PATIENT LENGTH OF STAY: PREDICTIVE VARIABILITY AMONG FORENSIC AND CIVIL PSYCHIATRIC UNITS AND POPULATIONS

Patient length of stay (LOS) in psychiatric inpatient services has become one of the most popular but least understood measures of treatment efficiency and total cost of care. The initiative to decrease LOS in an effort to reduce total treatment costs will ultimately be more costly, unless LOS predictors are appropriately applied and revised based on individual patient needs to decrease subsequent hospitalizations. Additionally, unit-specific variables such as staff/client ratio, unit size, and treatment approach have a direct impact on unit operations and should be expected to cloud LOS comparisons across facilities. These inherit differences make the prediction of LOS norms in advance more problematic. The present study investigated several patient demographic, clinic, and legal variables as predictors of LOS on forensic and civil psychiatric units at a major metropolitan public hospital. Data on 1,201 patients (forensic n = 767; civil n=434) were collected from the information in medical records routinely collected for quality assurance purposes at a single facility over a 10-year time span (1999-2008). A series of multiple regression analyses were then conducted to determine the best fit prediction equation for each psychiatric unit, and a cross validation approach was utilized to determine the generalizability of those equations.