High Cost, High Need Populations Data‐Driven Decision Making for Interventions with Complex Needs’ Patients: Identifying the Right Target Population

Abstract

Research ObjectiveWhile existing taxonomies describe the major groups of complex patients (prominently, the National Academy of Medicine [NAM] taxonomy), these are often at a conceptual level. Healthcare organizations often select patients for CN interventions using structured criteria based on administrative data, and program designers must make critical decisions in defining target populations. Our study aimed to provide insights to inform decisions about selecting CN populations for interventions, by conducting a review of approaches being used to operationalize CN populations in practice, and implementing various definitions in a large health plan member population to compare the characteristics of the resulting patient cohorts.Study DesignOur study involved multiple methods to identify and implement data‐driven definitions of CN patients. We interviewed internal and external subject matter experts to inform our research approach. We conducted a pragmatic literature review of studies published 2000‐2018 which used data‐driven definitions of CN populations. We conducted thematic analysis of the results to identify core concepts and variations in criteria used to define CN populations. We implemented variations of the core concepts and calculated descriptive statistics to understand the characteristics of patients within each of the resulting CN populations, including size of the population, demographics, chronic illness, health care utilization, spending, and survival.Population StudiedPatients with complex needs defined using criteria applied to the membership population of Kaiser Permanente Southern California (KPSC), a large, population‐based health plan with more than 4 million members.Principal FindingsOur literature review yielded 90 examples of CN populations; we cataloged population criteria for each. We grouped observed criteria into domains: age (reflected in 59 CN population definitions), income (12), costs (45), utilization (39), health status (35), and subjective criteria (15). We constructed a framework for classifying approaches to selecting CN populations (Figure 1). We then used KPSC data to assemble populations representing five CN patient concepts (high cost, high emergency department utilization, high inpatient utilization, multiple chronic conditions, and frail elderly); in each concept, we compared multiple population specifications. We observed within‐ and across‐concept differences in the characteristics of the resulting populations. Data presented during the session will include CN population size, demographics, utilization, comorbidity, costs, and survival.ConclusionsOur results illustrate tremendous variation in how CN populations are defined. Nearly every CN population identified through our literature review was unique in the criteria used to select patients. Furthermore, when we implemented a variety of definitional approaches in our membership, we found substantial variation in the cohorts that resulted from different approaches to patient selection. These findings have important implications for both research and health care operations. In a world of limited resources, it is imperative that we use available data to identify, understand, and target resources to the right CN populations.Implications for Policy or PracticeThese findings offer important considerations for intervention planning and implementation. Program planners should select population criteria after exploring the impact of their draft criteria and comparing alternate approaches to understand the implications of those decisions.Primary Funding SourceThe Garfield Memorial Fund.