Premature cessation of antibiotic therapy (i.e., non-adherence) is common in long treatment regimens and can severely compromise health outcomes.

We investigate the problem of designing a schedule of incentive payments to induce socially-optimal treatment adherence levels in an unsupervised setting. The novel elements of this problem stem from its institutional features; there is a single incentive schedule applied to all patients, incentive payments must increase in line with patients’ adherence, and patients cannot be a priori prohibited from any given levels of adherence.

We develop models to design optimal incentives incorporating these features, which are also applicable in other problem contexts that share the same features. In a numerical study using representative data from a tuberculosis epidemic in India, we show that our optimally-designed incentive schedules are generally cost-effective compared to a linear incentive benchmark.