The Incidence and Costs of Hypoglycemia in Type 2 Diabetes
Published Online: October 10, 2011
Brian J. Quilliam, PhD; Jason C. Simeone, PhD; A. Burak Ozbay, PhD; and Stephen J. Kogut, PhD
Hypoglycemia can occur during therapy with antidiabetic drug agents and has the potential to lead to serious complications, including morbidity and mortality. While these agents are effective in treating hyperglycemia, optimal glycemic control can be more difficult to attain. The risk of hypoglycemia remains a serious adverse effect of diabetic pharmacotherapy and a potential limit to aggressive treatment.
While hypoglycemia has been well-researched in patients with type 1 diabetes, previous studies in persons with type 2 diabetes have focused on populations with limited demographics,1-3 persons using insulin in type 2 diabetes,4 or specific oral medications,5 or were conducted outside of the United States.6 Furthermore, the incidence of hypoglycemia is often measured in clinical trials and may not approximate true rates in more real-world populations. Published incidence rates (IRs) of hypoglycemia in type 2 diabetes vary from 0.02 to 0.35 events/patient/year, but differences in study design, sample populations, and definitions of hypoglycemia make direct comparisons between studies problematic.7
Research indicates that hypoglycemia places an economic burden on healthcare payers, but estimates of the magnitude of that burden differ considerably. While several non-US–based cost studies of hypoglycemia have been published, broad differences in payer systems make these analyses less generalizable to the US healthcare system.8-12 Over the past 10 years, several US studies have evaluated the medical costs of hypoglycemia.13-15 Curkendall et al evaluated costs associated with development of hypoglycemia during an inpatient hospital stay,13 while Rhoads et al evaluated the costs of hypoglycemia in patients taking insulin.14 An additional study by Pelletier and colleagues15 assessed allowed health plan costs in a large sample of patients with type 2 diabetes who experienced at least 1 diabetes-related medical complication during the study. Within this study, 0.2% of the population had a hypoglycemic event in the first year, with a mean annual allowed charge for hypoglycemia of $345 (2007 US dollars). While several studies have assessed costs associated with hypoglycemia, none have comprehensively evaluated these costs in a population of patients with type 2 diabetes taking oral antidiabetic drugs (OADs).
While 1 in 10 US adults currently have diabetes, the Centers for Disease Control and Prevention estimate that 1 in 3 adults will have the disease by 2050.16 The management of diabetes complications such as hypoglycemia and resulting costs to the healthcare system will be of greater concern in upcoming years, and comprehensive analyses areneeded to measure hypoglycemia in the US population. Our objectives in this study were: 1) to estimate the incidence of hypoglycemia in a large sample of working-age patients with type 2 diabetes; and 2) to estimate the corresponding direct medical costs of hypoglycemia.
RESEARCH DESIGN AND METHODS
The data source was the Medstat MarketScan database, including pharmacy and medical (inpatient and outpatient) claims data for the years 2004 to 2008. The Medstat MarketScan database comprises more than 4 billion patient records, 69 million covered lives, 77 contributing employers, and 12 contributing heath plans.17 The Medstat MarketScan database compiles claims from work-based insurance plans including data from currently enrolled employees, early retirees, former employees covered by COBRA, and their dependents.18 Unlike many healthcare databases, the MarketScan database provides a near-complete capture of a member’s healthcare as it contains information on inpatient and outpatient medical visits, pharmacy claims, and carved-out services.18
We conducted a retrospective cohort study of patients with type 2 diabetes and identified all persons with 2 or more inpatient and/or outpatient claims for diabetes (International Classification of Diseases [ICD]-9 250.XX) during the study period. As hypoglycemia may be the result of diabetes medications, we excluded from this eligible sample persons who were not taking at least 1 OAD. We further excluded participants who did not have at least 12 months of continuous eligibility within a noncapitated health plan following the first fill date of an OAD. Finally, we excluded persons with at least 1 medical claim (inpatient or outpatient) for type 1 (ICD-9 250.X1 or 250.X3) or gestational diabetes (648.0X) during the study period.
For each cohort member, we initiated follow-up on the first date of fill for an OAD agent and identified the first lapse in continuous eligibility beyond the initial 12 months of eligibility used for sample identification. The last day of the month preceding the first gap in coverage served as the end of follow-up for each member. We characterized the study population by describing the demographics and estimated the prevalence of antidiabetic medication use, and micro- and macrovascular complications of diabetes. To identify medications taken within 90 days of cohort entry, we evaluated pharmacy claims for prescription fills of oral and injectable antidiabetic agents. To estimate the baseline prevalence of diabetes-related complications, we used methods employed in other studies,19 and assessed inpatient and outpatient medical claims within 90 days of cohort entry.
Hypoglycemic Event Characterization
We followed all cohort patients from their respective index dates to the end of their follow-up to identify instances of hypoglycemia requiring medical intervention (inpatient, emergency department [ED], and/or outpatient). To identify hypoglycemic episodes, we used the algorithm proposed by Ginde and colleagues.20 We identified the inpatient and outpatient claims for hypoglycemia using ICD-9 codes: 251.0, 251.1, 251.2, 270.3, and 962.3. In addition, we identified claims for ICD-9 code 250.8 in the absence of other contributing diagnoses (ICD-9 259.8, 272.7, 681.XX, 682.XX, 686.9X, 707.1-707.9, 709.3, 730.0-730.2, or 731.8). Our primary end point was the first medical encounter for hypoglycemia (inpatient, emergency department, or outpatient). We repeated similar procedures to look at the incidence for each of the 3 settings separately: the first inpatient admission, the first ED visit, or the first outpatient visit.
Cost Analyses Estimation
To estimate the direct medical costs associated with hypoglycemic events, we used the total gross payment to all providers associated with an admission in the inpatient claims database and the total gross payment to a provider for a specific service in the outpatient claims database. The cost estimates included all payments made to the provider by the patient including copayments, coinsurance, and deductibles. In addition to overall measures of cost, we calculated costs for 3 mutually exclusive groups of claims for all cohort members: 1) those related to hypoglycemia (identified using the Ginde algorithm20); 2) other diabetes-related claims (primary ICD-9 250.XX); and 3) all other claims. To capture episodes of care in the outpatient and ED settings, we summed costs associated occurwith claims on the same service date as 1 episode of care. The MarketScan data already provide summary costs associated with a particular inpatient admission within the database.18 In instances where ED visits resulted in inpatient hospitalization, the collective costs for both the ED visit and the resulting inpatient hospitalization were captured within the inpatient hospitalization. All cost variables from inpatient, ED, and outpatient claims were inflated to 2008 equivalents using the regional Consumer Price Index medical care expenditure category to allow for a proper comparison of costs across years.21 Lastly, we calculated per-member-per-month (PMPM) costs by dividing total costs for each category of care by the number of person-months in the study sample.
To describe the baseline characteristics of our study sample, we calculated frequencies for categorical variables and means for continuous covariates. To estimate the incidence of hypoglycemia resulting in an inpatient admission, ED, or outpatient visit, we conducted descriptive analyses. For each eligible member, we estimated person-years (p-yrs) of follow-up as the interval from the participant’s index date to the date of the first of the following events: a hypoglycemic episode; the end of continuous eligibility; or December 31, 2008. We calculated the IR by dividing the number of hypoglycemic events by the total p-yrs of follow-up and constructed 95% confidence intervals (CIs) around the estimated IRs. The final IRs and corresponding 95% CI are presented per 10,000 p-yrs. We further stratified these estimates by age (18-34, 35-49, 50-64, or 65 years), gender, and age and gender. Unadjusted Poisson regression models were used to conduct between-group comparisons of IRs (across age and gender) and t tests were calculated to compare mean total gross payments across cost categories (eg, hypoglycemia-related encounters, other diabetes-related encounters, and all other encounters). All analyses were performed using SAS Software (version 9.2; SAS Institute Inc, Cary, North Carolina).
PDF is available on the last page.