Detection of Undiagnosed Diabetes and Prediabetic States in High-risk Emergency Department Patients

Background: Diabetes is often not diagnosed until complications appear, and one‐third of those with diabetes may be undiagnosed. Prediabetes and diabetes are conditions in which early detection would be appropriate, because the duration of hyperglycemia is a predictor of adverse outcomes, and there are effective interventions to prevent disease progression and to reduce complications. Objectives: The objectives were to determine the prevalence of diabetes mellitus and prediabetes in emergency department (ED) patients with an elevated random glucose or risk factors for diabetes but without previously diagnosed diabetes and to identify which at‐risk ED patients should be considered for referral for confirmatory diagnostic testing. Methods: This two‐part study was composed of a prospective 2‐year cohort study, and a 1‐week cross‐sectional survey substudy, set in an urban ED in Los Angeles County, California. A convenience sample was enrolled of 528 ED patients without previously diagnosed diabetes with either 1) a random serum glucose ≥ 140 mg/dL regardless of the time of last food intake or a random serum glucose ≥ 126 mg/dL if more than 2 hours since last food intake or 2) at least two predefined diabetes risk factors. Measurements included presence of diabetes risk factors, ED glucose, cortisol, insulin and glycosylated hemoglobin (HbA_1c), and 2‐hour oral glucose tolerance test results, administered at 6‐week follow‐up. Results: Glycemic status was confirmed at follow‐up in 256 (48%) of the 528 patients. Twenty‐seven (11%) were found to have diabetes, 141 (55%) had prediabetes, and 88 (34%) had normal results. Age, ED glucose, HbA_1c, cortisol, and random serum glucose ≥ 140 mg/dL were associated with both diabetes and prediabetes on univariate analysis. A random serum glucose ≥ 126 mg/dL after 2 hours of fasting was associated with diabetes but not prediabetes; ED cortisol, insulin, age ≥ 45 years, race, and calculated body mass index (BMI) were associated with prediabetes but not diabetes. In multivariable models, among factors measurable in the ED, the only independent predictor of diabetes was ED glucose, while ED glucose, age ≥ 45 years, and symptoms of polyuria and polydipsia were independent predictors of prediabetes. All at‐risk subjects with a random ED blood glucose > 155 mg/dL had either prediabetes or diabetes on follow‐up testing. Conclusions: A substantial fraction of this urban ED study population was at risk for undiagnosed diabetes and prediabetes, and among the at‐risk patients referred for follow‐up, the majority demonstrated diabetes or prediabetes. Notably, all patients with two risk factors and a random serum glucose > 155 mg/dL were later diagnosed with prediabetes or diabetes. Consideration should be given to referring ED patients with risk factors and a random glucose > 155 mg/dL for follow‐up testing.