Dual Use of Department of Veterans Affairs and Medicare Benefits and Use of Test Strips in Veterans With Type 2 Diabetes Mellitus

Importance  Self-monitoring of blood glucose is a costly component of care for diabetes mellitus, with unclear benefits for patients not taking insulin. Veterans with dual Department of Veterans Affairs (VA) and Medicare benefits have access to test strips through both systems, raising the potential for overuse.

Objectives  To examine the patterns of test strip receipt among older veterans with diabetes and determine whether receipt of strips from dual health care systems is associated with overuse.

Design, Setting, and Participants  We performed a cross-sectional, retrospective cohort study using national VA administrative data linked to Medicare Parts A, B, and D claims for fiscal years 2008 and 2009. A total of 363 996 community-dwelling veterans 65 years or older with diabetes who used the VA health care system and received test strips in fiscal year 2009 were included in the study.

Exposures  Receipt of test strips from the VA only, Medicare only, or both the VA and Medicare; covariates included sociodemographics, comorbidity, diabetes complications, and hemoglobin A_1c level.

Main Outcomes and Measures  Quantity of test strips dispensed and overuse of test strips, defined as more than 1 strip per day (>365 strips per year) among those taking no diabetes medications, oral diabetes medications alone, or long-acting insulin without short-acting insulin or more than 4 strips per day (>1460 strips per year) among those taking short-acting insulin.

Results  Overall, 260 688 older veterans (71.6%) with diabetes received strips from the VA only, 82 826 (22.8%) from Medicare only, and 20 482 (5.6%) from the VA and Medicare. Veterans receiving strips from both the VA and Medicare received more strips (median, 600; interquartile range [IQR], 350-1000) than the Medicare only (median, 400; IQR, 200-700) and VA only (median, 200; IQR, 100-500) groups (P < .001) and had substantially greater odds of overuse than the VA only group (55.4% vs 15.8%) (adjusted odds ratio [OR], 16.3; 95% CI, 14.6-18.1 for no medications; 55.3% vs 6.0%; OR, 19.8; 95% CI, 18.9-20.8 for oral medications; 87.4% vs 65.5%; OR, 3.69; 95% CI, 3.30-4.14 for long-acting insulin; and 32.8% vs 13.5%; OR, 3.24; 95% CI, 3.05-3.45 for short-acting insulin). Patterns were similar when using more conservative thresholds of overuse.

Conclusions and Relevance  Veterans who receive glucose test strips through both the VA and Medicare use more strips and are more likely to potentially overuse strips. These results illustrate the profound importance of understanding dual VA and Medicare coverage and are emblematic of waste and inefficiency.

Type 2 diabetes mellitus affects more than 25 million individuals in the United States and more than 1 in 4 adults older than 65 years.¹ For individuals following regimens that require multiple daily insulin injections, regular self-monitoring of blood glucose (SMBG) is a critical part of diabetes self-management. The American Diabetes Association, the Canadian Diabetes Association, the National Institute for Health and Care Excellence, and the Department of Veterans Affairs and Department of Defense all recommend self-monitoring for patients managed with such regimens.^2-5

Although the American Diabetes Association recently revised upward its recommendation for glucose testing in individuals receiving multiple daily insulin injections,^2,6 little consensus exists regarding the optimal frequency of SMBG for patients with type 2 diabetes receiving noninsulin therapy or only basal insulin therapy.^7-9 Several trials and meta-analyses^10-15 have questioned the utility of routine SMBG for most patients with type 2 diabetes, some⁷ concluding that SMBG has no clinically relevant effect in lowering hemoglobin A_1c levels and no beneficial effect in identifying hypoglycemic events. In fact, one Choosing Wisely initiative suggests that physicians avoid recommending daily SMBG in patients with type 2 diabetes who are not using insulin.¹⁶

The costs of SMBG in the United States are substantial. Medicare contractors paid more than $1.2 billion for test strips and/or lancets in 2007,^17-20 and in the VA, where use and cost are lower, approximately $50 million per year is spent on test strips alone. Medicare coverage guidelines allow for 100 test strips every 3 months for patients not treated with insulin (once per day) and 100 test strips every month for patients treated with insulin (3 times per day),^17-20 with the ability to receive additional strips if clinically indicated. In 2010 and 2011, the Office of the Inspector General for Medicare reported that most sampled high-use claims for strips were not in line with these guidelines.^17-20 The VA places stricter limits on SMBG, typically allowing for 50 strips per 150 days (approximately 2 strips per week) for those not receiving insulin, with additional strips offered with proper clinical justification. For patients using insulin, the VA typically recommends dispensing enough strips to match the number of daily insulin injections.²¹

Veterans who use the VA with concomitant Medicare coverage have access to test strips through both the VA and Medicare, which may reduce the effectiveness of each system’s attempts to discourage overuse. No prior studies have examined the effect of dual system use on the receipt of test strips. Our objective was to examine receipt of glucose test strips from the VA and Medicare for a national cohort of VA patients 65 years or older with diabetes. We hypothesized that those obtaining test strips through both the VA and Medicare would be more likely to obtain a greater number of strips and a potentially excessive number of strips compared with veterans obtaining test strips through the VA or Medicare alone.

The study was approved by the institutional review board of the VA Pittsburgh Healthcare System with a waiver of informed consent. We linked national patient-level data from fiscal years (FYs) 2008 and 2009 (October 1, 2007, through September 31, 2009) from the VA and the Centers for Medicare & Medicaid Services (CMS). The VA data comprised the medical SAS data sets for patient demographics and diagnosis codes for all inpatient and outpatient visits, the VA Pharmacy Benefits Management data for dispensed outpatient prescriptions and glucose test strips, and Decision Support Service data for information on VA nursing home stays and hemoglobin A_1c values. The CMS data comprised the MedPAR file for inpatient and skilled nursing facility stays; outpatient, carrier, and durable medical equipment claims (fee for service); denominator files for patient enrollment and sociodemographic information; minimum data set for information on nursing home stay; and Part D event files for prescription drug use.

We previously identified a national cohort of all veterans with type 2 diabetes using International Classification of Diseases, Ninth Revision (ICD-9), diagnosis codes specific to type 2 diabetes and receipt of diabetes medications, adapted from Miller et al and Gellad et al.^22-24 The original cohort included 1 158 809 veterans with type 2 diabetes older than 18 years with at least one outpatient visit in the VA in 2009 (thus limited to VA health care users). For the current study, we identified a subgroup of patients 65 years or older with diabetes documented in FY 2008 (n = 565 715). We then excluded 29 849 veterans (5.3%) with hospital or nursing home stays of longer than 30 days in FY 2009, identified using the VA and CMS files. To focus on veterans receiving test strips from the VA or fee-for-service Medicare, we also excluded 171 870 veterans (32.1%) with no dispensed test strips from either source.

We measured the dispensing of glucose test strips in FY 2009 and determined whether veterans received excess strips. Strips are identified as dispensed prescriptions in the VA data and as dispensed durable medical equipment in Medicare data for the fee-for-service population (Healthcare Common Procedure Coding System code A4253).²⁵ Because recommended testing frequency differs depending on the medication regimen, we used both the VA Pharmacy Benefits Management data and Medicare Part D prescription drug events to categorize each veteran into 1 of 4 mutually exclusive groups based on the type of diabetes medication used in FY 2009: (1) none, (2) oral only (includes patients using the daily injectable exenatide [0.15%]), (3) long-acting insulin (without any short-acting insulin), and (4) short-acting insulin (including premixed preparations). Individuals in the long-acting insulin group, which included both glargine-detemir and insulin isophane suspension, inject once or twice daily without adjusting dosages based on glucose testing, whereas those in the short-acting insulin group typically inject 2 to 4 times daily. Veterans were assigned to insulin categories based on any use during the year, and veterans who used oral medications along with insulin were included in the insulin groups.

To define potential overuse of strips (hereafter referred to as overuse or excess strips), we assumed that, for most patients with type 2 diabetes who are taking no diabetes medications, oral medications only, or long-acting insulin without any short-acting insulin, once-daily testing or less would be sufficient to ensure high-quality diabetes self-management.^5,7,10,14-16 Thus, for these individuals, excess strip use was defined as the receipt of more than one test strip per day (or >365 strips in FY 2009). For individuals using any short-acting insulin, we assumed that testing 4 times per day (requiring 1460 strips per year) would be sufficient for most individuals, and any use of more than that amaunt would indicate overuse.

We performed several sensitivity analyses varying the definition of overuse. First, for individuals taking long-acting insulin, we used a cut point of more than 2 strips per day (>730 strips per year) instead of 1 strip per day, given that some individuals using twice-daily insulin isophane suspension could potentially test twice daily. Second, for individuals using short-acting insulin, we used a cut point of more than 6 strips per day (>2190 strips per year) rather than 4 strips per day. Finally, because concerns about hypoglycemia associated with sulfonylureas may justify increased testing in patients prescribed these medications, we repeated our analysis of the oral medication–only group and excluded patients who used sulfonylureas.

Our primary independent variable of interest was the source for strips dispensed, defined at the patient level and categorized as the VA only, Medicare only, or both the VA and Medicare (ie, dual use). We constructed variables for patient age and sex using the VA data. We defined patient race/ethnicity as non-Hispanic white, non-Hispanic black, Hispanic, or other or missing using the VA data, supplemented by data from the Medicare denominator file when missing from the VA.²⁶ To account for differences in health status that might affect use of strips, we also assessed baseline (FY 2008) comorbidity, number of diabetes complications, and hemoglobin A_1c levels, along with a VA or Medicare hospitalization during the study year (FY 2009). We calculated a count of comorbid conditions defined in the Charlson Comorbidity Index (excluding diabetes) using ICD-9 codes recorded in the VA or Medicare files.^27,28 We also recorded the number of diabetes complications (ie, diabetic retinopathy, nephropathy, neuropathy, and diabetes-related peripheral vascular diseases) using ICD-9 codes from both the VA and Medicare.²⁹ For individuals undergoing hemoglobin A_1c testing within the VA in the baseline study year, we defined 3 levels of glucose control (ie, hemoglobin A_1c <7%, hemoglobin A_1c of 7% to <9%, and hemoglobin A_1c ≥9%) using the last hemoglobin A_1c level of the year. We classified patients with hemoglobin A_1c testing in Medicare only (using Healthcare Common Procedure Coding System codes 83036 and 83038), for whom hemoglobin A_1c levels are not available, as monitored in Medicare only and those with no testing in the VA or Medicare as not monitored.

We also included an indicator for Medicaid enrollment (from the Medicare denominator file), given that Medicaid-enrolled veterans would have lower Medicare copayments for test strips. Veterans pay no copayment for strips received through the VA. Finally, we defined variables to describe the dual receipt of diabetes care more generally beyond simply the receipt of strips. The first was a count of the number of outpatient diabetes-related office visits in Medicare for each veteran, and the second was an indicator for having an outpatient diabetes-related office visit in both Medicare and the VA during the study year.

We compared characteristics of strip users across the VA only, Medicare only, and the VA and Medicare use groups. We tested differences in characteristics by source of strips (overall differences and pairwise comparisons) with χ² tests for categorical variables and Kruskal-Wallis tests for continuous variables due to nonnormality.

We calculated the median and interquartile range (IQR) for the total number of test strips received by patients during the study year. We compared differences in the total number of strips received by patients by source of strips (overall differences and pairwise comparisons) using Kruskal-Wallis tests. For each patient, we compared the total quantity of strips dispensed to the cut points used to define overuse. We quantified the proportion of patients with excess strips and the total number of excess strips dispensed.

To examine the association between test strip overuse and test strip source, we performed multivariable analyses separately for patients within each medication group, adjusting for patient demographics and clinical characteristics as described above using logistic regression models with random effects for the VA parent facilities.

All analyses were performed using SAS statistical software, version 9.3 (SAS Institute Inc), and STATA software, version 12 (Stata Corp). Statistical tests were 2-sided with P < .05 considered statistically significant.

The overall sample included 363 996 veterans 65 years and older who received test strips in FY 2009. The median age was 75 years, 98.7% were male, 82.5% were non-Hispanic white, and 71.3% had one or more comorbid conditions (Table 1). Veterans most frequently received strips from the VA only (71.6%); 22.8% received strips from Medicare only and 5.6% from both the VA and Medicare. Compared with veterans receiving strips in the VA only, those receiving strips from Medicare (either alone or in addition to the VA) were older and more likely to be non-Hispanic white, to have diabetes complications, to have a hospitalization during the study year, to have hemoglobin A_1c testing performed outside the VA, and to have a greater number of diabetes-related office visits in Medicare (P <.001 for all).

Overall, 87.1% of the study sample received diabetes medications from the VA or Medicare. Just more than half of the sample (52.9%) received oral medications only; 34.2% used some form of insulin, including regimens with (21.9%) and without (12.3%) short-acting insulins. Veterans receiving test strips from Medicare alone were more likely than those receiving strips from the VA or both the VA and Medicare to be taking no diabetes medications (23.1% vs 9.8% and 9.9%, respectively; P < .001) (Table 1).

The median number of strips dispensed was 250 (IQR, 100-600). Compared with veterans receiving strips from the VA only (median, 200; IQR, 100-500), the median number dispensed was 2 times higher in those receiving strips from Medicare only (median, 400; IQR, 200-700) and 3 times higher in those who obtained strips from both the VA and Medicare (median, 600; IQR, 350-1000) (P < .001 for both comparisons). Among those receiving no diabetes medications (Figure 1A), the median number of strips dispensed was 4 times as high in those receiving strips from Medicare only (median, 400; IQR, 200-600) or both Medicare and the VA (median, 400; IQR, 250-700) compared with those receiving strips from the VA only (median, 100; IQR, 50-200) (P < .001).

Among each of the baseline hemoglobin A_1c levels (Figure 1B), veterans receiving strips from the VA only received significantly fewer strips than those receiving strips from Medicare only or the VA and Medicare (P <.001 for all).

For patients who received strips from both the VA and Medicare, the median percentage of strips obtained from the VA was 38.5% (IQR, 25.0%-60.0%), but this percentage varied within and across medication subgroups. In the no medication and oral medication–only groups, a median of 33.3% of strips (IQR, 20.0%-50.0%) were received from the VA; in the long-acting and short-acting insulin groups, 48.4% (IQR, 25.0%-66.7%) and 50.0% (IQR, 26.4%-70.0%) of strips were received from the VA, respectively.

Overall, 24.7% of veterans received excess test strips, including 15.8% receiving strips from the VA only, 45.7% using strips from Medicare only, and 53.5% receiving strips from both the VA and Medicare (Table 2). Among veterans taking no diabetes medications, 26.2% received excess strips, including 6.1% of those receiving strips from the VA only compared with 50.1% of those receiving strips from Medicare only and 55.4% receiving strips from the VA and Medicare (P < .001). We observed a similar pattern for veterans receiving oral medications only. For veterans receiving any form of insulin, the effects of the test strip source were similar, but the magnitudes of the differences were smaller.

In sensitivity analyses that removed sulfonylurea users from the oral medication–only group and increased the threshold of overuse for veterans taking insulin, the proportions of veterans with overuse decreased but remained significantly higher in veterans receiving strips from Medicare only and the VA and Medicare compared with the VA only (P < .001 for all) (Table 2).

Of a total of 157.1 million strips dispensed to all veterans during the study year, 31.2 million (19.9%) were used in excess: 14.7% of strips dispensed to veterans receiving strips from the VA only, 27.5% of strips dispensed to those receiving strips from Medicare only, and 31.7% of strips dispensed to those receiving strips from both the VA and Medicare (Figure 2). Among veterans taking no diabetes medication who received strips from both the VA and Medicare, almost half (46.1%) of the strips dispensed were in excess. The differences between the VA only and Medicare only groups were less pronounced among insulin users.

In multivariable analyses, veterans receiving strips from both the VA and Medicare had significantly greater odds of test strip overuse than veterans receiving strips from the VA only across all diabetes medication groups (odds ratio [OR], 16.3; 95% CI, 14.6-18.1 for no medications; OR, 19.8; 95% CI, 18.9-20.8 for oral medications only; OR, 3.69; 95% CI, 3.30-4.14 for long-acting insulin; and OR, 3.24; 95% CI, 3.05-3.45 for short-acting insulin) (Table 3). Veterans receiving strips from both the VA and Medicare also had significantly higher odds of test strip overuse compared with veterans receiving strips from Medicare only. In sensitivity analyses that removed sulfonylurea users from the oral medication–only group and increased the threshold of overuse for veterans taking insulin, the ORs were of similar or larger magnitude and similar direction.

Among older veterans with type 2 diabetes, we found remarkable differences in the quantity of test strips received when veterans access strips in multiple health systems. Among veterans taking no diabetes medications, for whom long-term daily SMBG is almost never indicated, the median number of strips received was 4 times as high in those receiving strips from both the VA and Medicare compared with the VA only, increasing the odds of overuse by more than 15 times. Among all test strip users, 1 in 5 strips dispensed was potentially unnecessary, amounting to more than 30 million test strips in 1 year. This analysis joins a previous study²³ that compared brand name drug use to illustrate the profound differences in utilization between patients with diabetes treated in the VA and those treated through Medicare.

The use of SMBG illustrates the tension between encouraging self-management for patients with diabetes and the potential for overuse. The argument, based on these data, that the VA more efficiently manages test strips rests on 2 assumptions: (1) that veterans receiving care from the VA only do not have difficulty controlling their blood glucose levels or more frequent hypoglycemic events because of limitations in test strips and (2) that veterans receiving strips through Medicare are not substantially sicker and thus require additional test strips. There is no reason to suspect that diabetes control is weakened because of the limitations included on test strips within the VA given that several systematic reviews^30-33 have found only limited evidence of clinical benefit, if any, for patients not using insulin and because the VA performs as well or better than commercial health plans and Medicare on several measures of quality for diabetes. We do not suspect profound differences in hypoglycemic events between the test strip groups, although we are limited in our ability to accurately measure hypoglycemia using these data. Although veterans receiving strips though Medicare were older with more coded diabetes complications and hospitalizations and thus might require more strips, our findings were robust to statistical adjustment for these clinical characteristics. Furthermore, our defined thresholds for overuse were conservative for those taking no medications and oral medications only, in which the most striking differences were present between the VA and Medicare, and our results were robust to sensitivity analyses that altered the thresholds.

For Medicare policy makers, the findings should reinforce the notion that waste and inefficiency must be addressed. In the 2003 Medicare Modernization Act (the law that established Medicare Part D), policy makers included a competitive bidding program for durable medical equipment, including diabetes testing supplies, with the goal of reducing costs and unnecessary use. The law became fully implemented as a national mail order program in July 2013. In an analysis of the mail order program’s early effects, CMS concluded that many beneficiaries had previously been receiving excessive and inappropriate replacement supplies.³⁴ The early results of the program are promising, having reduced glucose testing strips and costs substantially, but whether Medicare is able to reach the levels of test strip use attained by the VA and whether dual receipt of strips will remain a problem for the VA remain to be seen.

For both VA and CMS policy makers, the findings will reinforce concerns about the effect of dual use of benefits on cost, quality, and safety.^27,35-45 Dual use, in fact, will become increasingly important for the VA as the Medicaid expansion and insurance exchanges created by the Patient Protection and Affordable Care Act take hold. A key finding in our analysis is not only that Medicare is a source of potential overuse for VA patients but also that the VA is a source of potential overuse for Medicare patients; the same will surely be true with an expanded Medicaid program. Dual receipt of test strips is accompanied by other types of dual system care: almost half of patients in the VA and Medicare test strip group underwent hemoglobin A_1c testing in both systems, and almost two-thirds had a diabetes-related office visit in both systems. Although claims information on veterans treated through Medicare and Medicaid is available to the VA (as we demonstrate in this analysis), it is typically outdated by several years, making it difficult to incorporate the information into real-time planning. To our knowledge, claims information from the VA is not available to CMS. The solution, short of the establishment of regional health information exchanges nationwide, is the more active sharing of claims information across CMS and the VA.

This study has important limitations. First, our data are from FY 2009, and it is unclear how policy changes since that time have affected inappropriate use by veterans receiving strips from both the VA and Medicare. We are limited by the data available, and it is only recently that Medicare Part D data on veterans, which are necessary to understand test strip use in the context of different medication regimens, have been made available. Our data provide a baseline for understanding the effect of recent Medicare changes on test strip use. Second, there may be unmeasured differences between veterans in the VA only, Medicare only, and the VA and Medicare groups that could affect the clinical need for strips (and thus the appropriateness of overuse thresholds), such as differences in health literacy, propensity for hypoglycemia, patient preference, and occupation (eg, checking glucose level before operating machinery); however, it is unlikely that these differences are of sufficient magnitude to explain a 4-fold difference in strip receipt. Third, we focused our analyses on veterans enrolled in the VA using test strips in the VA or fee-for-service Medicare; thus, our results are not generalizable to all veterans or to veterans dually enrolled in Medicare Advantage plans and may underestimate the prevalence of dual use of test strips. Fourth, it is possible that some veterans could be receiving $4 generic medications paid with cash and thus are misclassified into medication groups; however, based on the generous VA drug benefit and prior research,⁴⁶ we do not believe this population would be large or could explain our findings.

We report evidence of substantial overuse of glucose test strips among dual health care system users. The results are robust to adjustment for clinical characteristics and are especially pronounced in individuals who take no diabetes medications or oral medications only. These findings illustrate the profound importance of understanding dual health system care and are emblematic of waste and inefficiency that must be addressed.

Accepted for Publication: August 2, 2014.

Corresponding Author: Walid F. Gellad, MD, MPH, Center for Health Equity Research and Promotion, Veterans Affairs Pittsburgh Healthcare System, University Drive, Mail Code 151C, Pittsburgh, PA 15240 (walid.gellad@va.gov).

Published Online: November 10, 2014. doi:10.1001/jamainternmed.2014.5405.

Author Contributions: Dr Gellad had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Gellad, Thorpe, Mor, Good, Fine.

Acquisition, analysis, or interpretation of data: Gellad, Zhao, Thorpe, Mor, Good.

Drafting of the manuscript: Gellad.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Gellad, Zhao, Mor.

Obtaining funding: Gellad.

Administrative, technical, or material support: Good.

Study supervision: Gellad, Mor, Fine.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grant CDA 09-207 from the Veterans Affairs Health Services Research & Development Service and from pilot funding from the Veterans Affairs Center for Health Equity Research and Promotion (Dr Gellad).

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

Disclaimer: This work represents the opinions of the authors alone and does not necessarily represent the views of the Department of Veterans Affairs or the US government.