"Private insurance for prescription drugs is characterized by two regimes: flat copayments and variable co-insurance. We develop a simple model to show that patient compliance is lower under coinsurance due to uncertainty in cost-sharing. Empirically, we derive comparable models for compliance behavior in the two regimes. Using claims data from nine large firms, we focus our analysis on diabetes, a common chronic condition that leads to severe complications when inappropriately treated. In the coinsurance model, an increase in the coinsurance rate from 20% to 75% resulted in the share of persons who never comply to increase by 9.9%, and reduced the share of fully compliant persons by 24.6%. In the copayment model, an increase in the copayment from $6 to $10 resulted in a 6.2% increase in the share of never-compliers, and a concomitant 9% reduction in the share of full compliers. Similar results hold when the level of cost-sharing is held constant across regimes. While non-compliance reduces expenditures on prescription drugs it may also lead to increases in indirect medical costs due to avertable complications. Using available aggregate estimates of the cost of diabetic complications, we calculate that the $6-$10 increase in copayment would have the direct effect of reducing national drug spending for diabetes by $125 million. However, the increase in non-compliance rates is expected to increase the rate of diabetic complications resulting in an additional $360 million in treatment costs. The results suggest that both private payers and public payers may be able to reduce overall medical costs by switching from coinsurance to copayments in prescription drug plans"--National Bureau of Economic Research web site.