Predicting avoidable hospital events in Maryland.

Area of Science:

• Health Services Research
• Health Informatics
• Predictive Analytics

Background:

• Avoidable hospital events pose a significant burden on healthcare systems and beneficiaries.
• Accurate prediction of these events is crucial for targeted interventions and resource allocation.

Purpose of the Study:

• To develop and validate a prediction model for avoidable hospital events among Medicare fee-for-service beneficiaries in Maryland.
• To assess the predictive accuracy of the developed model using administrative claims data.

Main Methods:

• Utilized Medicare fee-for-service claims data from Maryland (2017-2020) and ZIP code-level data.
• Employed multivariable logistic regression, incorporating 198 individual and ZIP code-level risk factors across six domains.
• Performed feature selection to identify 73 statistically significant risk factors for the primary model.

Main Results:

• The developed risk prediction model identified individuals at high risk for future avoidable hospital events.
• The top 10% of individuals identified by the model accounted for 48.7% of avoidable hospital events in the subsequent month.
• The model demonstrated superior predictive power compared to existing Centers for Medicare & Medicaid Services hierarchical condition category risk scores.

Conclusions:

• A risk prediction model utilizing standard administrative claims data can accurately identify individuals likely to experience future avoidable hospital events.
• The findings support the use of such models for proactive healthcare management and intervention strategies.
• The model's performance indicates its potential for improving patient outcomes and reducing healthcare costs.