Development and Validation of a Primary Care Tool to Identify Patients With Type 2 Diabetes Mellitus at High Risk of Hypoglycemia-Related Inpatient Admissions

Abstract

Background: Hypoglycemia inpatient admissions are costly and potentially preventable. Using established risk factors for hypoglycemia, we set out to develop a risk-scoring tool using the data from an Asian population.

Methods: In this historical cohort study, we extracted the data of 47,404 type 2 diabetes mellitus (T2DM) patients with complete data based on their last visit in 2012 at selected National Healthcare Group Polyclinics in Singapore. The outcome variable is the occurrence of any hypoglycemia inpatient admission within 6 months from their last visit in 2012. We entered the following potential predictors into a logistic regression model: 1) Age; 2) Largest reduction in glycated hemoglobin within 1 year; 3) Last body mass index; 4) Last estimated glomerular filtration rate; 5) Usage of sulphonylurea and/or insulin; 6) Last glycated hemoglobin; 7) Any previous hypoglycemia inpatient admission in the past 1 year. The relative weightage of predictors were compared, and the model parameters were subsequently converted to a simple risk score (range: 0 to 100).

Results: We found predictors 1 to 5 to be statistically significant for subsequent hypoglycemia inpatient admission. In our study population, based on a sensitivity of 73.8% and a specificity of 73.1%, a cut-off score of 38 was selected. The area under the receiver-operating characteristic curve was 0.809 (CI: 0.763 - 0.855).

Conclusions: A risk score using commonly available clinical data can help to identify those at risk of hypoglycemia inpatient admission with satisfactory level of accuracy. This score needs to be further validated with randomized controlled studies.




J Endocrinol Metab. 2019;9(3):43-50
doi: https://doi.org/10.14740/jem563