We use a structural approach to separately estimate moral hazard and adverse selection effects in health care utilization using hospital invoices data. Our model explicitly accounts for the heterogeneity in the transactions costs associated with hospital visits which increase the individuals' total cost of health care and dampen the moral hazard effect. A measure of moral hazard is derived as the difference between the observed and the counterfactual health care consumption. In the population of patients with non life-threatening diagnoses, our results indicate statistically significant and economically meaningful moral hazard. We also test for the presence of adverse selection by investigating whether patients with different health status sort themselves into different health insurance plans. Adverse selection is confirmed in the data because patients with estimated worse health tend to buy the insurance coverage and patients with estimated better health choose not to buy the insurance coverage.