Diabetes Screening in the Emergency Department: Development of a Predictive Model for Elevated Hemoglobin A1c

Area of Science:

• Diabetes Mellitus Research
• Predictive Modeling in Healthcare
• Emergency Medicine

Background:

• Elevated Hemoglobin A1c (HbA1c) indicates poor glycemic control, a risk factor for diabetes.
• Identifying individuals with elevated HbA1c in the emergency department (ED) is crucial for timely intervention.
• Predictive models can aid in stratifying risk and guiding follow-up care for at-risk populations.

Purpose of the Study:

• To develop and validate a prediction model for elevated HbA1c (≥5.7%) in emergency department patients screened for diabetes.
• To identify key demographic, clinical, and historical factors associated with elevated HbA1c.
• To inform follow-up care strategies for patients identified with prediabetes or diabetes in the ED setting.

Main Methods:

• Retrospective analysis of electronic health records from 590 patients screened for diabetes in the ED.
• Development of a prediction model using a derivation set (80%) with backward elimination and cross-validation.
• Validation of the model on a separate test set (20%) with performance evaluation using C-statistic, sensitivity, specificity, and Brier score.

Main Results:

• The final prediction model incorporated nine variables: age, race/ethnicity, insurance, chief complaints (back pain, fever/chills), and medical history (obesity, hyperlipidemia, COPD, substance misuse).
• The model demonstrated acceptable performance with a C-statistic of 0.75 and predictive values exceeding 0.70.
• The Hosmer-Lemeshow test (0.29) and Brier score (0.21) indicated no significant model ill fit and moderate predictive accuracy.

Conclusions:

• Age, obesity, hyperlipidemia, and importantly, a history of substance misuse, are significant predictors of elevated HbA1c in ED patients.
• Substance misuse should be considered when planning follow-up care for individuals identified with prediabetes or diabetes in the ED.
• Further research, including external validation in larger ED populations, is warranted to confirm these findings.