Development and Assessment of Algorithms for Delivering Tailored or Targeted Patient Decision Support in Two Disease Models

Thesis (Ph.D.)--University of Washington, 2013Patient decision support refers to the provision of appropriate information to a patient about his or her health conditions to aid in the goal of informed medical decision-making. Medical decisions in which there are multiple valid options in which the risks and benefits differ, deemed "preference-sensitive" decisions, necessitate shared decision-making and patient decision support. Tailoring and targeting can be used to adapt the content or the amount of information delivered to the patient. Tailored or targeted health information for patient decision support has long been shown to be more effective than generic information at modifying health behaviors. To increase adoption in clinical care, one goal in developing targeting frameworks is to choose theory constructs and personal characteristics measured by a few data items that are simple to collect. Thus, in developing parsimonious algorithms, it is important to understand the relationships between socio-demographic characteristics and theories of health behavior. Finally, to ensure acceptance of algorithms in patient decision support, it is critical to understand patient perceptions of such algorithms, particularly what additional factors are important in decision-making. This dissertation addresses these needs in two parts: Part 1) development of algorithms to guide content and intensity of patient decision support in the context of colorectal cancer screening, and Part 2) assessment of patient perceptions of the use of algorithms to guide patient decision support in the context of heart disease prevention. In the Part 1 of this dissertation, I conducted two studies that consider health beliefs, demographics, and patient behavior in the context of colorectal cancer screening behavior. The first study sought to understand the relationship between socio-demographic characteristics and constructs of the health belief model in the context of colorectal cancer screening behavior in a racially, ethnically, and linguistically diverse population. The second study sought to develop a practical regression model to predict the probability of completing colorectal cancer screening and, from this model, a framework for targeting patient decision support materials based on level of probability of completing CRC screening. In Part 2 of this dissertation, I conducted one study to assess the use of an algorithm in patient decision support for heart disease prevention therapy. This study was a qualitative analysis of interviews conducted after a discrete choice experiment (DCE) for heart disease prevention. I sought to understand new factors that were influencing heart disease prevention therapy choices and perception of DCE-based "values concordant" choice results. In the first part of the dissertation, I observed several key differences across racial/ethnic and language groups in terms of health beliefs about CRC and CRC screening. These differences were largely dependent upon primary spoken language, which may approximate level of acculturation. Non-English speaking Hispanics typically reported lower perceived susceptibility to CRC than non-Hispanic Whites, and higher perception of several barriers to CRC screening (prior testing experience, preparation for the test, need for sedation). Non-English speaking Asians also reported lower perceived susceptibility than non-Hispanic Whites, but lower perception of several barriers to CRC screening (need for additional testing, fear of results of the test, concern about complications from the test, need for sedation, anxiety about the procedure). These results may suggest topic areas that could be highlighted in CRC screening promotion interventions that are targeted at specific racial/ethnic and language groups. In the second study, I developed a simple model for predicting CRC screening completion. From that model I developed an intervention framework that may be useful for targeting the amount of information to patients based on how likely they are to complete screening. In the final study, I found several new factors that were influencing heart disease prevention decisions: medication avoidance/naturalness, competing demands, and familiarity. Participants were receptive to the DCE-based "values concordant" choice, even if they did not ultimately agree with it. The results of this dissertation may be informative to patient decision support researchers considering methods for targeting or tailoring decision support intervention for CRC screening or heart disease prevention. Future research should confirm the differences in health beliefs about CRC that we observed across racial/ethnic and language groups. Additionally, the targeting framework that we developed for CRC screening promotion interventions should be evaluated. Finally, future work in DCE for heart disease prevention therapy decision-making should incorporate the new attributes