Testing an Automated Approach to Identify Variation in Outcomes among Children with Type 1 Diabetes across Multiple...

Area of Science:

• Pediatric Endocrinology
• Health Services Research
• Data Science in Healthcare

Background:

• Comparative quality assessment for type 1 diabetes (T1D) in children requires efficient data collection methods.
• PCORnet's Common Data Model (CDM) standardizes electronic health record (EHR)-derived data across multiple institutions.

Purpose of the Study:

• To determine the feasibility of using automated EHR data for comparative quality assessment in pediatric T1D.
• To evaluate the reliability and validity of EHR data for measuring T1D care and outcomes.

Main Methods:

• Assessed glycemic control, diabetic ketoacidosis (DKA) admissions, and clinic visits from 2016-2018 in two PCORnet networks (PEDSnet, OneFlorida).
• Developed EHR-based measure specifications and identified institution-specific rates using CDM data.
• Validated findings through agreement assessment with manual chart review.

Main Results:

• Analyzed data from 9,740 youth with T1D across 12 institutions.
• Found 26% had HbA1c >9% annually; median HbA1c was 8.5%.
• 4% were hospitalized for DKA; PCORnet CDM predictive value exceeded 75% for all measures.

Conclusions:

• Automated EHR data is a valid, efficient, and reliable tool for assessing T1D care quality.
• Observed significant institutional variations in care, with many institutions not meeting American Diabetes Association HbA1c goals (<7.5%).