Matched cohorts applied propensity scores inverse probability of weighting - the association between home-based primary care in dementia patients and their hospital utilization

• Activity Number: 63 - Statistical Methods in Precision and Personalized Medicine and Subgroup Analysis
• Type: Contributed
• Date/Time: Sunday, August 7, 2022 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): In-Lu Amy Liu* and Janet S Lee and Ernest Shen and Huong Q Nguyen
• Companies: Kaiser Permanente and Kaiser Permanente and Kaiser Permanente and Kaiser Permanente
• Keywords: Propensity scores inverse probability of treatment weighting ; Fine and Gray’s proportional hazards model ; Strata-matched comparison ; Data simulation; Matched cohorts; Stratified proportional hazards model

Abstract

This study included a retrospective observational cohort of persons living with dementia (PLWD) receiving home-based primary care (HBPC; n=287) from 2015-2020 and a strata-matched comparison PLWD (n=861) who never received HBPC. Three important factors (diagnosis year, age, gender) were selected as matching strata. Within each matching cell, three comparison patients were randomly selected and assigned anchor dates based on the corresponding HBPC patient’s first dementia diagnosis and HBPC admission date using data simulation. Propensity scores inverse probability of treatment weighting (IPTW PS) were generated from baseline socio-demographic and clinical data of 12 months prior to HBPC admission or anchor dates. We included IPTW PS in Fine and Gray’s proportional hazards model with death as a competing risk and adding matching ID as strata variable to assess time-to-first hospital-based utilization following HBPC admission through December 31, 2020. We found that HBPC for PLWD was associated with a similar risk of hospital utilization [HR: 1.06 (95% CI: 0.89–1.26), p = 0.51] compared to a matched non-HBPC comparison group after a median follow-up of 199 days (IQR: 81-432).