In medicine, computers are increasingly used not only to document patient care, but to support clinical decisions with relevant medical knowledge and patient information. This function is referred to as clinical decision support (CDS). Decades of scholarship have focused on evaluating CDS effectiveness, with promising but highly variable results. Having established that CDS can be useful, many have turned to more complex questions: what factors make CDS more effective and better-received by its users? How does CDS fit in with other strategies meant to ensure that clinical decisions are informed by current medical knowledge? When CDS works well, why and how does it do so? These are particularly important questions within the Department of Veterans Affairs (VA), which has a long history of using CDS, and has a massive scale that can enable CDS improvements to affect millions of patients. In this dissertation, I explore the use of CDS in VA primary care clinics. In study 1, I use VA-wide survey data to evaluate "user-centered design" strategies intended to make CDS easier to use and more effective. In study 2, I link that survey data with administrative records on colorectal cancer screening to examine CDS as one of several strategies for implementing evidence-based practices. In study 3, I use data from semi-structured interviews with primary care providers at VA clinics around Los Angeles to understand how CDS works for a specific clinical decision (prostate cancer screening).
I find that “analysis of impact on performance improvement” is positively associated with perceived utility of CDS, but no association is evident for the other three user-centered design strategies examined. In assessing CDS efficacy alongside other strategies to support colorectal cancer screening, I find that neither CDS, nor any other implementation strategy examined, is associated with screening. In the qualitative study, I identify key factors in the effectiveness and acceptability of CDS for prostate cancer screening, including: workflow compatibility, the use of a trusted clinical guideline, and consultation with the intended users of CDS as part of its implementation.
Conclusion: Judicious implementation and governance of CDS are important determinants of its usefulness.