Template matching for hospital comparison: an application to birth event data in Italy

Quality evaluation in healthcare has obtained a growing attention in the statistical literature. In order to evaluate hospital performances by comparing hospital outcomes it is necessary to remove the bias due to the different case-mix in each hospital, which is usually done using statistical modeling as a risk adjustment tool. Template matching is a new matching approach allowing to remove bias selection in a multi-treatment setting, resulting in a clean and easily interpretable evaluation. We adopted this method to compare caesarean section rates across the hospitals of two Italian regions: Sardinia and Lombardy. We found 5 (out of 79) hospitals with abnormal performance on the same template of expectant mothers, and all these hospitals are located in Lombardy, whereas we do not observe a relationship with the number of deliveries per year and the ownership. Despite this, fairness of the comparing procedure makes easier for the policy makers the identification of potential outliers with respect to both patients’ selection and outcomes.