Modeling the growth of long-stay populations in public mental hospitals

• Published in: Social science & medicine (1982) Vol. 30; no. 12; pp. 1341 - 1347
• Main Authors: Fisher, William H., Phillips, Barbara F.
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 1990; Elsevier; Pergamon Press Inc
• Series: Social Science & Medicine
• Subjects: Algorithms; Chronic Disease; Deinstitutionalization; dynamic modeling; Follow-Up Studies; Forecasting; Hospitalization; Hospitals, Psychiatric - statistics & numerical data; Hospitals, State - statistics & numerical data; Humans; Length of stay; Length of Stay - statistics & numerical data; Life Tables; Massachusetts; Mental Hospitals; Mental institutions; Mental Patients; Models, Statistical; Population Dynamics; Psychiatric hospitals; public mental hospitals; public mental hospitals length of stay survival analysis dynamic modeling; Social research; survival analysis; Time Factors; United States; length of stay; public mental hospitals; dynamic modeling; survival analysis
• ISSN: 0277-9536; 1873-5347
• DOI: 10.1016/0277-9536(90)90314-I

Long-stay, chronic patients have been a problematic subpopulation in public mental hospitals for over a century. Despite three decades of deinstitutionalization and a major shift toward shorter episodes of hospitalization, there continues to exist a group of patients who experience lengthy hospital stays. As the number of such patients increases in a facility, its ability to provide acute care may be compromised, and the size of this subpopulation must therefore be anticipated. This paper examines the length-of-stay patterns of a sample of public mental hospital admissions through the use of life table analysis, and develops a dynamic modeling algorithm using sample survival function values. Life table analysis revealed a declining hazard function, indicating a diminishing probability of discharge with increased hospital stay. The dynamic model showed that, after 2 years of operation of a hypothetical facility, current length-of-stay patterns would generate an inpatient population 40% of which had been hospitalized for over 6 months. Goodness-of-fit tests comparing the algorithm's forecast with actual hospital utilization data showed its predictions to be reliable. The authors discuss the use of this methodology to anticipate the effects of programmatic or other types of changes in mental hospitals, and also suggest other types of settings where such modeling techniques might profitably be applied.