UCSF

Reducing patient readmissions within 30 days of an initial hospitalization is fiscally responsible and important to patient-centered care. Home health (HH) agencies, have been publicly reported by Medicare on 30-day readmission to the hospital from HH since 2015, and are proposed to be measured on potentially preventable 30-day post-discharge readmissions in 2019. The body of 30-day readmissions evidence inclusive of HH is small and inconsistent, comprised in part from national claims data or federal-level reports. We cannot assume that while the hospital and HH settings share the same patients that the risks, determinants of readmission and preventative interventions are also the same, nor that Medicare-specific data reports the full story. This study was performed as a dissertation, and retrospective, secondary analysis of electronic health record data, aimed at determining HH operations predictors influencing readmission to the hospital during HH services in one northern California HH agency. Logistic regression was used to determine univariate and multiple predictors of two readmission models: 30-day readmission to the hospital during HH services and readmission to the hospital at any time during HH services. Univariate results with 30-day readmission produced 8 significant predictors which were then entered into the 30-day readmission multiple logistic regression (MLR) model simultaneously (omnibus test chi-square 32.058, p=0.000). Two variables continued to demonstrate unique contribution to the model: frontloaded contacts rate in the first week (AOR = 0.970, p = 0.005) and number of high-risk medications (AOR = 1.638, p = 0.027). Univariate results with readmission at anytime produced six significant predictors which were then entered simultaneously to the readmission at anytime MLR model (omnibus test chi-square 29.565, p=0.000). Two variables continued to demonstrate unique contribution to the model: homebound status by medical contraindication (AOR = 5.058, p = 0.011) and HH total length of stay (days) (AOR=1.034, p=0.033). Both the 30-day and anytime readmission models described a unique combination of significant predictors of readmission to the hospital during HH services. This study also defined a manner of calculating frontloaded contacts in the first week of care, which contributed to the 30-day readmission model, bearing operational interest to HH settings.