Applying Sequential Analytic Methods to Self-Reported Information to Anticipate Care Needs.

Area of Science:

• Health Services Research
• Health Informatics
• Predictive Modeling

Background:

• Identifying care needs for newly insured individuals is crucial under the Affordable Care Act.
• Systematically collected patient-reported information can proactively identify specific care needs before service utilization.

Purpose of the Study:

• To develop a predictive model using patient-reported data to identify high-cost care subgroups among newly enrolled individuals.
• To apply cluster analysis to patient data to define actionable subpopulations with distinct care requirements.

Main Methods:

• Retrospective cohort study of 6,047 individuals enrolled via the Colorado State Individual Exchange.
• Utilized Brief Health Questionnaire (BHQ) responses to predict top 25% cost and cluster analysis to identify subpopulations.
• Measured per-member, per-month cost 6-12 months post-enrollment.

Main Results:

• BHQ responses, including self-reported health, functional limitations, medication use, chronic conditions, prior ED use, and insurance history, predicted high costs.
• Predictive model showed a wide range of high-cost probabilities (3.5% to 96.4%).
• Identified patient clusters such as those with high morbidity and prior utilization, previously uninsured with few financial constraints, and previously insured with medication needs.

Conclusions:

• Sequential predictive modeling and cluster analysis of patient-reported data effectively identify subgroups with specific needs.
• This approach enables targeted interventions for heterogeneous populations to optimize initial care delivery.
• Proactive identification of care needs can lead to more efficient and effective healthcare resource allocation.