Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

Headline

Creating clinically meaningful profiles of patients with medically complex needs can support the delivery of more tailored interventions.

Context

Patients with complex health and social needs are often identified through prior year costs or utilization, number and type of comorbid conditions, and/or future predictions of hospitalization or costs. These methods, however, do not provide guidance that can be used to tailor care for patient subgroups with unique needs. This study used electronic health record data from Kaiser Permanente Northern California to define clinically meaningful and heterogenous patient profiles within the overarching group of medically complex patients.

Findings

Among a population of 104,869 adults that met the definition of complexity (high comorbidity and high health care utilization), the analysis identified seven patient subgroups:

1. Individuals with highest utilization and most comorbid conditions;

2. Older adults with cardiovascular disease-related conditions and complications;

3. Frail older adults;

4. Individuals with chronic pain management needs;

5. Individuals with active cancer treatment;

6. Individuals with severe mental illness complicated by social factors; and

7. Less clinically engaged individuals with prevalent comorbidities but fewer visits.

These subgroups were based on the variation in prevalence of 17 different key patient variables including hospital admissions, opioid analgesic use, and mental health visits. Subgroups had significant differences in outcomes for mortality, hospitalization, hospice admission, and mental health visits. Given this heterogeneity, subgroups will benefit from different care models and care management strategies. The study also describes key features of care management strategies suggested for each of the seven clinical profiles identified.

Takeaways

By defining these distinct subgroups of medically complex patient profiles, this study can help guide population identification strategies and efforts to tailor complex care management interventions. As a single care model may not meet the needs of all adults with high utilization of care, health care systems can develop care management interventions that address shared needs across groups and can also use tailored needs assessments to understand what resources would most benefit specific patient subgroups.