Advertisement

P93 Precision Behavioral Nutrition: Development of the NutriPCP Inference Engine for Data-driven Diet Goals in Primary Care

      Objective

      To develop a computational system that uses dietary recall data to prioritize behavioral goals to facilitate efficient, personalized collaborative goal-setting in primary care.

      Use of Theory or Research

      The Chronic Care Model posits that synergy between the healthcare system and patient self-management will improve chronic disease outcomes. Thus, improving how diet is addressed in primary care could augment the benefit of dietary self-management. Collaborative goal-setting with primary care providers (PCPs) can facilitate patient behavior change. However, PCPs lack time and training to set effective diet goals with patients. NutriPCP aims to address this gap by presenting PCPs with a set of evidence-based goals prioritized using patient data.

      Target Audience

      PCPs

      Program Description

      NutriPCP uses ASA24 diet recall data to compute patient status for each of 9 previously developed, MyPlate-based goals (eg, “Make half my grains whole”). NutriPCP's inference engine consists of Python rule statements that synthesize a patient's data and compare it to evidence-based targets for nutrient consumption personalized for patient characteristics (eg, kcal intake/sex/age). PCPs are then presented with a list of the patient's status for each goal prioritized by degree of improvement needed.

      Evaluation Methods

      We tested our inference engine with test data (n = 12), and our team of nutrition, technology, and clinical experts validated the output. We used NHANES data to establish reasonable population-wide estimates.

      Results

      Testing revealed challenges for goal prioritization because datasets reflected consumption far from evidence-based targets. Therefore, we created standardized ranges to improve variability for relative ranking across goals. For goals with upper and lower limits (eg, “Reduce portion size”) we added warnings for inadequate intake.

      Conclusion

      We demonstrated that computational rules can automatically process recall data into prioritized behavioral goals. To our knowledge, this is the first system that personalizes MyPlate recommendations based on an individual's data. This has implications for nutrition education in primary care. Future research will examine implementation feasibility for PCPs and patients.
      Funding Agency for Healthcare Research and Quality (AHRQ) R21HS027660 Burgermaster, PI.

      Appendix. Supplementary data