Understanding the Drivers and Effects of Treatment Variation Across Hospitals

Treatment variation across hospitals is seen as a sign of unwarranted care provision, especially if it leads to outcome variation. However, it is often unclear whether such variation is hospital-driven or related to (unobserved) patient characteristics. We study the variation in the surgical treatment of breast cancer in the Netherlands using individual claims data from a Dutch health insurer between 2016 and 2018. We estimate a case-mix-adjusted treatment propensity score that captures a hospital’s propensity to provide lumpectomy instead of mastectomy and use an instrument based on geographical distance to exploit variation between a group of patients (quasi-)randomly assigned to a hospital. Leveraging this variation through an instrumental variable approach, we find substantial hospital-driven treatment variation that, on average, impacts clinically important health outcomes for (quasi-)randomly selected patients. These effects suggest that, in the context of breast cancer, hospital variation is unwarranted. Further descriptive analyses do not provide strong evidence that this variation is driven by comparative advantages, tentatively pointing towards allocative inefficiencies.