Every day, millions of Americans use the Internet to obtain health information. To satisfy this demand, organizations deliver a variety of content that promotes awareness and education and informs health- related decision making. Given advances in web technology, new statistical models of disease, and the shift towards shared patient decision making, these e-health services are increasingly complex. Through applications such as personal health records and "health risk calculators" Internet users can obtain personalized and interactive feedback about their current health state, model-based predictions about their future health, and tailored education about healthy behavior. While providing the public with more content to inform health-related decisions is an appropriate goal, research in health psychology and behavioral decision making suggest the importance of clearly understanding the perceptual and behavioral responses when laypersons are presented with statistical results and personalized risk information. Little research has studied how web-based personalized and interactive health applications actually impact the beliefs and behavior of users. In two separate experiments, this research measured the effect of a type 2 diabetes "risk calculator" website on user information processing and subjective risk perceptions about diabetes. In the first experiment, 100 middle-aged and elderly adults were randomized to one of three conditions in order to determine how personalized risk estimates and interactive risk feedback influenced information usage and beliefs about future diabetes onset. Results showed that personalization and interactive features did not lead to increases in information utilization as expected. Instead, this research shows in some cases personalization actually reduced the amount of information accessed and the extent to which users attended to and carefully considered health risk content. The experiment did show that personalization was related to modest increases in the accuracy of absolute diabetes risk estimates but did not motivate significant changes in relative or affective risk perceptions. A second study of 34 university staff members was qualitatively suggestive of similar results. Future work is needed to further understand the behavioral implications when complex statistical models are integrated with publicly available health information websites. This may aid the design of health information applications and ensure that providers of these tools are effectively motivating improved awareness and education about health risks.
Adobe .pdf files require the Adobe Acrobat Reader.
Download and install the Adobe Acrobat Reader.