What is the optimal paid sick leave system? To answer this question, I combine individual-level data on paid sick leave claims with a model of sick pay insurance provision. I start by providing evidence that workers respond to the monetary incentives induced by the benefit scheme. I also document that workers behavior varies with the day of the week they fall sick. I use these patterns to inform a model of sick pay insurance. In the model, risk-averse workers face a health shock and decide how many days to be on leave. Workers are insured by a risk-neutral social planner who chooses the optimal contract to maximize social welfare, considering workers' behavioral responses. Social welfare is a function of workers'; utility and the potential production losses induced by sick pay provision. The main empirical challenge in estimating the model is to disentangle the underlying distribution of health from workers' preferences. To overcome this challenge, I combine the individual-level data on sick pay utilization with detailed medical assessments of recovery times associated with each health condition. This strategy allows me to construct the underlying distribution of health without imposing parametric assumptions. To estimate workers' preference parameters, I exploit the day of the week on which a sick leave claim is filed as a quasi-exogenous shifter of the temptation to extend a sick leave claim as the main source of variation. Finally, I use the estimated model to derive the optimal sick pay contract and estimate the welfare gains from its implementation. I find that, relative to the current system, the optimal system would provide more insurance for short-term sickness and less insurance, i.e., lower replacement rates, for longer sickness spells. I estimate that workers are willing to give up 1.53% of their earnings to be insured under the optimal policy.