Non-adherence to medications is a ubiquitous problem that is incompletely understood. It increases costs and wastes valuable resources. In hypertension, non-adherence places patients at risk of stroke and heart disease. In Human Immunodeficiency Virus (HIV) infection, non-adherence not only leads to progression of disease, but also threatens to overturn the hard-won advances by facilitating the breeding of resistant strains of HIV. The goal of this study is to further understanding of the root causes of non-adherence through the study of preferences. Our central hypothesis is that when a patient is non-adherent, it is often the result of his or her judgment that the overall cost-to-benefit ratio of proper use of a medication is unfavorable. Widely applied behavioral models of non-adherence, such as the Health Belief Model, typically acknowledge the central role of a patient's weighing of the risks and benefits in his or her subsequent adoption of desired health-behaviors. However, most previous research in this area has applied relatively crude methods to measure the perceived benefit of adherence. In this study, we will take a detailed look at the components of risk-benefit judgements using computer interviews to measure the utility of adherence and other aspects of preferences. We will then determine the relationship between the utility of adherence and medication use behavior. Most models of adherence recognize the influence of factors such as ethnicity, drug abuse, and perceived vulnerability to disease and on health behaviors. Many of these factors may be mediated through effects on preferences. Therefore, this study will also examine associations among demographic, psychosocial factors and patients' utility scores for health outcomes. The design of the study is a cross-sectional survey with longitudinal follow up. Studies will be conducted in two patient groups (HIV and hypertension) in order to assess the generalizability of findings to other populations. Data on patients' preferences will be collected using the standard gamble, time trade-off, and other utility metrics (including risk attitude) for favorable and unfavorable life courses with high and low adherence efforts and moderate and low adverse effects. We will then estimate the marginal disutility of adherence from these multi-attribute measures. Other data to be collected include demographic factors, psychosocial factors, and estimates of the effects of adherence on the probability of complications. The explanatory power of demographic and psychosocial factors and utilities will be compared using logistic regression and CART. We will also use an expert panel to assess the clinical match between patients' regimens and their preferences.
