A major focus of diabetes management is glucose control, since persistently elevated hemoglobin A1c (A1c) levels predict adverse health outcomes. There are evolving data showing that glucose variability plays a role in predicting risk of complications. Longer term variability as reflected in A1c fluctuations over time has been linked to risk of microvascular and macrovascular complications. We have shown that increasing A1c variability over a 3 year period is independently associated with risk of myocardial infarction, ambulatory care sensitive condition hospitalization, and mortality. In this study we are developing and validating a novel clinical measure of A1c variability for its potential clinical application.
We are conducting an observational study based on the following Specific Objectives:
Objective 1. Construct statistical measures and determine the predictors of A1c variability.
Objective 2. Develop and validate a more intuitive clinical measure of A1c variability defined as A1c time-in- range. We will calculate the percentage of days an individual has an A1c level in the appropriate range, based on clinical parameters and the VA-DoD clinical practice guidelines.
Objective 3. Estimate the relationship between A1c variability, % time-in-range, and adverse health outcomes- including micro- and macrovascular complications and mortality.
The project is a retrospective observational study of patient- level administrative and claims data from VA and Medicare. Utilization and pharmacy files will be used to determine patients diagnosed with diabetes between 2004 and 2015. A three-year baseline period will be used to calculate individual A1c variability and a % time in range. The % time in range measure will be the percentage of days an individual has an A1c level in the appropriate range based on VA-DoD Diabetes Clinical Practice Guideline (CPG). Patients will be assigned to the provider that orders the most A1c tests for them during the baseline period and provider A1c variability measures will be calculated. Provider A1c variability will be used as an instrumental variable to predict individual A1c variability, controlling for process quality (Objective 1, 2). Residuals will be captured from the equations estimated in these two objectives for Objective 3 to measure the effect of individual A1c variability on health outcomes. Significant relationships will then be further validated in a VA-Medicaid population.
All variables required for Objective 1 are coded except for the provider process quality measures. Those are in the process of being coded and then these models will be run. A1c variability and time-in-range measures require that Veterans be categorized into discreet A1c target ranges based on life expectancy, clinical conditions and co-morbidities. This is consistent with the recently released VA-DoD Diabetes CPG. We have constructed regression models to predict <5 and 5-10 years of life expectancy. Top predictors include age, marital status, prescription for sulfonylureas, and specific comorbidities such as liver disease and congestive heart failure. Overall c-statistics were 0.80 and 0.77 respectively, comparing favorably with previously published efforts to predict mortality. Results from this first step will be used to assign Veterans to target A1c ranges.
Diabetes is the leading cause of end stage renal disease, lower extremity amputations and is a significant contributor to myocardial infarction, stroke and mortality. Effective diabetes management is a policy priority for VHA. There is also a need to develop new quality measures beyond A1c alone to better identify patients at risk for diabetes complications and mortality. A1c variability, as measured by fluctuations in A1c over time, is a potential candidate. Thus, we are developing and validating a new measure of A1c variability - A1c time in range - that will help clinicians and patients control A1c in a way that balances their unique long-term benefits and risks.
None at this time.
Diabetes and Related Disorders
Treatment - Observational, TRL - Applied/Translational, Diagnosis
Diabetes, Research Measure Development, Risk Factors