1035-P: HbA1c Does Not Identify Nursing Home Residents at High Risk of Hypoglycemia and Hyperglycemia

Introduction: Preventing hypo- and hyper-glycemia are primary goals of diabetes management for NH residents with type 2 diabetes. Our objective was to compare discrimination and test characteristics of HbA1c and fingerstick-glucose derived metrics to predict subsequent hypoglycemia <54mg/dL and hyperglycemia >400mg/dL.

Methods: We used a national Veterans Affairs database to identify NH residents aged ≥65 years with type 2 diabetes and insulin use with a NH stay of ≥45 days from 2013-2018. To derive fingerstick-glucose based metrics, we included 5,346 NH residents with ≥3 fingersticks daily for ≥14 days during days 15-45 of their NH stay. The low blood glucose index (LBGI) and high blood glucose index (HBGI) are fingerstick-based metrics that summarize the frequency and severity of low and high glucoses over 1 month, calculated from Kovatchev’s formula.

Results: The mean age was 73.7 years, 98% were male, and mean HbA1c was 7.5%. During days 46-90 of the NH stay, 804 (15%) had hypoglycemia <54mg/dL and 1,254 (24%) had hyperglycemia >400mg/dL. HbA1c <7% performed poorly as predictor of hypoglycemia (Table). Fingerstick-based metrics performed well, with c-statistics around 0.7.

Conclusion: Our results suggest that fingerstick-glucose based metrics and not HbA1c should be used to identify NH residents at high risk of hypo- and hyper-glycemia and to drive glycemic treatment decisions in these patients.

View largeDownload slide

View largeDownload slide