OBJECTIVES: To identify claims and electronic health records (EHR) data that can be used to build a predictive model for need of palliative care in patients with chronic diseases in the US.

METHODS: A retrospective study using the Optum® de-identified Market Clarity Dataset (linked claims and EHR of patients) was done among adult (>=18 years) patients with >=1 claims and/ or EHR with a diagnosis or procedure code for Palliative care during 1^st Jan 2019 to 31^st Mar 2022. Only patients requiring Palliative care for the following 8 chronic diseases were included – hypertension, neoplasms, diabetes mellitus, depression, anxiety, chronic ischemic heart disease, COPD, and chronic kidney disease. Index date was defined as the first claim/ EHR with Palliative care code. Only patients with no Palliative care code in claims or EHR during preceding 6 months from index date were included. A control group of adult patients that didn’t require Palliative care but were matched on the 8 chronic diseases was identified (1:4 greedy match). Palliative care predictive model was built using demographic and claims data of cases and controls:

1. Potential demographic predictors: age, biological sex, race/ ethnicity
2. Potential claims predictors: in preceding 6 months, history of inpatient hospitalization or ICU/ mechanical ventilation, number of healthcare interactions, other comorbidities, procedures, and medications received

RESULTS: Total 96,378 cases and 417,170 controls were included. Significantly higher proportion of patients aged >=50 years, males, and African Americans required Palliative care. Further, significantly higher proportion of patients that had inpatient hospitalization or ICU/ mechanical ventilation in preceding 6 months required Palliative care. We are currently identifying potential EHR predictors like clinical features, disability scores, and biomarkers that can make the model more accurate and robust.

CONCLUSIONS: Claims and clinical data can be effectively used to predict need for Palliative care in patients.