Outcomes Research in Review

EMR-Based Tool for Identifying Type 2 Diabetic Patients at High Risk for Hypoglycemia

Karter AJ, Warton M, Lipska KJ, et al. Development and validation of a tool to identify patients with type 2 diabetes at high risk of hypoglycemia-related emergency department of hospital. JAMA Intern Med 2017 Aug 21.



Study Overview

Objective. To develop and validate a risk stratification tool to categorize 12-month risk of hypoglycemia-related emergency department (ED) or hospital use among patients with type 2 diabetes (T2D).

Design. Prospective cohort study.

Setting and participants. Patients with T2D from Kaiser Permanente Northern California were identified using electronic medical records (EMR). Patients had to be 21 years of age or older as of the baseline date of 1 January 2014, with continuous health plan membership for 24 months prebaseline and pharmacy benefits for 12 months prebaseline. Of the 233,330 adults identified, 24,719 were excluded for unknown diabetes type, and 3614 were excluded for type 1 diabetes. The remaining 206,435 eligible patients with T2D were randomly split into an 80% derivation sample (n = 165,148) for tool development and 20% internal validation sample (n = 41,287). Using similar eligibility criteria, 2 external validation samples were derived from the Veterans Administration Diabetes Epidemiology Cohort (VA) (n = 1,335,966 adults) as well as from Group Health Cooperative (GH) (n = 14,972).

Main outcome measure. The primary outcome was the occurrence of any hypoglycemia-related ED visit or hospital use during the 12 months postbaseline. A primary diagnosis of hypoglycemia was ascertained using the following International Classification of Diseases, Ninth Revision (ICD-9) codes: 251.0, 251.1, 251.2, 962.3, or 250.8, without concurrent 259.3, 272.7, 681.xx, 686.9x, 707.a-707.9, 709.3, 730.0-730.2, or 731.8 codes [1]. Secondary discharge diagnoses for hypoglycemia were not used because they are often attributable to events that occurred during the ED or hospital encounter.

Main results. Beginning with 156 (122 categorical and 34 continuous) candidate clinical, demographic, and behavioral predictor variables for model development, the final classification tree was based on 6 patient-specific variables: total number of prior episodes of hypoglycemia-related ED or hospital utilization (0, 1–2, ≥ 3 times), number of ED encounters for any reason in the prior 12 months (< 2, ≥ 2 times), insulin use (yes/no), sulfonylurea use (yes/no), presence of severe or end-stage kidney disease (dialysis or chronic kidney disease stage 4 or 5 determined by estimated glomerular filtration rate of ≤ 29 mL/min/1.73 m² (yes/no), and age younger than 77 years (yes/no). This classification tree resulted in 10 mutually exclusive leaf nodes, each yielding an estimated annual risk of hypoglycemia-related utilization, which were categorized as high (> 5%), intermediate (1%–5%), or low (< 1%).

The above classification model was then transcribed into a checklist-style hypoglycemia risk stratification tool by mapping the combination of risk factors to high, intermediate, or low risk of having any hypoglycemia-related utilization in the following 12 months.

Regarding patient characteristics, there were no significant differences in the distribution of the 6 predictors between the Kaiser derivation vs. validation samples, but there were significant differences across external validation samples. For example, the VA sample was predominantly men, with a higher proportion of patients older than 77 years, and had the highest proportion of patients with severe or end-stage kidney disease. Regarding model validation, the tool performed well in both internal validation (C statistic = 0.83) and external validation samples (VA C statistic = 0.81; GH C statistic = 0.79).

Conclusion. This hypoglycemia risk stratification tool categorizes the 12-month risk of hypoglycemia-related utilization in patients with T2D using 6 easily obtained inputs. This tool can facilitate efficient targeting of population management interventions to reduce hypoglycemia risk and improve patient safety.


It is estimated that 25 million people in the United States have diabetes [2]. Hypoglycemia is a frequent adverse event in patients with T2D, being more common than acute hyperglycemic emergencies such as hyperosmolar hyperglycemic state [3]. Iatrogenic hypoglycemia due to glucose-lowering medication can result in hypoglycemic crisis that requires administration of carbohydrates, glucagon, or other resuscitative actions in the ED or in hospital [4,5]. The estimated total annual direct medical costs of hypoglycemia-related utilization were estimated at approximately $1.8 billion in the United States in 2009.

The risk of hypoglycemia varies widely in patients with T2D and there are no validated methods to target interventions to the at-risk population. In this article, Karter and colleagues developed and validated a pragmatic hypoglycemia risk stratification tool that uses 6 factors to categorize the 12-month risk of hypoglycemia-related ED or hospital utilization.

Identifying patients at high-risk for hypoglycemia-related utilization provides an opportunity to mobilize resources to target this minority of patients with T2D, including deintensifying or simplifying medication regimens, prescribing glucagon kits or continuous glucose monitors, making referrals to clinical pharmacists or nurse care managers, and regularly asking about hypoglycemia events occurring outside the medical setting. This is important, as more than 95% of severe hypoglycemia events may go clinically unrecognized because they did not result in ED or hospital use [6]. In addition, as the 6 inputs were identified by EMR, intervention can include automated clinical alert flags in the EMR and automated messaging to patients with elevated risk.

Several limitations exist. The study excluded secondary discharge diagnoses for hypoglycemia as these may occur due to sepsis, acute renal failure, trauma, or other causes. In addition, the external validation populations had different distributions of disease severity and case mix. The authors attribute some of the inconsistent findings to sparse data in the GH validation sample (n = 14,972). Finally, this tool was developed to stratify the population into 3 levels of risk, and it should not be used to estimate the probability of hypoglycemic-related utilization for an individual patient.

Next Article: