Lead/Presenter: Matthew Maciejewski,
COIN - Durham
All Authors: Maciejewski ML (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC), Zulman D (Center for Innovation to Implementation, Palo Alto, CA), Grubber J (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC) Weidenbacher HJ (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC) Blalock DV (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC) Zullig LL (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC) Whitson HE (Geriatrics Research Education and Clinical Center, Durham Veterans Affairs Health Care System, Durham, NC) Hastings SN (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC) Smith VA (Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC)
Objectives:
Despite recognition of the influential role of social and behavioral determinants of health (SDH) on patient outcomes, little is known regarding the predictive value of SDH in explaining variation in outcomes for high-risk patients. We examined whether patient-reported SDH measures improves prediction of 90-day hospital admission in high-risk Veterans over a model including only electronic health record (EHR) variables.
Methods:
In a prospective cohort study, we used a modified Dillman method to mail surveys to a nationally representative sample of 10,000 high-risk VA primary care patients with 1-year risk of hospitalization or death > = the 75th percentile based on the VA's Care Assessment Need (CAN) score. The survey included patient-reported measures of SDH (e.g., recent life stressors, self-determination, resilience, activities of daily living (ADLs), social support, loneliness). Characteristics of survey respondents and non-respondents were compared with frequencies, means, and standardized mean differences. Three logistic regression models predicting all-cause 90-day hospital admission were conducted on survey respondents: a basic model of EHR covariates related to demographics and comorbidity alone, a full specification including EHR and all survey-based covariates, and a model including all EHR variables as well as the subset of survey variables identified in a stepwise-selected best model of survey-based covariates based on AIC.
Results:
4,685 of 10,000 high-risk Veterans responded. Respondents were more likely to be older, married, white, live in a rural locale, and less likely to have been diagnosed with a mental illness than non-responders. At 90 days, 5.6% of survey respondents had VA hospital admissions. The logistic regression with EHR-based + stepwise selected survey-based covariates (ADLs, resilience, general health, and life stressors) was more predictive of 90-day hospital admission than models with only EHR covariates or a fully-specified model (AIC = 1944 vs. 1981 vs. 1979, respectively).
Implications:
The addition of patient-reported SDH and functional status measures not available in EHR data improved prediction of 90-day hospital admission in high-risk Veterans.
Impacts:
Integration of select SDH measures into clinical assessments and EHRs may help identify individuals at high-risk for hospitalization who would benefit from medical and social service intervention.