Tyson Amundsen, MD1, John S. Kelley, MD, MBA2, Rajesh Shah, MD, MS3, Kenneth Coggins, MD1, Christopher M. Johnson, MD, PhD1 1Baylor Scott & White Health, Temple, TX; 2Baylor Scott and White, Lubbock, TX; 3Baylor Scott & White Health, Austin, TX
Introduction: Clinical information are typically used to predict adverse outcomes in patients with inflammatory bowel disease (IBD). There is a growing recognition of the importance of social determinants of health (SDOH) influencing important outcomes. Our study aimed to investigate the influence of SDOH such as food security, housing stability and transportation, in addition to clinical risk factors on adverse outcomes in patients with IBD.
Methods: We conducted a retrospective cohort study of adult patients with IBD from the years 1996 to 2019 seen within the Baylor Scott & White System. Patients were identified using ICD-10 codes for ulcerative colitis (UC) and Crohn’s disease (CD). Chart review was performed to validate the diagnosis and extract demographic and clinical information. SDOH information regarding food security, financial resources, and transportation were self-reported by the patient. The outcomes of interest were IBD related hospitalization or surgery, which were defined based on discharge or CPT coding. Random Forest models to predict IBD related hospitalizations or surgeries were trained and tested using the random forest and caret packages with either clinical only or clinical plus SDOH information.
Results: 175 patients met inclusion criteria and 75 patients had the outcome of interest. Among patients with an IBD related hospitalization or surgery, the majority of patients reported no financial resource strain (87%), no problems with food security (worried or scared, 89% and 91% respectively) or lack transportation resources to attend medical appointments. For the model using the whole cohort and only clinical information, the sensitivity was 0.77 and specificity was 0.55 with an area under the receiver operating characterisic curve (AUROC) of 0.74. With the addition of SDOH information, the sensitivity and specificity were 0.67 and 0.55 with an AUROC of 0.71. The most important variables influencing the model performance were tobacco use, disease extent, race, alcohol use and food security. Model performance improved when analyzing by phenotype with an AUROC of 0.91 for patients with CD and AUROC of 0.79 for patients with UC using combined clinical and SDOH information.
Discussion: The addition of SDOH information to clinical risk factors for adverse outcomes did not improve model accuracy; however, model performance improved when examining by phenotype. Further research is needed to understand the potential role of SDOH factors with respect to IBD related outcomes.
Tyson Amundsen indicated no relevant financial relationships.
John Kelley indicated no relevant financial relationships.
Rajesh Shah indicated no relevant financial relationships.
Kenneth Coggins indicated no relevant financial relationships.
Christopher Johnson indicated no relevant financial relationships.
Tyson Amundsen, MD1, John S. Kelley, MD, MBA2, Rajesh Shah, MD, MS3, Kenneth Coggins, MD1, Christopher M. Johnson, MD, PhD1. P1604 - Medical and Social Determinants of Health as Predictors of Adverse Outcomes in Patients With Inflammatory Bowel Disease, ACG 2021 Annual Scientific Meeting Abstracts. Las Vegas, Nevada: American College of Gastroenterology.