Unobserved health care expenditures: How important is censoring in register data?

Accurate information on individuals' health service use is important for evaluating health policies and analyzing health care demand. Although register data are considered to be more reliable than survey data, little is known about the extent and effect of censoring of the expenditure distribution in register data. We exploit a recent change in the health provider remuneration system in several Swiss cantons to empirically investigate whether censoring occurs when individuals do not have to disclose their health service use below their deductible level. Applying a difference-in-differences approach, we find that between CHF 6.70 (1.7%) to CHF 9.64 (2.4%) of all health service use paid out-of-pocket are not observed (per capita per year). This effect seems to be driven by high-deductible plans where observed out-of-pocket expenditures declined by CHF 30.34 (7.6%) after the change. Although statistically significant, these effects are almost negligible in economic terms. We therefore concluded that, if anything, censoring is a very limited issue in Swiss health insurance claims data.