300 - Evaluating the Predictive Utility of Hemoglobin A1c for Diabetic Ketoacidosis

Publication Number: 300.4717

Adam Cross, University of Illinois College of Medicine, Peoria, IL, United States; Maciej Wieczorek, University of Illinois at Urbana-Champaign, College of Engineering, Toronto, ON, Canada; Mark D. Miller, University of Illinois College of Medicine, Dunlap, IL, United States; Valeria C. Benavides, Children's Hospital of Illinois, Peoria, IL, United States; Jimeng Sun, University of Illinois College of Medicine, Las Vegas, NV, United States

Background: Hemoglobin A1c (HbA1c) is a widely used biomarker to stratify patients with type 1 diabetes mellitus (T1DM) into high-risk or low-risk categories for diabetic ketoacidosis (DKA). However, the effectiveness of HbA1c in predicting DKA risk remains uncertain.

Objective: This study aims to evaluate the predictive performance of HbA1c ≥9% in identifying patients at risk for DKA across different racial and ethnic groups and to determine the need for more sensitive diagnostic methods.

Design/Methods: A retrospective analysis was conducted on a cohort of 38,270 HbA1c tests among 2734 patients with T1DM. A confusion matrix was used to assess the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of HbA1c ≥9% for DKA in the next 3 months. The analysis was further stratified by race and ethnicity.

Results: • Overall Population (N=38,270): Sensitivity 83.0% (95% CI: 80.7-85.1), specificity 61.4% (95% CI: 60.9-61.9), PPV 6.1% (95% CI: 5.7-6.5), NPV 99.2% (95% CI: 99.0-99.3).
• White or Caucasian (N=30,921): Sensitivity 81.8% (95% CI: 79.1-84.3), specificity 64.2% (95% CI: 63.6-64.7), PPV 6.0% (95% CI: 5.6-6.5), NPV 99.2% (95% CI: 99.1-99.3).
• Black or African American (N=4,371): Sensitivity 90.6% (95% CI: 85.9-93.8), specificity 42.9% (95% CI: 41.4-44.4), PPV 7.5% (95% CI: 6.6-8.6), NPV 98.9% (95% CI: 98.3-99.3).
• Asian (N=199): Sensitivity 57.1% (95% CI: 25.0-84.2), specificity 76.0% (95% CI: 69.5-81.5), PPV 8.0% (95% CI: 3.2-18.8), NPV 98.0% (95% CI: 94.2-99.3).
• Not Hispanic, Latino/a, or Spanish origin (N=35,603): Sensitivity 83.5% (95% CI: 81.1-85.6), specificity 61.7% (95% CI: 61.1-62.2), PPV 6.3% (95% CI: 5.9-6.7), NPV 99.2% (95% CI: 99.1-99.3).
• Hispanic or Latino (N=2,132): Sensitivity 74.1% (95% CI: 61.6-83.7), specificity 55.9% (95% CI: 53.7-58.0), PPV 4.5% (95% CI: 3.3-6.0), NPV 98.7% (95% CI: 97.9-99.2).

Conclusion(s): While HbA1c ≥9% shows high sensitivity for identifying patients at risk for DKA, its specificity is moderate. This indicates that while many patients at risk for DKA are identified, there is also a high number of false positives. The performance varies across racial and ethnic groups, with Black/African American patients showing higher sensitivity but lower specificity compared to White/Caucasian patients. Conversely, Hispanic or Latino patients exhibit lower sensitivity but higher specificity. While NPV is high for all groups, PPV is very poor at 6.1%. This demonstrates a critical need for more accurate methods to identify individuals at risk for DKA in the T1DM population, considering racial and ethnic disparities in diagnostic performance.

DKA Rates by Race and Hemoglobin A1c Range
The rates of diabetic ketoacidosis (DKA) by race, stratified by hemoglobin A1c.

DKA Rates by Ethnicity and Hemoglobin A1c Range
The rates of diabetic ketoacidosis (DKA) by ethnicity, stratified by hemoglobin A1c.

ROC Curve with A1c Thresholds
The Receiver Operator Curve for the A1c test. Note the optimal cutoff of 9.5% based on maximum Youden's J Point of 0.46 (0.43-0.49), as compared to 0.44 (0.42-0.47) for the current standard cutoff of 9.0%. However, the difference is statistically insignificant.