Connor Drake1,2, Tyler Lian3, Justin G Trogdon4, David Edelman5,6, Howard Eisenson7,8, Morris Weinberger4, Kristin Reiter4, Christopher M Shea4. 1. Department of Population Health Sciences, Duke University School of Medicine, 215 Morris Street, Durham, NC, 27701, USA. connor.drake@duke.edu. 2. Department of Health Policy and Management, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, 135 Dauer Dr, Chapel Hill, NC, 27519, USA. connor.drake@duke.edu. 3. Department of Population Health Sciences, Duke University School of Medicine, 215 Morris Street, Durham, NC, 27701, USA. 4. Department of Health Policy and Management, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, 135 Dauer Dr, Chapel Hill, NC, 27519, USA. 5. Department of Medicine, Duke University School of Medicine, 2301 Erwin Rd, Durham, NC, 27705, USA. 6. Durham VA Healthcare System, 508 Fulton St, Durham, NC, 27705, USA. 7. Lincoln Community Health Center, 1301 Fayetteville St, Durham, NC, 27707, USA. 8. Department of Family Medicine and Community Health, Duke University School of Medicine, DUMC 2914, Durham, NC, 27710, USA.
Abstract
BACKGROUND: Health systems are increasingly using standardized social needs screening and response protocols including the Protocol for Responding to and Assessing Patients' Risks, Assets, and Experiences (PRAPARE) to improve population health and equity; despite established relationships between the social determinants of health and health outcomes, little is known about the associations between standardized social needs assessment information and patients' clinical condition. METHODS: In this cross-sectional study, we examined the relationship between social needs screening assessment data and measures of cardiometabolic clinical health from electronic health records data using two modelling approaches: a backward stepwise logistic regression and a least absolute selection and shrinkage operation (LASSO) logistic regression. Primary outcomes were dichotomized cardiometabolic measures related to obesity, hypertension, and atherosclerotic cardiovascular disease (ASCVD) 10-year risk. Nested models were built to evaluate the utility of social needs assessment data from PRAPARE for risk prediction, stratification, and population health management. RESULTS: Social needs related to lack of housing, unemployment, stress, access to medicine or health care, and inability to afford phone service were consistently associated with cardiometabolic risk across models. Model fit, as measured by the c-statistic, was poor for predicting obesity (logistic = 0.586; LASSO = 0.587), moderate for stage 1 hypertension (logistic = 0.703; LASSO = 0.688), and high for borderline ASCVD risk (logistic = 0.954; LASSO = 0.950). CONCLUSIONS: Associations between social needs assessment data and clinical outcomes vary by cardiometabolic condition. Social needs assessment data may be useful for prospectively identifying patients at heightened cardiometabolic risk; however, there are limits to the utility of social needs data for improving predictive performance.
BACKGROUND: Health systems are increasingly using standardized social needs screening and response protocols including the Protocol for Responding to and Assessing Patients' Risks, Assets, and Experiences (PRAPARE) to improve population health and equity; despite established relationships between the social determinants of health and health outcomes, little is known about the associations between standardized social needs assessment information and patients' clinical condition. METHODS: In this cross-sectional study, we examined the relationship between social needs screening assessment data and measures of cardiometabolic clinical health from electronic health records data using two modelling approaches: a backward stepwise logistic regression and a least absolute selection and shrinkage operation (LASSO) logistic regression. Primary outcomes were dichotomized cardiometabolic measures related to obesity, hypertension, and atherosclerotic cardiovascular disease (ASCVD) 10-year risk. Nested models were built to evaluate the utility of social needs assessment data from PRAPARE for risk prediction, stratification, and population health management. RESULTS: Social needs related to lack of housing, unemployment, stress, access to medicine or health care, and inability to afford phone service were consistently associated with cardiometabolic risk across models. Model fit, as measured by the c-statistic, was poor for predicting obesity (logistic = 0.586; LASSO = 0.587), moderate for stage 1 hypertension (logistic = 0.703; LASSO = 0.688), and high for borderline ASCVD risk (logistic = 0.954; LASSO = 0.950). CONCLUSIONS: Associations between social needs assessment data and clinical outcomes vary by cardiometabolic condition. Social needs assessment data may be useful for prospectively identifying patients at heightened cardiometabolic risk; however, there are limits to the utility of social needs data for improving predictive performance.