Rate and Predictors of Service Disengagement Among Patients With First-Episode Psychosis

The sample consisted of consecutive patients accepted by EPIP between 2001 and 2009 with two years of follow-up data. Patients in the sample met the following criteria: age between 15 and 40, a first episode of psychosis with no or minimal prior treatment, no current history of substance abuse or forensic involvement, and no history of major medical or neurological illness. Medications were prescribed on the basis of a standardized treatment algorithm that emphasizes use of antipsychotic monotherapy at a low dosage and adjunct medication (mood stabilizers or antidepressants) when necessary. Patients were assessed and managed at regular intervals by a multidisciplinary team of psychiatrists, case managers, psychologists, social workers, nurses, and occupational therapists.

Every effort was made to engage patients in the service by repeated phone calls and home visits. Even when patients refused medications, the team continued to engage them by offering help in other areas, such as employment and financial assistance. After two years, patients were discharged from EPIP and transferred to mental health services in the community (general practitioners and regular mental health clinics). The use of patient data for this study was approved by the institutional review board of the National Healthcare Group.

Sociodemographic data were recorded and stored in an electronic database. Case managers used a semistructured questionnaire to assess employment status (including age-appropriate employment, such as student or homemaker). Duration of untreated psychosis was defined as the time in months between onset of psychotic symptoms and establishment of a definitive diagnosis and treatment. Patients and their primary caregivers were asked to date the onset of psychotic symptoms, and duration of untreated psychosis was estimated by combining information from the interviews and case records. If discrepancies were found between patients’ and caregivers’ reported onset of psychotic symptoms, the patients’ reports were favored.

The Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) was used to establish a diagnosis at first contact with patients (9). Patients were also assessed at baseline and three, six, 12, and 24 months with the Positive and Negative Syndrome Scale (PANSS) for schizophrenia (10) and the Global Assessment of Functioning (GAF) (11). Ratings were conducted by five experienced psychiatrists trained in the use of the rating instruments. All raters participated in periodic interrater reliability sessions to avoid rater drift. Using the intraclass correlation coefficient, interrater reliability for PANSS was assessed to be .94. Reliability statistics for the SCID-I and GAF were not available.

A semistructured scale was used to assess service disengagement at two years: did not disengage; telephone contact with the patient or a family member or both; telephone contact with a family member only; and no contact. On the basis of EPIP’s service model of continued engagement with patients even when they refuse medication and criteria used in previous studies to indicate active contact between a patient and a treatment team (4,5), only the latter two ratings (telephone contact with a family member only and no contact) were taken to indicate patient disengagement. Patients who returned to EPIP within two years of dropping out were considered to have remained in the service. Assessment of service disengagement was conducted by 15 case managers who were trained in the use of the scale.

Study variables were selected a priori on the basis of predictors identified in previous studies, including age, gender, race-ethnicity, education, employment status, living situation, duration of untreated psychosis, insight, and severity of psychopathology. These factors were adjusted for confounders in the regression models.

Data were analyzed by using univariate logistic regression to obtain a basic profile of the sociodemographic and clinical characteristics of the sample and their association with service disengagement. Sociodemographic variables and clinical variables with a p value of less than .20 were then included in the multivariable model. Multivariate logistic regression was performed to determine predictors of service disengagement. Statistical analyses were carried out using SPSS, version 19.0.

A total of 839 patients enrolled in EPIP between 2001 and 2009 had two years of EPIP data. Two of these patients were excluded from the analysis because they exceeded the age limit, and one was excluded because the diagnosis was later changed to a nonpsychotic disorder. Sixty-one (7%) patients were discharged early from EPIP because they moved out of the country or wished to follow-up with a private psychiatrist. These patients tended to be younger (odds ratio [OR]=.94, 95% confidence interval [CI]=.88–.99) and to have a main diagnosis of “other psychosis” (OR=2.16, CI=1.15–4.07). Therefore, 775 patients were included in the analysis.

Most of the patients were male (N=395, 51%), unmarried at the time of study (N=643, 83%), and living with others (not living on their own) (N=737, 95%). Various ethnicities were well represented, and the sample was largely reflective of Singapore’s ethnic composition, with 594 (77%) Chinese, 108 (14%) Malays, 55 (7%) Indians, and 18 (2%) of other races. For this study, patients’ diagnoses were grouped into three categories. The most common was schizophrenia spectrum disorders (N=578, 75%), which included schizophrenia, schizophreniform disorder, and schizoaffective disorder. This was followed by other psychoses (N=53, 19%), which included brief psychotic disorder, delusional disorder, and psychosis not otherwise specified; and by affective psychoses (N=144, 7%), which included bipolar disorder and depression. A total of 460 patients (59%) were admitted from the inpatient service.

Of the 775 patients included in the study, 548 (71%) did not disengage; for 118 patients (15%), telephone contact was maintained, and for 55 (7%) telephone contact was maintained with family only. For 54 (7%), there was no contact. In sum, 109 (14%) were considered to have disengaged from the service within two years.

In a univariate regression model, sociodemographic factors associated with service disengagement (p<.20) were older age at admission (OR=1.02, CI=.99–1.06), Malay ethnicity (reference Chinese) (OR=2.15, CI=1.29–3.58), being currently married (OR=1.46, CI=.89–2.40) or unemployed (OR=1.39, CI=.93–2.08), and having a lower level of education (OR=2.87, CI=1.59–5.18). Clinically, patients more likely to disengage from the service had a longer duration of untreated psychosis (OR=1.01, CI=1.00–1.01), a higher baseline PANSS negative subscale score (OR=1.02, CI=.99–1.05), a higher baseline PANSS general psychopathology subscale score (excluding item G12, lack of judgment and insight) (OR=1.01, CI=.99–1.04), and a higher baseline PANSS G12 score (OR=1.21, CI=1.05–1.38).

After adjustment for confounders with a multivariate model, Malay ethnicity (reference Chinese) (OR=1.93) and a low level of education—particularly, less than six years (reference tertiary education—that is, polytechnic or university education) (OR=2.30)—remained highly significant in predicting service disengagement (Table 1). Unlike other studies, this study found that having less severe symptoms at baseline (based on PANSS scores) and living alone were not significant predictors of service disengagement.

The study found that 109 (14%) EPIP patients disengaged from the program within two years, which is lower than the range of 23%–31% found for other early-intervention programs (3–5) and is comparable to the lower end of the reported range for nonspecialized psychiatric services (8). The lower rate of disengagement may be explained by EPIP’s exclusion of persons with a forensic history or a history of substance abuse, both of which have been shown to be predictors of service disengagement from other early intervention programs (3,4). Other possible explanations for the lower rate are inclusion of patients with telephone contact only (passive engagement) in the group regarded as engaged in services and efforts by case managers to reengage patients through repeated phone calls and home visits. In addition, Singapore is an island state, and the geographical concentration of patients may have made it easier for case managers to follow up and for patients to access EPIP services.

In this study, Malay ethnicity was found to be a predictor of service disengagement. Unfortunately, there are few data from similar studies for comparison. However, possible explanations for this finding include a higher level of tolerance and support for persons with mental illness in the Malay family unit compared with Chinese and Indian families, in accordance with Islamic laws and the teachings of the Quran (12,13); a prominent belief among Muslims that the causes of mental illness are religious and spiritual, rather than medical (12,13); and a stronger preference among Muslims than among Buddhists and Hindus for spiritual and religious therapies (14). Although these are plausible reasons for service disengagement among Malay patients, it would be of interest to explore how specific ethnic, cultural, and religious factors affect service disengagement.

Having a low level of education—specifically, less than six years of education—was also a significant predictor of service disengagement. Low educational attainment has been implicated in other studies as a factor associated with higher dropout from psychiatric services (8). As discussed by Dalgard and colleagues (15), low educational attainment may be associated with a decreased ability to manage and control one’s life, including compliance with a treatment regimen. Moreover, clinical experience suggests that this group of patients is more difficult to contact via telephone and thus less traceable when they do not keep appointments.

Duration of untreated psychosis was a significant predictor of service disengagement in this study, albeit one with minimal magnitude of effect. Other studies have found mixed results for this variable (3); in some, a short duration of untreated psychosis (5) is a predictor of disengagement, and in other studies the variable has not had a significant association with disengagement (4).

The naturalistic nature of this study precluded the standardization of the amount of contact between case managers and patients. The assessment of disengagement was also limited to the use of a semistructured scale measuring the amount and type of contact maintained between case managers and patients. In addition, although all the case managers received training in use of the scale, interrater reliability was not measured. Although the scale has not been validated, it is a good tool for assessing disengagement in terms of the EPIP service model (face-to-face contact, phone contact, or none). Future studies should consider the use of validated and structured scales, such as the Service Engagement Scale (6), to obtain additional information, such as the patient’s availability, collaboration, help-seeking behavior, and treatment adherence. Other limitations include the loss of patients with other psychoses to early discharge, which may have biased the generalizability of the results toward those with more severe psychosis (that is, schizophrenia spectrum disorders or affective psychoses).

The strength of this study lies in its sampling of a large number of incident cases from a nationwide catchment area and inclusion of both first-contact patients in the community and inpatients. Patients were referred from schools, community counseling centers, the army, outpatient clinics, and hospitals across Singapore. This study is also the first of its kind in an Asian population and contributes to the growth of research into service models for early psychosis programs in Asia.

The engagement strategy of EPIP needs to target patients of Malay ethnicity and patients with less than six years of education. Further studies are necessary to elucidate the mechanism through which these factors affect service engagement so that appropriate counterstrategies can be developed. In addition, the finding that patients who disengaged from EPIP had a longer duration of untreated psychosis highlights the potential benefits of reducing the duration of untreated psychosis among patients experiencing a first episode of psychosis. The findings provide additional motivation to EPIP staff to persist in and expand outreach programs in the community and linkages with institutions to encourage earlier referrals to EPIP and to educate the public.

The authors report no competing interests.