Optimizing Healthcare Policies through Healthcare Delivery and Insurance Design

Thesis (Ph.D.)--University of Washington, 2019The problems of efficient prevention, screening, and treatment interventions for chronic diseases and acute infections have received much attention in recent years because of the rise in American healthcare costs. Chronic diseases generally cannot be prevented by vaccines or cured by one-time medication; however, most of them are controllable through lifestyle changes, early detection, and treatment. Improving the healthcare delivery system and encouraging access and adherence to medical care provide key opportunities for better prevention and control of multi-stage chronic diseases. On the other hand, an acute outbreak of a disease can be prevented with vaccination, an effective prevention intervention; however, low vaccination rate and free-riding behavior in vaccination decisions exist. In this case, healthcare policies need to be optimally and efficiently designed to improve sequential decision making to implement interventions for chronic disease and to prevent the outbreak of acute infectious diseases. This research has two contributions: building sequential decision support models for three healthcare delivery and insurance design problems, and developing corresponding methodologies that can be used to effectively investigate these healthcare problems. The three healthcare policy optimization problems in this dissertation are: (i) the implementation of optimal sequential screening and treatment policy with a budget constraint for a chronic hepatitis C virus (HCV) birth-cohort to improve population health outcomes; (ii) a long-term healthcare insurance cost-sharing policy design for patients with hypertensive heart disease to minimize public/private insurer’s and insurants’ costs; (iii) the design of a vaccination reimbursement and cost-sharing policy to reduce medical treatment cost and prevent the spread of seasonal influenza. The first contribution of this dissertation is to build decision support models that represent disease transmission and progression for these three health healthcare applications. These models capture the uncertainties of disease transmission/progression and interactions between decision makers and target population. The three problems are modeled with stochastic processes, game theoretic approaches, and agent-based simulations to forecast population health, intervention effects, and costs over time; however, these problems are complicated and determining an optimal healthcare policy is challenging. Therefore, efficient methodologies need to be developed so that optimal/near-optimal healthcare policies can be easily implemented. The second contribution is to develop three efficient methodologies that can be applied to three corresponding healthcare policy optimization problems. The methodologies include several common important characteristics. The first characteristic is to approximate models to alleviate the curse of dimensionality. These methodologies focus on building low-fidelity optimization problems or models so that a healthcare policy with good quality can be easily calculated. The low-fidelity approximation allows decision makers to efficiently identify a near-optimal healthcare intervention policy. The second characteristic is to use the structure of the problems to bound or guarantee an optimal solution so that these methodologies can be extended to large-scale healthcare management problems and further improve solution quality. The second objective focuses on incorporating global optimization techniques or optimal control theorems into methodologies to provide an optimal or near-optimal solution with improved computation time. The improvement allows decision makers to trade-off computation effort with policy quality in large-scale healthcare management problems. These decision support models: (i) provide useful insights regarding HCV elimination to maximize overall health outcomes measured by quality-adjusted life-years (QALYs) for different annual budgets and birth-cohorts in baby boomers; (ii) design flexible and differential cost-sharing provisions for specific patient groups to incent the patients with multi-stage chronic diseases for promoting healthcare service choices; (iii) enable a public/private insurer to set optimal insurance policies, including reimbursement and cost-sharing rate, to maximize its utility which is in terms of medication and vaccination cost and social benefit during the flu season. The illustrations of intervention implementation for these three diseases pave the way for new, and more efficient methods for preventing and controlling other disease problems. Also, the models and methodologies applied to these three healthcare policies provide contributions to understanding and future implementations to healthcare delivery and insurance design