Psychiatric length of stay determinants in a military medical center

In this 2-year prospective study, 1040 consecutive admissions to an adult inpatient psychiatric unit of the United States Air Force's largest tertiary care medical center were assessed for length of stay (LOS) determinants. Twenty-one demographic, clinical, and diagnostic variables were examined for their effect on LOS. The median LOS at our medical center was 9 days and was similar to the national median LOS of 10 days for psychiatric units in general hospitals. We found several clinical, nonclinical, and diagnostic variables to be independent predictors of LOS, accounting for 31% of the variance, but that much of the variance remained unaccounted for. Primary DSM-III-R discharge diagnoses were helpful in distinguishing a nonoverlapping, short-stay group from a long-stay group. HCFA Diagnostic Related Group (DRG) estimates for LOS were assessed for their ability to predict LOS in our institution. With the exception of substance abuse/dependence disorders, DRGs consistently underestimated LOS. Previous studies using the coefficient of variation (COV) have demonstrated the inability of DRGs to distinguish homogeneous diagnostic groups. However, in this study, COV was helpful in differentiating a majority of DRGs by LOS despite the general underestimation of LOS by DRG. These results continue to support the inadequacy of the DRGs in determining equitable reimbursement and the difficulties in predicting psychiatric LOS despite the inclusion of clinical and nonclinical variables.