Rheumatologist and Assistant Professor, Bella Mehta, MBBS, MS, MD, Division of Rheumatology at Hospital for Special Surgery, Department of Medicine, Weill Cornell Medicine explored whether social determinants of health (SDOH), particularly community-level factors, had an impact on surgical outcomes after total hip arthroplasty (THA). Dr. Mehta and her research team used statewide data from the Pennsylvania Health Care Cost Containment Council (PHC4) and modern, interpretable machine learning methods for this study entitled, Comparing Community-Level Social Determinants of Health With Patient Race in Total Hip Arthroplasty Outcomes1.
Discharge records for 105,000 patients who underwent elective primary THA in Pennsylvania from 2012 to 2018 were linked with U.S. Census data to obtain community-level SDOH metrics, such as income, education, internet access, health insurance coverage, and walkability of the neighborhood. Using this data, the team trained explainable boosting machine (EBM) models and evaluated the importance of each variable by its contribution to prediction accuracy. Dr. Mehta explained, “These supervised machine learning models showed community-level factors consistently exhibited greater power than race in predicting all outcomes including 90-day readmission, mortality, revision, and length of stay. These factors were strongly associated with recovery environments and support systems following THA. Discharge plans and postoperative support should be tailored with these elements in mind.”
Dr. Mehta highlighted the major results stating, “Our study highlights the importance of integrating machine learning with traditional health services research to uncover actionable insights. We believe future policy and clinical interventions should incorporate neighborhood and environmental factors, which may be more modifiable and impactful than immutable demographic attributes like race.“
Dr. Mehta concluded, “PHC4 is a valuable, comprehensive resource for researchers studying healthcare utilization, outcomes, or disparities in Pennsylvania, especially since their data can be linked with external datasets for enhanced contextual insights. The documentation and structure of PHC4’s data were clear and manageable, and PHC4 staff were available to clarify all nuances when analyzing the data.”
1 Mehta B, Yiyuan Y, Pearce-Fisher D, Ho K, Goodman S, Parks M, Wang F, Fontana M, Ibrahim S, Cram P, Caruana R. Comparing Community-Level Social Determinants of Health With Patient Race in Total Hip Arthroplasty Outcomes. American College of Rheumatology. 24 February 2025. doi: 10.1002/acr.25511.
Return to the Case Studies Page: VIEW ALL CASE STUDIES
