Motivated by the need for new techniques for analyses of health care costs from clinical and epidemiological studies, the goal of this study is to develop statistical techniques that fill methodological gaps in current CEA models. When a health care intervention is deployed, cost are engendered though the use of resources. These occur in random amounts at random times that might differ by patient attributes, clinical and intervention characteristics. Study design and sampling may lead to incomplete observation of key outcomes in some patients. Incorporating these components into statistical models that accurately reflect the experience of patients as their health histories manifest over time permits consideration of health outcomes and costs jointly. This study proposes statistical models of costs and patient outcomes that incorporate the temporal dynamics that generate these data and the availability of concomitant exogenous information. They propose to use a Markovian regime to model the dynamics of movement of patients through different health states, with costs incurred at transitions between states and sojourns within states. Applying a constant discount rate where appropriate, they will estimate summary health outcome measures (e.g. life expectancy, quality-adjusted life years, net present value, net health cost and benefit, cost-effectiveness ratios). They will assess the impact of exogenous factors on these parameters and provide a unified framework for statistical inference and then test the performance and sensitivity of the procedures with real and simulated data. They propose applications of the methods to several peer-reviewed investigations. 1) They will assess the determinants of total health care costs and the use and cost of pharmacologic treatments in a longitudinal study of quality and continuity of care to 9,000 Medicaid recipients with a diagnosis of attention deficit hyperactivity disorder. 2) They will estimate cost of hospitalization in relation to co-morbidity, patient demographic, and clinical attributes in a retrospective study of cardiac procedures performed in two cohorts of patients with myocardial infarction. 3) From the MADIT they will assess resource utilization and and survival in patients at high risk for ventricular arrhythmia, and re-examine the cost-effectiveness of the defibrillator compared to conventional therapy. 4) They will assess (relative to usual care) incremental costs, net health benefit and net cost measures in a trial of an intervention designed to improve mental health and physical health functioning and life course development in low-income pregnant women. 5) They will assess the determinants of cost by providing treatment in a study of stage of diagnosis, treatment, and survival in patients with cancer in Medicaid In summary, by developing and testing new methods for cost-effectiveness studies, and demonstrating their application in several ongoing clinical studies, this study not only offers an array of promising techniques, but also bridges the gap between methodological development and implementation.
