Socioeconomic Deprivation and Extended Hospitalization in Severe Mental Disorder: A Two-Year Follow-Up Study Melanie Amna Abas, M.D. Jane Vanderpyl, Ph.D. Elizabeth Robinson, M.Sc.
Objective: This study examined the association between socioeconomic deprivation and extended hospitalization in severe mental disorder, after taking account of confounding variables. Methods: A representative sample of 660 inpatients from South Auckland, New Zealand, was followed for two years from their index admission. Additional data were collected during the index admission for a subsample of 291 patients. Results: Greater levels of socioeconomic deprivation in the inpatient’s neighborhood of residence was associated with extended hospitalization after adjustment for demographic factors and primary diagnosis but not after adjustment for comorbid diagnosis, chronicity, function, and severity. Most extended hospitalizations were related to poor illness recovery. Conclusions: People from more deprived areas are likely to need longer psychiatric admissions, mostly because of the asso-
Dr. Abas is affiliated with the Health Services and Population Research Department, Section of Epidemiology, Institute of Psychiatry, Kings College London, P.O. 60, London SE5 8AF, United Kingdom (e-mail:
[email protected]. uk). Dr. Vanderpyl is with the Research, Evaluation and Audit in Mental Health Services Team, Counties Manukau District Health Board, Auckland, New Zealand. Ms. Robinson is with the Section of Epidemiology and Biostatistics, University of Auckland. 322
ciation between deprivation and having more disabling symptoms and a comorbid psychiatric diagnosis. Interventions to prevent psychiatric hospitalization, reduce duration of stay, and enhance recovery must be tested among those with greater levels of socioeconomic deprivation. (Psychiatric Services 59:322–325, 2008)
P
sychiatric admissions are more common in geographic areas of greater socioeconomic deprivation (1), but it is not clear whether deprivation is also associated with extended hospitalization. If so, the deprivation itself may explain delayed recovery (for instance, poverty may bring about resignation and low self-esteem), or the apparent association may be explained by confounding variables, such as comorbid substance abuse or more severe psychotic symptoms. Our hypothesis was that greater deprivation at admission would predict total duration of psychiatric hospitalization over the two-year followup, even after we controlled for clinical and service factors. To examine our hypothesis, we looked at the socioeconomic levels of small areas of residence (2) among inpatients from South Auckland, New Zealand. We used deprivation of residence as an individual attribute. We also aimed to explore whether extended stay was related to poor illness recovery or to having no suitable accommodation to be discharged to. PSYCHIATRIC SERVICES
Methods Compared with most of New Zealand, the South Auckland district (378,000 residents) has a high level of deprivation. For example, in 1996 34% of persons in that district lived in areas ranked nationally as being in the most deprived quintile (3), and there was a high proportion of ethnic minorities in the district (18% Maori, 17% Pacific Islander, 8% Asian, 58% New Zealand European and other groups). Most residents live in urban areas. Any residents needing secondary mental health care would normally use the free public district mental health service. Private care for severe mental disorders is negligible. In 2000 the district mental health service had 40% to 45% fewer psychiatric beds and community mental health staff than recommended nationally (4), which can lead to pressure to keep admissions short and to a high workload for staff. The cohort comprised consecutive admissions from within the district to the psychiatric inpatient unit from June 1998 to December 2000. We excluded admissions of persons who did not live in the district and of those who had no address. We followed each participant for two years from his or her index admission. For a random subsample, made up of consecutive index admissions between November 1999 and August 2000, we collected additional clinical and service-related data within the first 24 hours of the admission by using structured questionnaires, including the Health of the Nation Outcome Scales
♦ ps.psychiatryonline.org ♦ March 2008 Vol. 59 No. 3
(5), the Global Assessment of Functioning (6), and the Reasons for Admission Schedule (7). For those from this subsample who were still admitted after five weeks, we collected data by using the Reasons for Continued Stay Schedule (8). A trained psychiatric research nurse collected these data from clinical case notes and by interview with the primary nurse and consultant psychiatrist. Interrater reliability of the research nurse’s ratings with those of a research psychiatrist (MA) was mostly good. For 35 patients, interrater reliability for 80% of the items on the Health of the Nation Outcome Scales had a mean Cohen’s kappa between .61 and .77. For items 8 (other symptoms) and 9 (relationships) the kappas were .55 and .50, respectively. The research psychiatrist (MA) and the research nurse discussed the ratings and repeated the reliability exercise later with 30 additional patients. Interrater reliability for the 30 additional patients showed slightly improved agreement. The Auckland Ethics Committee (now known as the University of Auckland Human Participants Ethics Committee) approved the study. Total length of stay was the total number of days of hospitalization at the district psychiatric inpatient unit for each patient in the 24 months after the date of their index admission. We measured deprivation with the NZDep96—a deprivation index that was created from 1996 census data (9). The index was derived from data from small areas (median of 90 persons per small area) by using a weighted combination of the age- and sex-standardized variables. We defined persons as being in the most deprived category if they lived in areas ranked in the lowest quintile (in line with the New Zealand definition of “poor populations”); the moderately deprived category if they lived in areas ranked in the next highest quintile; and in the least deprived category if they lived in areas ranked in the top three quintiles. An independent firm assigned a geographical smallarea code to each patient’s address at admission, which enabled assignment of the correct deprivation score. We analyzed the log of total length PSYCHIATRIC SERVICES
of stay and the geometric mean, because of the lognormal distribution of the total length of stay. We used generalized linear modeling to model the effect of deprivation in three categories. We calculated an estimate of the days of stay for each category of deprivation with all other variables in the model at baseline by exponentiating the regression coefficients. [Two figures showing the average total length of stay and the log of the average total length of stay, along with calculations for the estimated length of stay, are available as an online supplement to this brief report at ps.psychiatryonline.org.]. After adding in each potential confounding variable, we carried out a likelihood ratio test to see whether deprivation made an independent contribution to the model.
Results A total of 872 patients from the district were admitted to the psychiatric inpatient unit during the study period. We excluded 17 patients because they had no address to code for deprivation. Of the 855 remaining patients, 291 constituted the subsample of admissions between November 1999 and August 2000 for which we had detailed information. Of the other 564 patients in the sample, we were able to include 369 who had
enough information about residential address to enable coding at small-area level. Those excluded were slightly more likely to be from an ethnic minority group or unmarried, but these differences were not statistically significant. Table 1 shows characteristics of the total sample (N=660) at index admission. Thirty-eight percent (N=250) of patients in the total sample had one or more readmissions during the twoyear follow-up. The median total length of stay was 28 days. Greater duration of stay was associated with more deprivation (F=4.40, df=4 and 655, p=.002), ethnicity (Pacific Islander and Maori ethnicity) (F=6.88, df=3 and 650, p