# Clinical Validation of Type 1 Diabetes Coding in Hospital Discharge Records Using ADA Criteria: Implications for Spanish and European Health Data Spaces

## TL;DR

This study validated the accuracy of Type 1 diabetes diagnoses in a Spanish hospital database, finding high reliability in younger patients but lower accuracy in older adults.

## Contribution

The study provides clinical validation of T1DM coding in the CMBD using ADA criteria, highlighting age-related diagnostic accuracy.

## Key findings

- T1DM coding in the CMBD showed 100% sensitivity and NPV, with 80.2% specificity.
- Specificity dropped to 60% in patients aged 80 and older due to misclassification with type 2 diabetes.
- Interobserver agreement was excellent (κ = 0.869), indicating high reliability in diagnosis confirmation.

## Abstract

Background/Objectives: Administrative and clinical databases are increasingly used for research, but their value depends on coding accuracy. The Spanish National Hospital Discharge Database (CMBD) is a standardised registry widely applied in epidemiology. Type 1 diabetes mellitus (T1DM) is an autoimmune disease with early onset and long-term complications. This study aimed to validate the accuracy of T1DM diagnoses recorded in the CMBD. Methods: A cross-sectional validation study was conducted at Hospital Clínico San Carlos (Madrid, Spain) including discharges from 2016–2023. Two age- and sex-matched samples of 384 admissions each (with and without T1DM coding, ICD-10 E10) were randomly selected. The gold standard was the confirmation of T1DM based on the diagnostic criteria established by the 2016 American Diabetes Association (ADA) consensus, which remained valid through 2025, verified by a detailed review of electronic health records (EHRs). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated with 95% confidence intervals (CIs), and interobserver concordance was assessed with Cohen’s kappa. Results: Of the 245,206 discharges, 1324 (0.54%) included a T1DM diagnosis. Validation showed a sensitivity of 100% (95% CI: 98.7–100), specificity of 80.2% (95% CI: 76.4–83.5), PPV of 75.3% (95% CI: 70.7–79.3), and NPV of 100% (95% CI: 99.0–100). Interobserver agreement was excellent (κ = 0.869). Specificity declined with age, from 100% in patients < 30 years to 60% in those ≥ 80 years, mainly due to misclassification with insulin-treated type 2 diabetes. Conclusions: T1DM diagnoses in the CMBD show very high validity and reliability in younger patients, supporting their use in epidemiological and clinical research, while complementary verification is advisable in older adults.