8%). Other prevalent stressors are separation (essentially from partner) for 9.7%, problems at work (5.6%), or financial difficulties (5.1%). Only two (less prevalent) stressors proved to differ between repeaters: grief with 7% of FR, 2.4% of non-repeaters and 1% of OR; and craving with 3% of FR, 0.5% of non-repeaters and 0.2% of OR.

Self-referral concerns about half of the patients (49%). Almost two thirds of FR (65%) are self-referred to PER, 54.7% of OR and 46.6% of non-repeaters. Family (12.7%) or friend (2.6%) referral is similar across repeater groups. Non-repeaters (9.2%), OR (6.5%) and FR (3%) are referred by a general practitioner.

The main arrangement for patients after their first contact is distributed as follows: 11.7% were sent back home without any treatment follow-up, 29.4% received outpatient appointments, and 16.8% were oriented to the crisis unit of the same psychiatric department.

Two orientations differentiate repeaters: The absence of any follow-up treatment (home return) is only significant for FR (19%) compared to other patients.

Multivariate analysis

Table 2 presents odds ratios (Odr) with confidence intervals and statistical significance of forward conditional logistic regression of repeaters (total of OR and FR: n=575) compared to non-repeaters (n=1,870). All dichotomized characteristics were entered in a stepwise forward conditional logistic regression to build a predictive model of contact repetition at first contact.

Table 2 Logistic regression of repeaters
Full size table

Repetition is predicted by younger age (Odr 0.99), lowest welfare status (Odr 1.51) and sick leave (Odr 1.66). Lack of income (Odr 0.65) is inversely related to repetition. The risk of living in a non-psychiatric institution is greater for repeaters (Odr 2.13). All these socio-demographic characteristics draw a picture of greater social disability and dependence. The only stressor predicting repetition is conflict with child (Odr 1.96).

Diagnoses at first contact that positively predict repetition are any psychosis (Odr 1.90) but also the absence of psychiatric diagnosis (Odr 1.83) or postponed diagnosis (Odr 1.70) and mental retardation (Odr 4.39; p=.04). Inversely major depression (Odr 0.49) predicts non-repetition of contact. Prescribed methadone (Odr 2.46) is a strong predictor and misuse or dependence of alcohol (Odr 1.55) a less intense one. Substance and social dependence are thus strong predictors of contact repetition. Previous hospitalization (Odr 0.81) is inversely related to repetition. Finally, service utilization with self-referral (Odr 1.41) is a positive predictor of repetition, while on the contrary, referral to another hospital (Odr 0.39) at first contact predicts non-repetition (Table 3).

Table 3 Logistic regression of occasional repeaters
Full size table

FR are analyzed in Table 4 by reference to all other patients and also to OR. Predictors of heavy use of PER are somewhat different from those for OR. To be male (Odr 1.85), younger (Odr for age 0.97) with a low welfare status (Odr 2.35) increases the risk of frequent contact repetition. Grief as a stressor (Odr 5.56) appears very significant. Two diagnoses are selected in the equation: depressive disorder as a complementary diagnosis (Odr 10.56) and non-active psychosis (Odr 2.30).

Table 4 Logistic regressions of frequent repeaters
Full size table

Self-referral (Odr 1.99) and disposition by temporary hospitalization at PER (Odr 2.95) and home return without any treatment proposal (Odr 1.83) increase the risk of frequent repetition.

Further multivariate analysis of contact repetition has been performed according to sex. For FR to be younger (Odr 0.97) and on low welfare (Odr 2.55) predicts frequent contact for men. Grief (Odr 6.08) is a powerful predictor and craving (Odr 5.41) is also significant. Self-referral (Odr 2.20) and temporary hospitalization at PER (Odr 2.50) are also selected as predictors. Female FR are characterized like men by Grief (Odr 4.05), but other predictors are different: complementary depressive disorder with a massive Odr of 57.78, pharmaceutical drug misuse (Odr 4.94) and heroin dependence (Odr 4.51).

Discussion

Size of the problem

In other studies, FR accounted for a higher percentage of patients (up to 8%) and contacts (up to 30%). The difference with the present study cannot be explained by the accounting of cases (in some research, cases were not identified by name), as the full verification of repeaters would have the opposite effect. We are left to hypothesize some hidden detail of the organization or the populations under study (strong commitment to reduction of hospitalization by the PER staff). Another explanation might involve the existence of mental health services in the community providing the type of help sought by repeaters.

Predictors of repetition

Results confirm previous research on FR while adding some new insights and predictors. Repeaters are younger and have an unambiguously lower social profile than non-repeaters. Not only do they depend on welfare, but they are also more often on sick leave and tend to have a greater risk of living in an institution. A parallel can be drawn with repeaters' higher risk of methadone or alcohol misuse or dependence illustrating a dependent profile for these patients. As with other repeater subgroups, the characteristic type of admission is self-referral. This has already been proposed in scientific literature and appears in our analyses as a steady constant. It can be interpreted as being a sign of greater impulsiveness and a stronger urge to seek help, combined with another predictor, the inverse probability of a recent hospitalization, it could be suggested that repeaters appear, albeit falsely, as ‘new’ users of health services, due to the fact that they rarely follow-up on a treatment plan. For PER staff, this lack of previous professional evaluation added to an ambiguous symptom profile is a source of difficulty: repeaters have a higher risk of having their diagnosis postponed or not receiving any psychiatric diagnosis. This is however not true for all repeaters as revealed by a higher probability of psychosis or mental retardation. This combination of lack of diagnosis and severe diagnoses is an indicator of the difficulty of PER staff to deal with this population. The result is consequently an absence of hospitalization or treatment planning, even if the presence of a psychosis diagnosis should have increased the probability of such a provision.

Predictors of occasional repetition

For the subgroup of OR, predictors (see Table 3) are in line with the total repeater group, but in a somewhat more limited scope. OR are characterized by their sick role status, lack of active social participation and multiple drug and alcohol dependence. Self-referral proves to be a constant predictor for repeater subgroups. Psychiatric characteristics provide a complex of severe diagnosis with psychosis and secondary personality disorder, but also a lack of any diagnosis. Comments made above on this duality apply for OR.

Gender specificities of occasional repetition

To obtain a more detailed portrait of OR, separate analyses according to sex and age for OR indicate that specific predictors for OR men (compared to other men) show a greater probability of conflicting with staff planning by leaving the unit without notice, but also a greater disposition to receive some form of treatment, albeit not by referral to another hospital. Men OR have thus a greater likelihood of receiving planned treatment (expressed inversely by the lesser probability of returning home without treatment), which is an exception in this study's results. Although men and women OR are both more probably psychotic, there is a trend for women to have more negative symptoms, to be diagnosed with mental retardation and to have not received psychiatric diagnosis. The already noted ambiguity of diagnosis thus describes more specifically women patients. The other main difference lies in a different type of drug misuse: OR women are more alcohol dependent than other women, and pharmaceutical drug dependence is more probable for OR men. These are quite specific characteristics for OR compared to the prevalence according to sex of alcohol problems (more men) and pharmaceuticals (more women) in the general population.

Predictors of frequent repetition

FR are more likely to be younger males and socially disabled. They manifest a greater risk of being in a state of grief, even compared to OR with whom they also have a tendency to exhibit more family conflict and craving. FR are also more likely the same as OR, to have a substance abuse problem with pharmaceuticals. FR as OR have a higher risk to be self-referred. This tends to confirm their personal high demands of treatment. The trigger role of grief is remarkable and totally unique for FR. In the absence of any study on this stressor for PER patients, interpretation should be cautious. It reveals nevertheless a specific fragility and cause of suffering. More probable diagnoses of FR are non-active psychosis and a secondary depressive disorder. Compared to OR, this last diagnosis of secondary depressive disorder is the only differentiating factor. FR exhibit thus more unspecific depression-like complaints than OR. Although absence of diagnosis is not a predictor, per se, FR also have characteristics that do not provide PER staff with a straightforward clue as to how to react at first contact. Temporary hospitalization for evaluation at PER and absence of treatment planning illustrate this vividly.

Time elapsed between first and second contacts

As noted by Hansel and Elliott [8], we confirm in this study a shorter period between first contact and return of FR than OR. FR returned to PER 47.7 days (±78.1) after their first contact, while for OR, 74.3 (±89.5) days elapsed (p<.01). A short period for the first return to PER can thus alert staff to an FR's profile. It underlines also the greater impulsivity of these patients.

Gender specificities of frequent repetition

Predictors for men FR compared to other men include younger age, social disability, self-referral and temporary hospitalization and are absent from the model for women. Both genders have grief, predicting frequent repetition (with a more significant odds ratio for men), and exhibit some form of substance misuse. Diagnosis is a differentiating predictor of frequent repetition with non-active psychosis for men and secondary depressive disorder for women, confirming for FR the observation made for OR of a less-structured symptom profile for women.

A complex picture

This study confirms the social disability of repeaters and the role of psychosis, especially for men and younger patients. A more complex picture is however emerging. Firstly, non-active psychosis or disorganized-residual schizophrenia is the main type of psychosis concerned, and secondly, other diagnoses involve a less severe or structured psychopathology than previously proposed in scientific literature. Indeed, results support a new hypothesis of a general unspecified complaint in a socially deprived context with substance misuse and grief triggering frequent repetition of contact. Although both groups of repeaters have an ambiguous symptom profile—more so for women and older patients—there is some continuum from OR to FR in the depressive intensity of this non-structured complaint with grief as a very specific stressor for FR.

Many FR exhibit a less structured symptom profile than most patients leading them to seek help in an easily accessible health service with low demands such as PER. An important aspect of the emergency room situation, critically underlined by Bassuk and Gerson [4], is its impersonal and transient nature. But, paradoxically, the loosely structured care environment is favored by these patients who are undecided about their commitment to treatment. Although their complaints and suffering seem to be recognized, the absence of any clear-cut symptom profile explains the difficulty for PER staff to efficiently dispose of this group of patients. A disappointing result is the lack of efficiency of temporary hospitalization at PER, being on the contrary a predictor of frequent contact repetition. Warwick and Solkovskis [19] have proposed two separate, but related processes leading to medical contact: (a) handicap, pain or inconvenience caused by a complaint and (b) worry and fear of threats that could arise from not seeking professional (medical) help. Repeaters of PER contact thus feel, like other patients, a strong need for professional help stimulated by the major inconvenience caused by their complaint. The non-specific nature of this complaint does not, however, elicit a precise reaction from the PER staff. This in turn creates new fears for the patient who attempts to seek help again. This cyclical process increases the patient's dependency. Indeed, the already existing social and substance dependence of these patients constitutes a favorable ground for a revolving door use of PER. Although it was not possible to verify whether or not these patients also attempt to seek help elsewhere in addition to PER, we would hypothesize that they did not. Further studies should empirically address this issue.

Toward a prevention of repeated contacts

Can this revolving door process at PER be broken? Results of this study show that the professional staff, in reaction to their difficulty in clarifying and meeting the patient's demand, pave the way for the patient's return by not planning for treatment. Another reaction would be more helpful. If, as we have shown, self-referred socially deprived patients with substance misuse and a lack of specific symptoms are at high risk of contact repetition, this group should be addressed more straightforwardly and referred to services who are equipped to deal with this kind of population. The development of crisis units in psychiatric hospitals has been efficient in reducing hospitalization rates. To reduce the burden of repetition of contacts at PER, we would advocate the development of a program specialized in non-specific mental health pathologies for patients with characteristics as detailed in this study. If such services already exist in the community, they could network with PER for this type of patient. This proposal should of course be put to the test empirically to demonstrate its effectiveness in reducing heavy use of PER