Algorithms to define diabetes type using data from administrative databases: A systematic review of the evidence

Area of Science:

• Health Informatics
• Diabetes Research
• Algorithm Development

Background:

• Distinguishing between type 1 and type 2 diabetes is crucial for effective patient management.
• Administrative data offers a valuable resource for diabetes classification but requires robust algorithms.

Purpose of the Study:

• To identify the most effective algorithms for differentiating type 1 and type 2 diabetes within administrative datasets.
• To evaluate the performance of various algorithmic approaches using diagnostic metrics.

Main Methods:

• A systematic literature search was conducted in Embase and MEDLINE (January 2000 - January 2023).
• Studies reporting diagnostic metrics for diabetes type classification algorithms against reference standards were included.
• The Quality Assessment of Diagnostic Accuracy Studies tool was used to evaluate study quality.

Main Results:

• Nineteen of 24 included studies showed low risk of bias and applicability concerns.
• Algorithms using multiple diabetes diagnostic codes achieved high sensitivity and specificity (>92.1% for type 1, >86.9% for type 2).
• Insulin use criteria were more sensitive and specific than oral hypoglycemic agent criteria for diabetes type detection.

Conclusions:

• Algorithms incorporating multiple diabetes diagnostic codes and insulin use are optimal for distinguishing diabetes types in administrative data.
• Employing multiple criteria within algorithms enhances the accuracy and sensitivity of diabetes type classification.