Forecasting length of stay in child residential treatment

A sample of 126 consecutively admitted residential treatment children (mean age = 9.86, SD=1.84; 70.6% male; 42.1% African American; 50% Caucasian) were studied over a five-year period to identify predictors of length-of-stay. Cox regression was the primary statistical method used to analyze psychiatric and behavioral rating data for children assessed by teachers and treatment staff using the Devereux Scales of Mental Disorders (DSMD). Parental alcohol abuse, and children's age, medication status, race, initial DSMD total and critical pathology scores, were predictive of length-of-stay. Residential length-of-stay was strongly linked to initial levels of psychiatric symptomatology. Models that can help forecast length of stay are vital tools in helping to improve both clinical and utilization management strategies.