and. Methodological. Issues in the Comparison of Inpatient. Psychiatric. Facilities. Nancy. A. Goodban,. Ph.D., Paul B. Lieberman,. M.D., Michael. A. Levine, M.A.,.
A. Goodban, Ph.D., Paul Boris M. Astrachan,
esearch has approached the study of length of stay in psychiatric hospitals from two different perspectives. One looks at length of stay as a predictor of outcome. In general, these studies have found that more hospital days are no more beneficial than shorter lengths of stay (1-8). The other approach, represented in this paper, concentrates on predicting length of stay and leaves aside the question of whether there is an optimum length of stay representing the most appropriate care (9-12). This approach has become more common as concern has increased about the cost of inpatient psychiatric care. Although the research literature presents a mixed picture regarding whether the full constellation of necessary outpatient services is tess costly than hospitat-based care for disabled, seriously disturbed patients, it is generally thought that hospitalization
School, Hanover, NH 03756. Supported in part by NIMH training The authors thank Phil Leaf, Ph.D.,
and Dave © 1987
Mental Health Center, New Haven; Middletown; and the Department of New Haven. Address reprint requests to
1 44: 1 1 , November
MH-15783. Wells, Ph.D.,
B. Lieberman, M.D., Michael M.D., and Vincent Cocilovo,
The authors compared the length of stay of acute admission patients at a mental health center and a nearby state hospital. The two facilities had significantly different length of stay distributions; the mean was not an adequate index to describe these patterns. Despite careful matching, patients at the state hospital were more disabled. Different patient characteristics were associated with length of stay at the two facilities, and these were also characteristics on which the patient populations differed at admission. The authors conclude that comparisons of hospitals, for example, on mean or median length of stay can be misleading unless the different f unctions, policies, and constraints of the facilities are taken into account. (AmJ Psychiatry 1987; 144:1437-1443)
1987. From the Connecticut Connecticut Valley Hospital, Psychiatry, Yale University,
A. Levine, M.D.
should be limited to those who absolutely need it. Therefore, length of stay is rigorously controlled and attempts are made to provide care at other levels of service as appropriate. Within this framework the question arises: “What different types of patients are there, and what is the normative length of stay for each type?” Although practitioners resist equating what is normative with what is optimal, programs for cost containment such as prospective payment systems based on diagnosis-related groups (DRGs) have focused increased attention on efforts to establish norms for length of stay. Studies predicting length of stay tend to took at the characteristics of either patients or the hospital system. There is evidence for the importance of both. Patient characteristics, including sociodemographic (employment history, insurance status, age, sex, race, and marital status) and clinical (diagnosis, axis IV and V ratings, and psychiatric history) variables have been found to be associated with length of stay in a number of studies (9-17). Relevant hospital and system charactenistics include staffing patterns, availability of aftercare, treatment goals, referral patterns, and ward milieu (2, 18, 19). Recent research has examined the effect of benefit structure (20), hospital ownership, and type of hospital (21, 22). In addition, refinements of the DRG system have been suggested (23-25) in order to predict length of stay more accurately. Thus far, however, little consensus exists about the factors that account for length of stay. Studies that have identified predictive factors do not agree on what these factors are. In any case, length of stay analyses tend to use mean length of stay as an index of length of stay and to predict length of stay by using parametric statistics such as linear multiple regression (10-12, 26). The current study explored length of stay patterns at two public sector psychiatric facilities that serve the same geographic area in order to learn whether length of stay for acute care populations from a specified geographic area differed at the two facilities and, if so, what factors might account for differences between the two facilities. In the process of the analysis, several methodological problems were uncovered that may account in part for the difficulty in understanding and predicting length of stay in psychiatry.
METHOD The facilities studied were a state hospital and a mental health center in south central Connecticut. They are both public sector institutions, part of the state department of mental health system, and subject to the same fiscal constraints and payment policies. The mental health center, administered jointly by a university department of psychiatry and the department of mental health, is primarily an outpatient facility with a small acute care inpatient unit consisting of 32 beds. It is located in the center of the largest city in its 13-town catchment area, which has a population of 425,000. Mental health center patients who need tong-term hospitalization are transferred to the state hospital, which has facilities for the provision of both shortand tong-term care. The state hospital is a large regional facility and is mostly inpatient. It is located 30 mites to the north and serves a broader geographic area. For the purposes of this study, we used only subjects from the mental health center catchment area who were admitted to the acute care unit. This pair of hospitals was selected for comparison for several reasons. First, both facilities share the same formal criteria for admission to their acute inpatient treatment units: individuals who are admitted must be public patients with severe psychiatric disorders and must live in the geographic area of the facility. Second, the pressures of deinstitutionalization have led to a gradual phasing down of long-term care at the state hospital and an attempt at the convergence of its functions with those of the mental health center so that programs for acutely itt inpatients are now similar. Last, the two facilities are part of the same system of care. Ninety percent of the patients admitted to the state hospital from the mental health center catchment area have a history of inpatient or outpatient treatment at the mental health center. At the same time, the majority of patients discharged from the state hospital are referred to the mental health center for outpatient treatment. The two facilities share access to the same range of community and aftercare programs. The study population included alt patients admitted for acute psychiatric care to either facility from the mental health center catchment area during the 6month period July to December 1983. The data were obtained from the computer database maintained by the department of mental health. Derivation of the study sample is shown in table 1. Patients admitted to either facility for specialized detoxification (the bulk of the patients admitted to the state hospital) or geriatric treatment were excluded from the study. Patients who were committed to the state hospital by the courts through civil commitment proceedings, insanity acquittals, and so forth were also excluded, since patients with these legal statuses were not eligible for admission to the mental health center. As might be expected from the structure of the two facilities, the study population made up a much
TABLE 1. Derivation of Sample of Acutely Ill Patients a State Hospital or Mental Health Center, July 1-Dec.
Admitted to 31, 1983 Mental Health Center Patients
State Hospital Patients Group Total
Drug and alcohol admissions Geriatric admissions Nonresidents of area
1,105 33 346
65 2 20
33 8 34
12 3 12
Court-mandated admissions Repeat admissions’
to one of the two facilities only
8 more episode
187 than was
67 during for
smaller proportion of total admissions at the state hospital than at the mental health center. Patients who were transferred were categorized according to the facility of admission. Patients who were transferred from the mental health center to the state hospital were counted as mental health center patients only (there were no transfers in the other direction during the study period). Finally, repeat episodes for any patient during the period were excluded. These criteria yielded 133 patients at the state hospital and 187 at the mental health center.
The first question was whether length of stay differed between the two facilities. We were interested in the total hospitalization for each illness episode, so if a patient was transferred from the mental health center to the state hospital, the length of stay assigned for the mental health center included the accumulated days at both facilities. Comparable state hospital patients needing long-term care were transferred to other units within the hospital. The median lengths of stay at the two facilities were very similar-20 days at the mental health center and 17 days at the state hospital. On the other hand, the means±SDs were very different-26.1±34.2 days at the mental health center and 42.6±61.2 days at the state hospital, indicating greater positive skew for the state hospital. The range was much greater at the mental health center (1-378 days) than at the state hospital (1-357 days). It is clear from figure 1 that not only were the mean lengths of stay different, but so were the distributions. (Because of the tong tail on the state hospital distnibution, those with lengths of stay greater than 100 days were grouped at more than 100 days for the purposes of the graph.) The Kolmogorov-Smirnov two-sample
1 44: 1 1 , November
1. Length of Stay of Acutely Ill Patients Admitted to a State or Mental Health Center, July 1-Dec. 31, 1983
State hospital patients (N - 133)
30 25 20
LENGTH OF STAY (days)
test of the difference in distributions (27) yielded a D of .20 (p