(C-133) Identifying physician beliefs associated with higher concordance with risk assessment model treatment recommendations for the prevention of venous thromboembolism in an inpatient setting

Background: Risk assessment models (RAMs) can support physicians in personalizing care. For example, RAM can inform on the appropriateness of chemoprophylaxis to prevent venous thromboembolism (VTE) in hospitalized patients.

Objectives: To quantify physicians’ beliefs about RAM and to identify which beliefs are associated with higher concordance with the RAM treatment recommendations for chemoprophylaxis for VTE.

Methods: We surveyed physicians within a large health system who had used the RAM at least 50 times. Physician beliefs about RAM using best-worst scaling (BWS), a theory-driven choice experiment. The BWS assessed seven possible beliefs across seven choice tasks. In each task, four beliefs were displayed, and physicians were asked to select concepts that were most and least reflective of them. The BWS was analyzed using a conditional logistic regression, stratified those who were, and were not, concordant with the RAM recommendation most of the time (i.e. exceeding >66%).

Results: Of 434 eligible physicians, 188 (43%) completed the survey. The average rate of concordance with RAM recommendations was 59% and varied from 11-100%. Beliefs were different between high and low concordance physicians (p < 0.001). High concordance physicians were more likely to believe that ‘the model increases decision-making confidence’ (p-value < 0.001), ‘the model reduces overtreatment for low-risk patients’ (p < 0.001), and ‘the model is trustworthy’ (p=0.003). Low concordance physicians were more likely to believe that ‘the model disrupts their workflow’ (p < 0.001), that ‘important risk factors are missing from the model’ (p < 0.001), and that they ‘use the model because they need to follow institutional policies’ (p=0.008).

Conclusions: Physicians’ beliefs are associated with their concordance with RAM recommendations for VTE. The development and implementation of RAM in the hospital setting must be guided by a deeper understanding of physician beliefs and experiences.