Social Support and Risk of Compulsory Admission: Part IV of the Amsterdam Study of Acute Psychiatry

This study was part of a prospective cohort study of all consecutive patients who received emergency consultations from the Psychiatric Emergency Service Amsterdam (PESA) and from the Acute Treatment Unit (ATU) between September 15, 2004, and September 15, 2006 (5). The PESA is the main decision-making body for emergency compulsory admissions in the Amsterdam area (about 743,000 inhabitants). It provides 24-hour psychiatric emergency services, seven days a week. The ATU is a specialized outpatient emergency service that covers the Centre and Old West area of Amsterdam (about 200,000 inhabitants); it provides services during office hours only. Patients are referred by family doctors, police, emergency wards at general hospitals, and mental health care workers. During office hours, ATU takes the referrals of family doctors and mental health care workers within the specific catchment area. The PESA handles police referrals 24 hours a day. Three other catchment areas in the city are served by ATUs not covered by our study.

People who come into contact with the police are first screened at the police station by a community mental health nurse from the City Health Service. Individuals in need of immediate psychiatric consultation are brought by the police to the PESA's psychiatric emergency room. Patients who are referred to the PESA or the ATU by primary health care providers, such as family doctors or doctors from emergency wards, are often seen in their homes or at the referring emergency ward. Patients themselves have no direct access to consultations without a formal referral. In general, consultations are conducted on workdays between 9 a.m. and 10 p.m. by a team of residents, community mental health nurses, and a psychiatrist. Psychiatrists play an active role during consultations, either by acting as a supervisor or by seeing patients themselves. At other times, residents and a nurse work together during the consultations, and psychiatrists provide telephone backup. The PESA and the ATU use electronic patient files.

Two groups of participants were selected at random from among the patients who had at least one consultation with either the PESA or the ATU during the study period. One group consisted of patients whose consultation resulted in compulsory admission, and the other group consisted of patients whose consultation resulted in voluntary admission to a psychiatric hospital or crisis center, in outpatient treatment, or in no further intervention. Every patient in the compulsory-admission group and every fourth patient in the other group were asked to participate in the study. For pragmatic reasons we selected only patients with a fixed abode in Amsterdam whose command of Dutch was sufficient to permit them to be interviewed without an interpreter. We excluded people living elsewhere in the Netherlands or abroad and people with no known place of residence.

We collected three types of data. First, the clinician collected data during the emergency consultation about the patient's age, gender, domestic situation, and DSM-IV-TR axis I diagnosis. We used four clusters of categories: schizophrenia or other psychotic disorder, manic state, depressive disorder, and other or no disorder. The last group included drug- and alcohol-related disorders if they were the patient's main axis I diagnosis, a wide variety of other diagnostic categories, and psychosocial problems in the absence of a DSM axis I disorder. The outcome of the consultation was also recorded.

Second, data were collected by research assistants during an interview approximately one month after the emergency consultation. The research assistants were five trainee psychologists and a team leader who graduated in human geography with experience in mental health care research. The entire process was supervised by the first author. During the interviews we counted the number of people in the patients' social network by using the Social Network Structure Questionnaire (SNSQ) developed by van Sonderen (12). This Dutch ten-item self-report inventory maps the natural networks of people (partner, family, neighbors, coworkers, and people at school) who play a role in the patient's life. The SNSQ assesses the quantitative aspect of social support. We also used the 13-item Social Resources Questionnaire from the Multidimensional Health Profile (MHP) that was developed and validated by Ruehlman and colleagues (13). This self-report screening instrument evaluates positive and negative aspects of social interaction, by including nine items that assess social support (possible scores of 9 to 45) and four items that assess negative social exchange (possible scores of 4 to 20). A score between 11 and 20 on negative social exchange implies a high level of negative interaction with family or friends.

Third, we collected data about the number of emergency compulsory admissions of each patient recorded in the Amsterdam Mental Health Care database during the two-year period after the interview.

We expected a clinically relevant difference between the voluntary and compulsory groups, and an effect size of approximately .5. To detect a difference of this kind in effectiveness between the groups (α=.05 and β=.80), 64 respondents were needed in each group. Because we expected a considerable level of dropout (50%) between the inclusion stage and the final follow-up measurement, we estimated that 125 patients per group would be required initially.

A bivariate analysis with a Pearson chi square test and an independent-samples t test for variables with normally distributed values or an independent-samples Mann-Whitney U test for variables with values that were not normally distributed was conducted to examine differences in sociodemographic factors, clinical presentation, mean sum scores on the SNSQ and the MHP Social Resources Questionnaire among patients with and without at least one emergency compulsory admission during follow-up. A bivariate analysis with Pearson chi square testing to examine the relationship between the patients' domestic situation and the number of members of their social networks was also conducted.

To prevent bias by intercolinearity in our multivariate analyses, we checked for the presence of couples of variables with a contingency or correlation coefficient greater than ±.6. To analyze the influence of effect modification, we conducted two multiple stepwise logistic regression procedures with the variables that were found by the bivariate analysis to be correlated (p≤.15) with having at least one emergency compulsory admission during follow-up.

We calculated the odds ratios (with a confidence interval [CI] of 95%) for the incidence of emergency compulsory admission during follow-up and scores on the SNSQ and MHP Social Resources Questionnaire, domestic situation, and their interaction terms. For interaction terms that proved to be significantly associated with the outcome measure, further analysis of the modification effect was conducted by means of bivariate analysis with Pearson chi square testing of the association between the SNSQ and MHP scores for patients who lived alone or with other people.

Finally, we conducted a multiple stepwise logistic regression procedure with the variables found by the bivariate analysis to be correlated (p≤.15) to develop a prediction model. First, we calculated the odds ratios and CIs for the incidence of emergency compulsory admission during follow-up by using numbers of network members as predictors. In the second step, the subscales of the MHP for social support and negative interactions were added. Domestic situation was added in the third step of the analysis.

Before introducing variables to the logistic regression analyses, we tested all possible couples of independent variables for intercolinearity by calculating the contingency coefficient for nominal-nominal or nominal-ordinal couples, eta for nominal-linear couples, Spearman's rho for linear-ordinal couples, and Pearson's correlation for linear-linear couples. We used SPSS, version 19, software.

Patients randomly selected for the study cohort received a letter with a request to participate in the study. In the follow-up to this request, patients were contacted by a research assistant by telephone, in writing, or in person at the patients' place of residence. As a result, patients received information verbally as well as in writing before they were asked to give written informed consent. The procedure was endorsed by the Medical Ethics Committee for Mental Health Care Institutions. Confidential information and participants' names were subject to medical confidentiality rules and were handled in accordance with the code of conduct for medical research developed by the Federation of Biomedical Scientific Societies.

A total of 2,682 patients had at least one emergency consultation during the study period, and a group of 954 patients, evenly divided among those who had or had not been admitted compulsorily, were randomly approached and asked to participate in the study. Of those, 317 patients (33%) could not be contacted in spite of repeated phone calls or home visits; 243 (25%) refused to participate; 99 (10%) were too ill as a result of a psychiatric or general medical condition to give informed consent; 28 (3%) could not give informed consent because they did not speak sufficient Dutch and 15 (2%), including three who died before they could be contacted, did not give informed consent for various other reasons. This left a study cohort of 252 patients, 118 with and 134 without an emergency involuntary admission. Half the patients were interviewed within six weeks of the emergency consultation, and 75% were interviewed within ten weeks. Eight patients died or left the country during the first year of follow-up, leaving a follow-up cohort of 244 patients.

The study cohort of 252 patients was compared with the 2,430 patients who fulfilled the inclusion criteria and were not included in the follow-up study. We conducted separate representativeness tests for the compulsorily admitted sample (N=118) and for the sample in which consultation did not result in involuntary admission (N=134). We compared each sample with the part of the population from which it was drawn.

Among patients admitted compulsorily, members of the study cohort were significantly younger than nonmembers (38±11 versus 43±17, p≤.001), and a smaller percentage were classified as psychotic (58% versus 70%, p≤.01). There were no significant differences between the groups in gender, country of origin, and domestic situation. Among patients who were not admitted compulsorily, the proportion of native Dutch was larger (75% versus 60%) and the proportion of non-Western immigrants was smaller (16% versus 34%) in the study cohort than in the population as a whole (p≤.001). There were no significant differences between the groups in gender, age, domestic situation, and diagnostic category.