BACKGROUND: Inpatient treatment is the most costly sector of treatment for depressive disorders in Germany. However, little is known about which patient and hospital characteristics contribute to costs of inpatient episodes. PATIENTS AND METHODS: To take part in this study, patients had to fullfill criteria for ICD-10 diagnosis of F31.3-F31.5, F32, F33, F34.1, F43.20, or F43.21. Episodes were recorded between September 9 2001 and March 3 2003 in ten hospitals in three German states. Inpatient records of 1,202 persons were analysed. Multiple regression analysis was performed to identify significant patient predictors of cost per inpatient episode, and the predictive function of hospital characteristics was analysed by applying hierarchical linear modeling. RESULTS: Patient characteristics at admission could not explain a substantial part of the variance in episode costs. Better prediction was possible including variables from the whole treatment process. Also, conditions for admission and patient-related factors did not well explain cost differences between hospitals, but characteristics of the whole treatment were. CONCLUSION: For predicting costs of inpatient depressive episodes, the complete course treatment has to be considered. As in the physiologic sector, therapeutic and diagnostic procedures have a great effect on cost prediction.
BACKGROUND: Inpatient treatment is the most costly sector of treatment for depressive disorders in Germany. However, little is known about which patient and hospital characteristics contribute to costs of inpatient episodes. PATIENTS AND METHODS: To take part in this study, patients had to fullfill criteria for ICD-10 diagnosis of F31.3-F31.5, F32, F33, F34.1, F43.20, or F43.21. Episodes were recorded between September 9 2001 and March 3 2003 in ten hospitals in three German states. Inpatient records of 1,202 persons were analysed. Multiple regression analysis was performed to identify significant patient predictors of cost per inpatient episode, and the predictive function of hospital characteristics was analysed by applying hierarchical linear modeling. RESULTS: Patient characteristics at admission could not explain a substantial part of the variance in episode costs. Better prediction was possible including variables from the whole treatment process. Also, conditions for admission and patient-related factors did not well explain cost differences between hospitals, but characteristics of the whole treatment were. CONCLUSION: For predicting costs of inpatient depressive episodes, the complete course treatment has to be considered. As in the physiologic sector, therapeutic and diagnostic procedures have a great effect on cost prediction.