AUTHOR=Yun Jae-Seung , Han Kyungdo , Choi Soo-Yeon , Cha Seon-Ah , Ahn Yu-Bae , Ko Seung-Hyun TITLE=External validation and clinical application of the predictive model for severe hypoglycemia JOURNAL=Frontiers in Endocrinology VOLUME=13 YEAR=2022 URL=https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2022.1006470 DOI=10.3389/fendo.2022.1006470 ISSN=1664-2392 ABSTRACT=Objective

An internally validated, one-year risk prediction model for severe hypoglycemia (SH) in type 2 diabetes was evaluated in a general hospital setting to externally verify and validate its performance.

Research design and methods

Between December 2017 to December 2019, 2,645 adult patients with type 2 diabetes who visited the diabetes center were enrolled. The receiver operating characteristics curve and Harrell C-statistics were compared to identify the discrimination of the model. The predicted and actual incidence of SH for one year in the development and validation cohorts were compared by ranking participants by deciles of predicted risk.

Results

The concordance index was 0.878 in the external validation cohort. The sensitivity and specificity of the predictive model were 0.833 and 0.847, respectively. Based on the predicted risk, we stratified the groups into four categories: low (<0.05%), intermediate (0.05% to <0.5%), high (0.5% to <2.0%), and very high-risk group (≥2.0%). The actual annual incidence of SH gradually increased with the increased risk score level for the decile group (P for trend <0.001). The actual annual SH incidence significantly increased with increase in SH risk scores, which proportionately increased with age, duration of diabetes, glycated hemoglobin, and albuminuria and decreased with body mass index, renal function (p for trends <0.001 for all) in type 2 diabetes.

Conclusion

On external validation, the novel one-year SH prediction model showed excellent discrimination in participants with type 2 diabetes and can effectively screen high-risk patients for SH, even in the general hospital setting.