BACKGROUND/RATIONALE:
An important goal in improving diabetes care is increased control of hyperglycemia to prevent or delay the onset of diabetes complications. One major barrier to more intensive treatment is the risk of hypoglycemia, a common and potentially hazardous side effect of treatment with insulin and oral secretagogues, such as sulfonylureas. Severe hypoglycemia may result in serious consequences and has been associated with increased mortality in several recent trials. Yet, little is known about the frequency and predictors of hypoglycemia in diabetes patients, in VA, and in other patient populations.

OBJECTIVE(S):
We sought to fill this gap and provide critical information on the incidence of hypoglycemia, patterns of its occurrence in relation to treatments, and other predictors that will be useful to clinicians and managers.

METHODS:
We conducted a series of cross-sectional and longitudinal studies using linked national VA patient care data. We used the Diabetes Epidemiology Cohorts (DEpiC), a national linked database of all veteran VA diabetes patients since 1998 developed and maintained by our research team. Hypoglycemic events among VA diabetes patients were identified using previously developed methods based on ICD-9-CM codes from inpatient and outpatient encounters in VA care and non-VA care covered by Medicare. Hypoglycemia rates were estimated overall, separately for inpatient and outpatient events, and in patients stratified by specific medication regime and other factors. Particular attention was given to patients recently initiating insulin or sulfonylurea, or discharged from hospital. A large number of potential predictors of hypoglycemia were evaluated including diabetes treatment history, predisposing medical conditions, and recent events, such as medication changes, surgeries, and other hospitalization. Multivariable logistic regression was used to estimate independent risks of hypoglycemia for various potential predictors. Robust and parsimonious prediction models were developed for hypoglycemia risk using recursive partition classification trees (CART).

FINDINGS/RESULTS:
Nearly 1.5 million hypoglycemia events were identified as occurring in diabetes patients in VA from 1998 through 2009. Our methods for identifying hypoglycemia probably found only the most serious events and, while a formal evaluation was not done, we do provide some evidence of their value. Nearly 24% of all diabetes patients had at least one event, with annual rates of 7.0% per year for all patients and 15.4% for those using insulin. Many predictors of hypoglycemia were identified. Strong risks were found for indicators of labile or brittle diabetes: prior hypoglycemia, history of ketoacidosis or hyperosmolar coma, and high hemoglobin A1c levels. Increased risk was associated with a number of other medical conditions, particularly other diabetes complications, and with initiation of a new medication or hospitalization in the past two weeks. Risk increased with age, and was higher in minority ethnic groups. The CART analysis identified three key variables for classifying patients at high risk: prior hypoglycemia, recent hospitalization, and other diabetes complications, and together these form a simple but powerful algorithm for identifying diabetes patients at high risk of hypoglycemia.

IMPACT:
The findings and models from this study provide critical information on hypoglycemia among VA diabetes patients, including estimates of the magnitude of the risk and the means to identify who is at highest risk under various clinical circumstances. These patients should be monitored carefully and should only receive intensification of treatment with sufficient care and oversight. These findings are being disseminated and may be used directly in clinical practice and in diabetes care management with improvements in pharmacy practices, coordination of care, and in developing management strategies and tools. In this way, they may help to prevent this dangerous treatment complication and improve future prescribing for safe, effective treatment of diabetes.

---

External Links for this Project

---

PUBLICATIONS:

Journal Articles

1. Vimalananda VG, Miller DR, Christiansen CL, Wang W, Tremblay P, Fincke BG. Cardiovascular disease risk factors among women veterans at VA medical facilities. Journal of general internal medicine. 2013 Jul 1; 28 Suppl 2:S517-23. [view]

DRA: none
DRE: Treatment - Observational
Keywords: Adverse events, Care Management, Diabetes
MeSH Terms: none