A Risk Stratification Tool for a Serious Illness Population to Support Specialty Palliative Care Services Planning and Outreach (GP742)

A Risk Stratification Tool for a Serious Illness Population to Support Specialty Palliative Care Services Planning and Outreach (GP742)

Outcomes. 1. Understand the development and validation of a risk stratification tool for a population of seriously ill patients that combines the strengths of existing models 2. Understand why it is important to define and start with a seriously ill population definition and use a population-based a...

Full description

Saved in:
Bibliographic Details
Published in:Journal of pain and symptom management Vol. 63; no. 6; p. 1135
Main Authors: Wang, Susan, Viveros, Lori, Hong, Benjamin, Padilla, Ariadna, Nguyen, Huong, Schuller, Sean, Park, Seryoung, Ryan, Suzanne, Vargas, Sue Del Carpio Bellido, Nau, Claudia
Format: Journal Article
Language:English
Published: Madison Elsevier Limited 01-06-2022
Subjects:
Candidates
Computerized medical records
Coronaviruses
COVID-19
End of life decisions
Health records
High risk
Hospitalization
Literature reviews
Medical referrals
Medical screening
Mortality
Outreach programs
Palliative care
Pandemics
Pharmacy
Population
Prediction models
Stratification
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Outcomes. 1. Understand the development and validation of a risk stratification tool for a population of seriously ill patients that combines the strengths of existing models 2. Understand why it is important to define and start with a seriously ill population definition and use a population-based approach to developing a risk stratification tool Importance. Reliable prognostication for seriously ill patients can support specialty palliative care (SPC) service planning and systematic referrals to SPC. Objective(s). Develop a predictive tool to support SPC planning and outreach by combining the strengths of existing mortality prediction models and integrating COVID-19 diagnoses into risk stratification. Method(s). We used data from the Kaiser Permanente Southern California electronic health records of 185,240 patients diagnosed with a serious illness (SI) in 2018. Patients with SI were identified from an adapted list of ICD codes provided by the Center to Advance Palliative Care. Predictors included variables from two well-known models, the EPIC-End-of-Life-Care risk score and the PMR3 risk score. We further added predictors identified from literature review and physician input. The 300+ candidate variables included data on diagnoses, utilization, laboratory values, durable medical equipment, and pharmacy orders. We used group LASSO with logistic regression to predict the risk of dying in 2019. We split the sample into 80% training and 20% test data. Results. Sample mortality was 7%. The area under the receiving operating curve of the final model was 0.85. We identified a high-risk population of about 34,000 patients with SI that includes 65% of all deaths. This high-risk group has a mortality risk of 25% (number needed to screen [NNS] = 4). This population can be further divided into a very high-risk group of 8,660 patients with a mortality risk of 40% (NNS = 2.5) to prioritize screening. The low-risk group includes 150,920 of the 185,240 patients with SI and has a mortality risk of 3%. The 2019 model performs comparably in 2020. We will add a COVID-19 indicator for patients with a COVID-19 hospitalization and recalculate sensitivity and NNS. Conclusion and Impact. We combined the strengths of existing models to build a high-performing risk stratification model to support planning and outreach for SPC during the COVID-19 pandemic.
ISSN:0885-3924
DOI:10.1016/j.jpainsymman.2022.04.133