Development and validation of a model for predicting inpatient hospitalization

Area of Science:

• Health Services Research
• Predictive Analytics
• Public Health

Background:

• Hospitalizations represent a significant financial burden on healthcare systems and society.
• Effective prediction of acute care hospitalizations is crucial for resource management and cost reduction.

Purpose of the Study:

• To develop and validate a predictive model for acute care hospitalizations.
• The model aims to cover all age groups using administrative claims data.

Main Methods:

• A retrospective cohort study utilized a US health plan claims database.
• The Adjusted Clinical Group (ACG) system and multivariate logistic regression were employed to identify risk factors.
• Model validation used a separate dataset of 4.7 million individuals.

Main Results:

• The developed model demonstrated good performance with an Area Under the Curve (AUC) of 0.80.
• It outperformed previous hospitalization prediction models (AUC=0.75 and AUC=0.78).
• Key predictors included advanced age, prior hospitalizations, emergency room visits, and specific high-impact diseases.

Conclusions:

• A validated, population-based predictive model for hospitalization risk has been developed.
• This model can effectively estimate an individual's future hospitalization risk.
• The tool holds potential utility for health plans and care managers in proactive patient management.