Shirlene Obuobi1, Rhys F M Chua2, Stephanie A Besser2, Corey E Tabit3. 1. Department of Medicine, University of Chicago, Chicago, IL, USA. 2. Section of Cardiology, Department of Medicine, Chicago, IL, USA. 3. Section of Cardiology, Department of Medicine, Chicago, IL, USA. corey.tabit@uchospitals.edu.
Abstract
BACKGROUND: The HOSPITAL Risk Score (HRS) predicts 30-day hospital readmissions and is internationally validated. Social determinants of health (SDOH) such as low socioeconomic status (SES) affect health outcomes and have been postulated to affect readmission rates. We hypothesized that adding SDOH to the HRS could improve its predictive accuracy. METHODS: Records of 37,105 inpatient admissions at the University of Chicago Medical Center were reviewed. HRS was calculated for each patient. Census tract-level SDOH then were combined with the HRS and the performance of the resultant "Social HRS" was compared against the HRS. Patients then were assigned to 1 of 7 typologies defined by their SDOH and a balanced dataset of 14,235 admissions was sampled from the larger dataset to avoid over-representation by any 1 sociodemographic group. Principal component analysis and multivariable linear regression then were performed to determine the effect of SDOH on the HRS. RESULTS: The c-statistic for the HRS predicting 30-day readmission was 0.74, consistent with published values. However, the addition of SDOH to the HRS did not improve the c-statistic (0.71). Patients with unfavorable SDOH (no high-school, limited English, crowded housing, disabilities, and age > 65 yrs) had significantly higher HRS (p < 0.05 for all). Overall, SDOH explained 0.2% of the HRS. CONCLUSION: At an urban tertiary care center, the addition of census tract-level SDOH to the HRS did not improve its predictive power. Rather, the effects of SDOH are already reflected in the HRS.
BACKGROUND: The HOSPITAL Risk Score (HRS) predicts 30-day hospital readmissions and is internationally validated. Social determinants of health (SDOH) such as low socioeconomic status (SES) affect health outcomes and have been postulated to affect readmission rates. We hypothesized that adding SDOH to the HRS could improve its predictive accuracy. METHODS: Records of 37,105 inpatient admissions at the University of Chicago Medical Center were reviewed. HRS was calculated for each patient. Census tract-level SDOH then were combined with the HRS and the performance of the resultant "Social HRS" was compared against the HRS. Patients then were assigned to 1 of 7 typologies defined by their SDOH and a balanced dataset of 14,235 admissions was sampled from the larger dataset to avoid over-representation by any 1 sociodemographic group. Principal component analysis and multivariable linear regression then were performed to determine the effect of SDOH on the HRS. RESULTS: The c-statistic for the HRS predicting 30-day readmission was 0.74, consistent with published values. However, the addition of SDOH to the HRS did not improve the c-statistic (0.71). Patients with unfavorable SDOH (no high-school, limited English, crowded housing, disabilities, and age > 65 yrs) had significantly higher HRS (p < 0.05 for all). Overall, SDOH explained 0.2% of the HRS. CONCLUSION: At an urban tertiary care center, the addition of census tract-level SDOH to the HRS did not improve its predictive power. Rather, the effects of SDOH are already reflected in the HRS.
Authors: Susannah M Bernheim; Craig S Parzynski; Leora Horwitz; Zhenqiu Lin; Michael J Araas; Joseph S Ross; Elizabeth E Drye; Lisa G Suter; Sharon-Lise T Normand; Harlan M Krumholz Journal: Health Aff (Millwood) Date: 2016-08-01 Impact factor: 6.301
Authors: Eric T Roberts; Alan M Zaslavsky; Michael L Barnett; Bruce E Landon; Lin Ding; J Michael McWilliams Journal: JAMA Intern Med Date: 2018-11-01 Impact factor: 21.873