AIMS: The importance of using valid case-mix systems in long-term care is addressed by comparing the predictive power of different case-mix models, and by applying them in the calculation of technical efficiency scores of care units. METHODS: To construct different case-mix models a statistical clustering technique (Automatic Interaction Detection) was used. Technical efficiency score were calculated using data envelopment analysis (DEA). RESULTS: The Resource Utilization Groups (RUG-III/22) classification explained 39% of resident specific cost, compared with 16% for a functional dependency scale in the Finnish patient information system HILMO. CONCLUSION: When assessing the economic performance of long-term care units it is important to pay attention to the predictive validity of the case-mix measure to be used. The choice of case-mix measure significantly affected how units were rated in efficiency.
AIMS: The importance of using valid case-mix systems in long-term care is addressed by comparing the predictive power of different case-mix models, and by applying them in the calculation of technical efficiency scores of care units. METHODS: To construct different case-mix models a statistical clustering technique (Automatic Interaction Detection) was used. Technical efficiency score were calculated using data envelopment analysis (DEA). RESULTS: The Resource Utilization Groups (RUG-III/22) classification explained 39% of resident specific cost, compared with 16% for a functional dependency scale in the Finnish patient information system HILMO. CONCLUSION: When assessing the economic performance of long-term care units it is important to pay attention to the predictive validity of the case-mix measure to be used. The choice of case-mix measure significantly affected how units were rated in efficiency.