Comparing hospital costs: what is gained by accounting for more than a case-mix index?

We explore what effect controlling for various patient characteristics beyond a case-mix index (DRG) has on inferences drawn about the relative cost performance of hospital departments. We estimate fixed effect cost models in which 3754 patients are clustered within six Danish vascular departments. We compare a basic model including a DRG index only with models also including age and gender, health related characteristics, such as smoking status, diabetes, and American Society of Anesthesiogists score (ASA-score), and socioeconomic characteristics such as income, employment and whether the patient lives alone. We find that the DRG index is a robust and important explanatory factor and adding other routinely collected characteristics such as age and gender and other health related or socioeconomic characteristics do not seem to alter the results significantly. The results are more sensitive to choice of functional form, i.e. in particular to whether costs are log transformed. Our results suggest that the routinely collected characteristics such as DRG index, age and gender are sufficient when drawing inferences about relative cost performance. Adding health related or socioeconomic patient characteristics only slightly improves our model in terms of explanatory power but not when drawing inferences about relative performance. The results are, however, sensitive to whether costs are log transformed.