Economic Grand Rounds: Patterns of Hospital Costs for Depression in General Hospital Wards and Specialized Psychiatric Settings

This study analyzed the Nova Scotia Longitudinal File (NSLF), a database created by AlgoPlus Consulting Limited in collaboration with the Nova Scotia Department of Health. The NSLF is a statistical psychiatric database that includes all hospitals (general and psychiatric) in Nova Scotia. Patient confidentiality was maintained through the use of a Nova Scotia Department of Health encryption algorithm, which transformed the patient identifiers in a consistent manner. All hospitalizations over the study period for individual patients were cumulated and organized by the patient identifier, which enabled analysis of the longitudinal course and costs of hospital care (8).

The study group consisted of all patients who were age 18 years or older at first hospitalization and who had a primary diagnosis of any mental disorder as well as a depressive-spectrum diagnosis between January 1, 1989, and December 31, 1991.

Specifically, all patients included had to have an ICD-9 diagnosis of depression during the period from 1989 to 1991. The diagnoses included the affective psychoses (ICD-9 codes 296.1, 296.3, 296.4, 296.5, 296.6, 296.8, and 296.9), neurotic depression (code 300.4), affective personality disorder (code 301.1), and depression not elsewhere classified (code 311).

Hospital costs were cumulated from the first hospitalization during 1989 through 1991 until December 31, 1994, which constituted a three- to six-year follow-up period. Hospital costs were calculated by multiplying the number of patient days in each hospital setting by the cost per patient day determined by the Nova Scotia Department of Health in 1994 Canadian dollars.

Results are reported in May 1998 U.S. dollars. First, the 1994 Canadian dollars were transformed into May 1998 Canadian dollars using the Statistics Canada Consumer Price Index (1994 Canadian dollars were multiplied by 108.7 divided by 103.8). Next, the May 1998 Canadian dollars were transformed into May 1998 U.S. dollars at the May 15, 1998, noon exchange rate of .6894. The costs per hospital day were $289 U.S. 1998 ($400 Canadian 1994) for general wards in general hospitals, $399 U.S. 1998 ($552 Canadian 1994) for psychiatric units in general hospitals, and $321 U.S. 1998 ($445 Canadian 1994) for units in psychiatric hospitals.

Between 1989 and 1991, a total of 4,383 individuals were hospitalized in Nova Scotia for depressive disorders, for a rate of seven patients per 1,000 population.

As Figure 1 shows, 51 percent of the 4,383 patients (N=2,217) were hospitalized at least once in nonspecialized general hospital wards; 33 percent (N=1,427) were hospitalized exclusively in this setting, and 18 percent (N=790) were hospitalized in mixed settings. Forty-nine percent of the 4,383 patients (N=2,166) were hospitalized exclusively in specialized units, either in general hospitals or in psychiatric hospitals.

Over the three- to six-year follow-up, the mean±SE cumulated hospital cost per patient was more than six times higher in specialized units than in general wards—$30,266±$1,026 versus $4,729±$173. Those treated in mixed settings had a mean cumulated hospital cost of $29,422±$1,301, which was similar to the cost for those hospitalized in specialized units.

Within the same hospital setting, marked variability was noted in cumulated costs per patient. The coefficient of variation was 138 percent for general wards and 158 percent for specialized units, which means that relative to their means, the cumulated costs within general wards varied less than within specialized units.

The distributions of cumulated costs were skewed. The 25th, 50th, 75th, and 90th cost percentiles for patients treated exclusively in specialized units were $2,888, $8,361, $22,842, and $54,052, respectively. The same figures for patients treated exclusively in general hospital nonspecialized wards were $1,155, $2,599, $5,487, and $10,107.

Although the mean cumulative cost for all patients with depression was $21,800 per patient, the mean was only $3,376 per patient for the patients in the lower-cost half of the sample. The mean cumulated hospital costs were $66,304 for a patient in the top quartile and $111,644 for a patient in the top decile.

In Nova Scotia, .7 percent of the adult population (age 18 and over) was hospitalized for clinical depression during 1989–1991. This finding conflicts with the general perception of massive unrecognized need due to missed diagnoses of depression. Upper-limit estimates of three-year hospitalization rates of .39 to .75 percent can be derived by making four assumptions (9). The first is that all mood disorders during one year, 7.1 percent according to the U.S. Surgeon General's recent report (10), are depression. The second assumption is that all cases in which depression is recognized, 35 percent according to one recent study (11), are cases of major depression (35 percent × 7.1 percent = 2.5 percent). The third assumption is the Surgeon General's estimate that 5 to 10 percent of cases of major depression require hospitalization (10) (5 percent × 2.5 percent = .13 percent to 10 percent × 2.5 percent = .25 percent). The fourth assumption is that the three-year rate of hospitalization would be three times the one-year rate, which would result in a hospitalization rate in the adult population of .39 to .75 percent. The actual Nova Scotia hospitalization rate for depression of .7 percent is at the high end of this theoretical upper limit.

A major methodological contribution of this paper has been determining longitudinal hospital costs for depression. Major cost differences were found between hospital settings. The mean cumulated cost for patients hospitalized only in nonspecilized wards was $4,729, or 85 percent less than the cost of hospitalization for those hospitalized only in specialized units, which was $30,266. Similar disparities between costs for hospital care have previously been reported (6).

If such studies are to have an impact on policy, further analyses are required to confirm that the differences in hospital costs for depressed patients in Nova Scotia were not the results of severity or complexity of illness. We have previously reported that nonspecialized wards in general hospitals treat patients who are older than those treated in specialized units (12). The older patients would be expected to be a more complex patient group. The assumption that additional days of hospitalization enhance patient outcomes has been shown to be dubious in research on hospital cost-containment pressures of prospective payment systems (6,13).

Severity of depression is difficult to determine. Available methods are limited and lack rigor. We have no equivalent of the Apgar score for newborns. Current methods include a symptom count as in the Hamilton Depression scale, diagnosis-related-group classification, proprietary algorithms that are not disclosed (14), and an examination of the clinical course of care, encompassing previous history, complications, rehospitalizations, and resource use (costs).

In comparing patients, it would be useful to answer four questions. First, are similar proportions of patients with specific diagnoses treated in general wards and in specialized units? Second, what is the influence on costs of gender, age, geographic proximity to the hospital, previous hospitalizations, and comorbidity? Third, among patients hospitalized in both specialized and nonspecialized settings, what is the sequence of hospitalization settings? Fourth, what are the intervals between hospitalizations for patients treated in general wards and in specialized units?

Major variations in hospital costs were found within the same hospital setting. This finding is consistent with findings about treatment of depression in the Veterans Affairs health care system (15), which has standardized treatment protocols. Twenty-nine percent of VA medical centers reported mean lengths of stay for their patients with depression that were significantly different (p<.01) from the mean in the study sample.

In our study the distribution of costs was skewed. In a nonselected hospitalized population of patients with depression from which no patients were excluded, 10 percent of the patients accounted for 51 percent of the hospital costs. The mean per-patient cost for all of the hospitalized patients with depression was $21,800. However, the mean cost for a patient in the decile with the highest costs was $111,644. Among patients in specialized settings, the mean cost was $30,266, while the median cost was $8,361. A few high-use patients had a major impact on the overall system costs. This finding has important implications regarding future long-term costs resulting from mental health parity legislation in the United States.

This study confirmed that nonspecialized wards in general hospitals play a major role in hospitalizing depressed patients. Economic studies of depressive disorders must include these wards.

The authors are affiliated with AlgoPlus Consulting Limited, 5675 Spring Garden Road, Suite 502, Halifax, Nova Scotia, Canada B3J 1H1 (e-mail, [email protected]). This work was presented as a poster at the 12th International Conference on Mental Health Problems in the General Health Care Sector, sponsored by the National Institute of Mental Health, held July 13–14, 1998, in Baltimore. Steven S. Sharfstein, M.D., is editor of this column.