It is well-recognized that individuals’ spending on medical care is not entirely necessary but is to some extent discretionary. However, the composition of medical spending into necessary and discretionary parts is not (perfectly) observable. In this paper, we study ways to improve upon existing public health insurance policies using a framework where both the discretionary and necessary components of medical spending are explicitly modeled. We start by constructing a simple theoretical framework in the spirit of Mirrlees (1971) in which only total medical spending is observable, but not its division into discretionary and non-discretionary parts. We show that individuals with low consumption of medical care should be rewarded with high consumption of regular goods. Next, we construct a rich structural life-cycle model and evaluate the quantitative impact of this type of policy with application to public health insurance. The success of each policy is measured by benchmarking it against the full information case in which the division of medical spending into discretionary and non-discretionary components is observable. We show that the best results are achieved by introducing an option to substitute public health insurance with cash transfers. This is because this policy creates a trade-off between regular and medical consumption, which is the main mechanism behind our theoretical results.