Clinical Decision Science emphasizes unique social context in a way that Evidence-Based Medicine does not

INTRODUCTION: The goal of evidence-based medicine includes the integration of clinical experience and patient values with research evidence. We introduce clinical decision science, a new framework that includes patient social context to demonstrate this integration, which has been absent from evidence-based medicine sources. METHODS: This is an observational study comparing published articles within the domains of clinical decision science and evidence-based medicine. In a standardized manner, investigators identified and counted instances of social interaction within the publications. RESULTS: Publications of Clinical Decision Science had a higher number of markers of social interaction per paper and greater proportion of papers that included any markers of social interaction compared to publications in the Evidence-based medicine domain. DISCUSSION: We identified a framework that allows exploration of a new scientific domain that includes both research evidence and individual patient social context.


Introduction
The introduction of evidence-based medicine (EBM) in the early 1990's has fundamentally changed the process by which physicians and patients decide their next steps in patient care. Evidence-based medicine has sought to bring a more rigorous and scientific approach to both the development of guidelines and policies, as well as clinical decisions in the care of individual patients. 1 Early on, there was recognition that rigor, science, and evidence were necessary, but not sufficient, for optimal decision making. Evidencebased medicine needed to "[integrate] clinical experience and patient values with the best available research information." 2 While individual physicians may achieve this lofty goal, we can find no description of this process in the medical literature. There is no user's manual, no formal teaching, and no defined skillset. We rely on physicians to develop this ability on their own. Once they escape the confines of post-graduate training, most physicians will be left with only minimal and infrequent guidance and even less feedback. We have no evidence that EBM has achieved rigor and application of science to individual decision making in clinical settings. Prior to data collection, the investigators met as a group to train and standardize data collection using the codebook. A draft of the codebook was given to the group and discussed in detail, including any clarifying questions or concerns being addressed by the investigators. A training journal article from Clinical Research in Practice: The Journal of Team Hippocrates (CRP-JTH) was then distributed and reviewed by each investigator individually. The team then reconvened and read through the article as a group and discussed similarities and differences in their coding until a consensus was reached. The codebook was then revised to reflect this consensus. For example, the category "behavior" was removed, as the investigation team thought it was too broadly defined, and redundant with other categories of social context. Further, the group listened to several podcast episodes and coded each individually. Similar to the journal article, the group then reconvened and discussed similarities and differences of how they graded each podcast until a consensus was achieved, and the codebook similarly revised. The final codebook is included in the supplemental material as Appendix A.
In the domain of clinical decision science, 30 sequential, most recently published articles available at the time were taken from CRP-JTH, each published between August 2018 and September 2019. Thirty articles were used based on a guestimate for the power calculation, given an estimated 0.5 prevalence of recorded social markers in EBM compared to 2.0 in CRP-JTH. Representing the domain of evidence-based medicine, 31 total articles were randomly chosen from various EBM databases (indexed EBM aggregators) and publications. Up to 6 contiguous pages were taken from each article, starting at a random page number. These randomly selected articles were taken from DynaMed (n=5 articles), Essential Evidence Plus (n=5), UpToDate (n=6), Family Physician Inquiries Network (n=8), American Family Physician (n=4) and American College of Physicians Journal Club (n=4). Each article was assigned to two investigators to code separately and independently, using the codebook. These investigators were non-contributors and non-readers of CRP-JTH. Different colored highlighters were used to indicate text that described each descriptor of social context in the codebook.
Another EBM format exists that we thought it important to compare-podcasts. Podcasts tend to be informal discussions between clinicians and might have a greater possibility for describing how evidence is used in clinical care. So, 10 podcasts were sampled as well. Random single segments were selected from various podcast programs, with each segment having a duration of about 2-3 minutes. Sampled podcasts include American Family Physician Podcast (n=4), Broome Docs (n=3) and Primary Care Medical Abstracts (n=3). Investigators met together to listen to each podcast simultaneously, but each investigator coded the podcast separately and independently. Investigators used a common abstraction form to take rough notes on phrases that indicated social context.
The duplicate coding of each case was reviewed and compared to each other. Disagreements between coders were resolved by consensus of two authors (JM, NY), both of whom are familiar with the two domains and the formats reviewed and coded. The total number of instances of each social indicator was collected and recorded for each individual article and podcast. The total number of instances of social context were aggregated for CRP-JTH, EBM print articles, and EBM podcasts.

Results
A total of 71 cases were coded (CRP-JTH n=30, EBM print articles n=31, EBM podcasts n=10). For one EBM print article, consensus could not be reached and this case was not included in analysis. We compared the average social markers per publication in each group, which revealed 4.7 per CRP-JTH publication, 0.5 per EBM print article and 0.2 per EBM podcast (p<0.00001, Kruskal-Wallis). Data for individual social marker categories is available in Figure 1. These data highlight that the two groups were most similar in reporting "Institutional Relationships" and had a wide discrepancy of frequency in the other.

Figure 1. Average Social Markers per Publication
We also analyzed the percentage of cases in each domain containing a marker of social context. Results are demonstrated in Figure  2. These data highlight that social information comes to clinicians most often in the form of conversations and patient values or preferences. Emotions and family or personal information are also important sources of social context.