Engineering Incentives in Distributed Systems with Healthcare Applications
MetadataShow full item record
U.S. healthcare costs have experienced unsustainable growth, with expenditures of $2.5 trillion in 2009, and are rising at a rate faster than that of the U.S. economy. A major factor in the cost of the U.S. healthcare system is related to the strategic behavior of system participants based on their incentives. This dissertation addresses the challenge of designing incentives to solve problems in healthcare systems. Principal agent theory and Markov decision processes are the primary methods used to construct incentives. The first problem considered is how to design contracts in order to align consumer and provider incentives with respect to preventive efforts. The model consists of an insurer contracting with two agents, a consumer and a provider, and focuses on the trade off between ex ante moral hazard and insurance. Two classes of efforts on behalf of the provider are studied: those which complement consumer efforts, and those which substitute with consumer efforts. The results show that the provider must be given incentives when the consumer is healthy to induce effort, and that inducing provider effort allows an insurer to save on incentives given to the consumer. The insurer can save on the cost of incentives by using a multilateral contract compared to the bilateral benchmark. These savings are illustrated by an example showing which model features affect the savings achieved. The second problem addresses the decision to provide knowledge to consumers regarding the consequences of health behaviors. The model developed to address this second problem extends the literature on incentives in healthcare systems to consider dynamic environments and includes a behavioral model of healthcare consumers. By using a learning model of consumer behavior, a policy maker's knowledge provision problem is transformed into a Markov decision process. This framework is used to solve for optimal knowledge provision policies regarding behaviors affecting coronary health. Sensitivity analysis shows robust threshold features of optimal policies. The results show that knowledge about smoking should be provided at most health and behavior states. As the cost of providing knowledge increases or aptitude for behavioral change decreases, fewer states are in the optimal knowledge provision policy, with healthy consumers dropping out first. Knowledge about diet and physical activity is provided more selectively due to the to uncertainty in the health benefits, and the time delay in accrued rewards.
Pope, Brandon 1984- (2011). Engineering Incentives in Distributed Systems with Healthcare Applications. Doctoral dissertation, Texas A & M University. Available electronically from