Developing and validating a predictive model for future emergency hospital admissions.

Area of Science:

• Health Services Research
• Predictive Analytics
• Public Health

Background:

• A significant portion of emergency hospital admissions are avoidable, indicating failures in community care systems.
• Avoidable admissions lead to adverse patient outcomes, increased burden on caregivers, and substantial healthcare costs.

Purpose of the Study:

• To develop and validate a predictive model for estimating the individual probability of emergency hospital admission within the next 12 months.
• To create a tool that can identify high-risk patients for proactive community-based management.

Main Methods:

• Utilized routinely collected secondary care data linked with population-level data for a cohort of 190,466 individuals.
• Developed a logistic regression model incorporating five independent variables to predict emergency admission risk.
• Validated the model's predictive accuracy using the area under the receiver operating characteristic curve (AUC) and probability cut-off analysis.

Main Results:

• The developed risk prediction model demonstrated high accuracy, with an AUC of 0.9384 (95% CI 0.9325-0.9443).
• The model's overall prediction accuracy exceeded 95% across various probability thresholds.
• Internal validation confirmed the model's robust performance in predicting individual emergency admission risk.

Conclusions:

• The internally validated model accurately predicts the individual risk of emergency hospital admission within one year.
• The model's simplicity facilitates integration into decision support tools for community-based patient management.
• Implementation of this tool can aid in proactively managing at-risk patients, potentially reducing avoidable hospitalizations and associated costs.