Identifying diabetic ketoacidosis encounters at diagnosis among children with type 1 diabetes in claims data.

Abstract

Background: Quantify the performance of claims data in identifying diabetic ketoacidosis (DKA) cases and estimate healthcare resource utilization among newly diagnosed type 1 diabetes patients with and without DKA.

Methods: We matched individuals <18 years of age from the Barbara Davis Center for Diabetes Registry to the Colorado all-payer claims database to estimate variation in the performance of claims (ie, type 1 diabetes and DKA codes in emergency department [ER] and inpatient [IP] settings) in identifying DKA cases from 2014 to 2019. We estimated the unadjusted and adjusted associations of laboratory-confirmed DKA with resource utilization using encounters and incidence rate ratios with 95% confidence intervals.

Results: After applying algorithms for identifying type 1 diabetes with continuous enrollment criteria, n = 728 (62% match of the sample) represented an insured population residing in Colorado. Mean age at onset was approximately 9.48 years, with a proportion of DKA events at diagnosis of 56%. Strict coding definitions of DKA (DKA and type 1 diabetes) had lower sensitivity (63%) and higher specificity (76%) than looser definitions (type 1 diabetes or DKA), which had higher sensitivity (89%) and lower specificity (60%). Laboratory-confirmed DKA was associated with over 5 times the rate of IP encounters compared with no confirmed DKA (incidence rate ratio [IRR], 5.49; 95% confidence interval, 4.03-7.47) but fewer ER-only encounters (IRR, 0.36; 95% confidence interval, 0.24-0.54).

Conclusion: Claims data may not capture all DKA encounters at type 1 diabetes diagnosis. Future research should provide sensitivity analyses across coding and settings to estimate the burden of DKA.