Context, automated decision support, and clinical practice guidelines. Does the literature apply to the United States practice environment?
With the goal of determining whether existing research on clinical practice guideline use is applicable to United States physician practice, Edmonson and colleagues systematically reviewed articles published in the past 10 years that addressed effects of automated decision support (ADS) on guideline adherence. Articles were also included if they addressed related patient outcomes or physician opinions about ADS implementation. Literature was ascertained through searching of Medline, Web of Science, Ovid EBM and EBSCO databases as well as scanning of retrieved article bibliographies for missed references.
High inter-rater reliability was observed between the authors in identifying appropriate references and assigning specific characteristics to each reference. Overall, only 55% of articles provided a quantitative measure of behavioral change with ADS as compared to a control condition with other articles providing qualitative data (12%) or measuring existing behavior (6%) or behavioral changes (27%) without a control group. Because adherence to clinical practice guidelines may be context dependent, the authors examined each article for the guideline topic being addressed and the practice setting as well as for physician characteristics (specialty vs. primary care practice; academic vs. private practice; fee for service vs. salaried compensation).
As compared to national proportions for each of these factors, the authors found highly significant differences (p<.0001) in published studies of ADS and practice guidelines. More specifically, 65.3% of the reviewed studies involved academic physicians yet only 11% of physicians report academic affiliations nationally. Approximately 24% of the reviewed studies involved specialists, yet specialists represent 65% of physicians nationally and account for about half of physician visits. Only 14% of reviewed articles explicitly included salaried physicians whereas primary salaried compensation is reported by about half of U.S. physicians. In terms of practice settings, national comparison figures were not reported but the majority of studies occurred in outpatient or acute hospital settings with fewer articles addressing effects of ADS in emergency department, skilled nursing or long-term care facilities. Mismatches were also observed between the number of studies addressing specific guideline content areas (e.g., infectious disease, cardiovascular, endocrine) and either the number of published guidelines or the prevalence or seriousness of specific diagnoses.
Overall, the authors concluded that the clinical contexts in which recent studies occurred is not representative of national contexts. Just as importantly, some guideline topics, physician characteristics and settings are inadequately studied making it difficult to draw conclusions about ADS utility or make policy decisions about their use.
Comment: In this article, the authors have provided a methodologically rigorous systematic review of recent literature on ADS and clinical practice guidelines. Their conclusion about the limited generalizability of recent research on ADS and practice guideline adherence is not surprising given the huge number of clinical practice guidelines available for study and the concentration of most research in academic inpatient and outpatient settings.
However, even if a broader range of settings, practitioners and guideline topics were studied in greater detail, the findings may still not be application to an individual physician practice or institution given the wide variability in patient populations and physician behaviors even with specialties or settings.
Despite this, it would have been interesting to see whether physicians with particular characteristics differed in the extent to which they adhered to practice guidelines with ADS or whether physician behaviors or patient outcomes differed across settings although the variety of studies may have made comparisons difficult. In addition, the fact that few of the studies described physician characteristics (such as primary form of compensation) suggests that future work should be more specific in describing the research context.