Characteristics and Medication Use of Veterans in Medicare Advantage Plans

Talar W. Markossian, PhD, MPH; Katie J. Suda, PharmD, MS; Lauren Abderhalden, MS; Zhiping Huo, MS; Bridget M. Smith, PhD; and Kevin T. Stroupe, PhD

The Medicare Advantage (MA) program, introduced in 1982, authorized Medicare to contract with private insurers to provide healthcare coverage for a monthly prospective per enrollee capitated payment. Typically, MA plans offered supplemental benefits, such as prescription drug coverage1,2 or lower out-of-pocket premiums,1-3 that exceeded the Medicare entitlement. Studies in age-eligible Medicare beneficiaries showed that beneficiaries who enrolled in MA plans had lower pre-enrollment expenditures, fewer health conditions, lower use of services, and lower postenrollment mortality compared with beneficiaries who remained in the fee-for-service (FFS) sector.4-6 Studies also showed that MA plans attracted a higher percentage of African American beneficiaries,7 Latino beneficiaries, and beneficiaries with low education and fewer financial resources.7 MA plans treated chronic conditions more aggressively once they were diagnosed8,9 and had fewer disparities in quality of care than the FFS sector.10 Veterans were more likely to enroll in MA plans if they were nonwhite, had better self-reported functioning, exercised, or did not smoke. However, veterans were also more likely to enroll if they consumed alcohol or had more health conditions.11

Many veterans who were dually enrolled in the US Department of Veterans Affairs (VA) healthcare system and in Medicare used both VA and Medicare healthcare.12 The VA is able to track healthcare use and diagnoses of veterans in the VA system and FFS Medicare facilities using diagnosis-based instruments (eg, diagnosis-related groups,13 Diagnostic Cost Groups,14 and Ambulatory Clinical Groups15). However, the VA is unable to track healthcare use or diagnoses at MA facilities because MA plans are not required by the federal government to submit encounter-level records. As a result, this data incompleteness underestimates the prevalence of comorbidities for MA-enrolled Medicare-eligible veterans. Therefore, many studies of healthcare and medication use in Medicare-eligible veterans have excluded veterans enrolled in MA plans from their analyses to avoid this measurement error,16-20 despite the growing numbers of MA-enrolled veterans. In 2011, MA-enrolled veterans represented 21% of all Medicare-eligible VA users.21

There are limited data from empirical evaluations of the choice between FFS versus MA plans in Medicare-eligible veterans. Also, no previous study has compared medication acquisition patterns, overlapping medications, and receipt of high-risk medications (HRMs) from VA and Part D–reimbursed pharmacies by plan enrollment type. With the implementation of Medicare Part D and availability of prescription data obtained from VA and Medicare Part D–reimbursed pharmacies, this study was able to adjust for comorbidities using a pharmacy-based case mix instrument and assess patient characteristics and health conditions associated with enrollment in an FFS or MA plan by Medicare-eligible veterans during 2008 and 2009. In addition, we examined medication acquisition patterns by MA plan enrollees compared with Medicare FFS enrollees.


Subject Selection

The sampling frame for this study consists of all women and a random 10% sample of men from the VA Vital Status file. The Vital Status file contains demographic information and dates of death of individuals who received VA care, were enrolled in the VA system, or received VA compensation or pension benefits since 1992. Veterans were included if they were 65 years or older as of January 1, 2004; alive through December 2009; and enrolled in Medicare. All women veterans were included because they represent a smaller portion of the overall veteran population. Veterans who were enrolled in Part D during some but not all of 2009 were excluded. Participants were also excluded if they had missing or discrepant demographic, geographic, or medication use data.

Data Sources and Measures

Data for this study were derived from the VA and Medicare administrative databases for years 2008 and 2009. Demographic, geographic, and RxRisk-V risk classes were ascertained in 2008. Medicare Part D enrollment and medication acquisition was examined in 2009. Veterans’ Part D enrollment status was obtained from Medicare enrollment files. Consistent with previous research, a veteran was classified as an MA enrollee if the veteran was enrolled in an MA plan for at least 1 month from January 1, 2008, through December 31, 2009.6 Patient demographic characteristics, including age, race, gender, ethnicity, zip code of residence, and socioeconomic status of the veteran’s zip code of residence (eg, median household income in the zip code), were obtained from the VA Enrollment file22 and US Census Bureau data. Veterans’ priority categories, indicating which veterans are required to make co-payments for no, some, or all medications from VA pharmacies,23 were obtained from the VA Enrollment file. The differential distance between the nearest VA and Medicare-reimbursed facility was calculated using zip code information.12 Rural/urban status was based on Rural Urban Commuting Area Codes derived from zip codes.24

We obtained pharmacy use data from the VA Managerial Cost Accounting National Data Extract Pharmacy data sets25 and the Medicare Part D “Slim” file that contains all prescriptions that have been dispensed and paid through the Part D program. Based on pharmacy use, veterans were classified as 1 of 3 types of pharmacy users: VA only, Part D–reimbursed only, or dual (ie, both VA and Part D–reimbursed). Pharmacy use was measured as the number of 30-day medication supplies (eg, one 90-day supply was equivalent to three 30-day supplies) and the number of drug classes that veterans received from VA and Part D–reimbursed pharmacies.20 A medication with less than a 30-day supply was coded as one 30-day supply. The drug classes were based on the VA drug classification system that provides general categories of drugs and mostly follows the American Hospital Formulary Service Drug Information drug classification.26 For dual pharmacy users, overlapping days’ supply was defined as drug classes that veterans received from both VA and Part D–reimbursed pharmacies with at least 7 overlapping days.

The RxRisk-V is a pharmacy-based case-mix instrument that has been validated in the veteran population14 and can address the issue of missing claims data for veterans enrolled in MA plans. The RxRisk-V contains 45 disease categories based on medication classes (listed in eAppendix Table 1 [eAppendix available at]).27 To construct RxRisk-V scores for a patient population, patient-level pharmacy data are mapped into the RxRisk-V categories. We solicited pharmacist input from our team to review unclassified and new drug classes to identify whether their primary indication is associated with a specific disease state.

The Healthcare Effectiveness Data and Information Set’s list of HRMs28 was used to identify veterans who received HRMs from VA and Part D–reimbursed pharmacies in 2009. HRMs, originally codified by the Beers29 and Zhan30 criteria, are medications that should be avoided in patients 65 years or older because either the associated adverse effects outweigh the potential benefits or there are safer alternatives.
Statistical Analyses

Demographic, geographic, and medication use differences between veterans who obtained Medicare in the FFS sector and those who were enrolled in an MA plan in 2008-2009 were assessed by descriptive and bivariate statistics, including t tests and χ2 tests. We compared the number of 30-day supplies, the number of medication classes, and the most frequently occurring drug classes from VA or Part D–reimbursed pharmacies by Medicare FFS or MA plan enrollment status. We also compared receipt of HRMs from VA and Part D–reimbursed pharmacies by plan type (FFS vs MA) and pharmacy use groups (VA only, Part D–reimbursed only, or dual). Logistic regression analyses were conducted to assess factors associated with MA versus FFS enrollment in 2008-2009, which included veterans’ demographic characteristics, health status, and zip code–level factors. All analyses were conducted using Stata version 14.2 (Stata Corp; College Station, Texas). Human subject approval for this research was obtained from the Edward Hines, Jr. VA Hospital Institutional Review Board.


Veterans’ Characteristics by Medicare Type

There were 262,371 veterans who met the study inclusion criteria, of whom 203,723 (77.6%) were enrolled in Medicare FFS and 58,648 (22.4%) in MA. A majority (81.3%) of MA-enrolled veterans were enrolled for 12 months in MA plans during the year. The majority of veterans were non-Hispanic white, aged between 76 and 85 years, and residing in an urban location. Descriptive characteristics of veterans by MA enrollment status are presented in Table 1 [part A and part B]. Compared with those in the FFS sector, MA enrollees were more likely to receive medications for the treatment of pain, which had the largest difference between the 2 groups (29.4% vs 23.9%), followed by medications for congestive heart failure/hypertension (59.0% vs 56.7%), diabetes (24.8% vs 22.5%), hyperlipidemia (66.4% vs 63.1%), and benign prostatic hypertrophy (29.4% vs 27.0%).

In the adjusted analyses, compared with those in the FFS sector, MA enrollees had a 2-fold increase in the odds of having VA co-payments for some (odds ratio [OR], 2.25; 95% CI, 2.15-2.35) or all (OR, 2.41; 95% CI, 2.30-2.52) drugs. They also had greater odds of residing in urban settings (OR, 2.01; 95% CI, 1.96-2.06) and in the West (OR, 1.13; 95% CI, 1.10-1.16) compared with all other regions. MA and FFS enrollees differed significantly regarding 9 of the 10 most common clinical conditions in this population (listed in Table 1), excluding hypertension. Pain was the condition most associated with MA enrollment; MA enrollees had 33% (OR, 1.33; CI, 1.30-1.37) greater odds of having pain medication prescribed compared with FFS sector enrollees.

Medication Use by Medicare Type

A higher percentage of MA enrollees were enrolled in Medicare Part D compared with FFS sector enrollees (73.7% vs 24.3%). In Figure 1, we present unadjusted analyses comparing medication use by MA enrollment status. MA enrollees were more likely to use any VA or Medicare-reimbursed medication (92.8% vs 70.2%) and to be dual (32.3% vs 7.0%) or Medicare-only (31.1% vs 14.5%) pharmacy users. For VA or Medicare-reimbursed medication users, Medicare FFS enrollees and MA enrollees received a similar number of 30-day medication supplies in 2009 (mean [SD], 62.6 [43.9] vs 62.9 [40.9], respectively; P = .3482) from VA or Part D–reimbursed pharmacies (data not shown). Medicare FFS enrollees obtained medications from fewer drug classes (mean [SD] = 8.3 [5.2]) compared with MA enrollees (mean [SD] = 8.8 [5.1]; P <.001) from VA or Part D–reimbursed pharmacies. A comparison of the prevalence for all the RxRisk-V classes and a detailed comparison of medication use from VA or Part D–reimbursed pharmacies by Medicare plan enrollment status are presented in eAppendix Tables 1 and 2.

The 10 most frequently prescribed drug classes by Medicare enrollment status are presented in Table 2 [part A and part B]. In both groups, the most frequently obtained drug class from VA and Part D–reimbursed pharmacies were antilipemic agents, followed by β-blockers. For example, 38.5% of Medicare FFS enrollees obtained antilipemic agents from VA pharmacies and 12.1% obtained them from Part D–reimbursed pharmacies. Opioid analgesics, quinolones, and antidepressants were drug classes dispensed from Part D–reimbursed pharmacies that were not included in the list of 10 most frequently obtained drug classes from VA pharmacies. The most frequent drug classes that dual pharmacy users obtained with an overlapping days’ supply from both VA and Part D–reimbursed pharmacies for more than 7 days are also presented; the list included antilipemic agents, β-blockers, antidepressants, and opioid analgesics, among others (Table 2).

In total, 15.1% of the veterans received HRMs; of these, 6.7% received the medication from VA pharmacies and 7.9% from Part D–reimbursed pharmacies (Figure 2A). A higher proportion of MA enrollees received HRMs compared with the FFS sector enrollees (17.0% vs 14.3%; P <.001) (Figure 2B). A higher proportion of dual users (24.4%) received HRMs compared with users of Part D–reimbursed (20.7%) or VA-only (10.1%) pharmacies (P <.001) (Figure 2C). The 10 most frequent HRMs that veterans obtained from VA and Part D–reimbursed pharmacies are presented in Table 3.

Almost one-fourth of Medicare-eligible veterans were enrolled in an MA plan for at least 1 month in 2008-2009, representing more than 1 million veterans enrolled in the VA and Medicare. A higher proportion of MA patients (73.7%) enrolled in Medicare Part D compared with FFS patients (24.3%). Moreover, a significantly higher portion of MA enrollees (17.0%) compared with the FFS enrollees (14.3%) were receiving HRMs. MA plan enrollees also differed from FFS sector enrollees in several of the demographic, geographic, and clinical characteristics that we examined.

The percentage of patients with common clinical conditions differed significantly between MA and FFS enrollees, with the largest percentage difference between the 2 groups being for pain. Veterans with pain had higher odds of enrollment in MA plans versus FFS plans. One possible explanation could be access to pain medications through the Medicare Part D program, with 73.7% of MA enrollees being enrolled in Part D versus just 24.3% of the FFS enrollees. Chronic pain is frequently encountered in the United States; it affects an estimated 20% of the population.31 Opiates are used routinely for the treatment of pain and frequently overprescribed.32 Prescriptions for opioids have decreased slightly, but opioid-related overdoses continue to increase and have been associated with an epidemic of opioid abuse, dependence, and overdose.33 Opioid analgesics were the third most frequent class of medications acquired from Part D–reimbursed pharmacies for both Medicare FFS and MA enrollees. Notably, opioid analgesics were not on the list of frequent classes of medications acquired from VA pharmacies. Since the recent implementation of state-level prescription drug monitoring programs, providers within the VA can henceforth explore the receipt of these medications by veterans from outside the VA system.34

We found that MA-enrolled veterans used more medications and were more likely dual or Medicare-only pharmacy users compared with the FFS sector enrollees. In a previous study of FFS-enrolled veterans, results showed that a substantial proportion of veterans received their medications from Part D–reimbursed and VA pharmacies.35 According to the authors, this might pose safety risks if providers in one healthcare system did not know about medications prescribed by providers in another system. Our results showed that the potential safety risks associated with dual pharmacy use among MA-enrolled veterans might be much greater because these veterans are more likely to be dual pharmacy users than those in the FFS sector (32.3% vs 7.0%). Similar to the previous study, we also found that some of the medications filled at Part D–reimbursed pharmacies were drugs that carried a significant drug–drug interaction profile (eg, quinolones) or included narcotics that require cross-system monitoring to control potential abuse and diversion. Antidepressants and opioid analgesics were on the list of most frequent drug classes that dual pharmacy users obtained from both Part D–reimbursed and VA pharmacies with overlapping supply days.

In 2009, a previous study found that 21.5% of MA enrollees received at least 1 HRM.36 In our study, 15.1% of the veterans received HRMs. The proportion was significantly higher among dual (24.4%) and Part D–reimbursed (20.7%) pharmacy users compared with VA-only pharmacy users (10.1%) (P <.001). Because MA enrollees were more likely to be dual or Medicare-only pharmacy users, the proportion of HRMs was significantly higher for MA enrollees (17.0%) compared with those in the FFS sector (14.3%) (P <.001). The computerized patient record system at the VA alerts physicians about prescribing HRMs for elderly veterans. Although alert fatigue is a well-documented phenomenon,37 our data demonstrate that a significant proportion of HRMs are received from outside of the VA system. Other possible reasons for lower receipt of HRMs from VA pharmacies include availability of alternative nonpharmacological treatments at the VA, norms and training practices of VA providers, and availability of onsite pharmacists at the VA.


There are limitations to this descriptive study. Due to constraints in VA administrative data, we were not able to assess relevant veteran characteristics, including education level, income, or preference for care, that might impact their choice of Medicare plan. As proxies for these variables, we used the socioeconomic status of the veterans’ zip code of residence, veterans’ priority categories, and the differential distance between the nearest VA and Medicare-reimbursed facility. Veterans who did not use medications in 2008 from VA or Medicare-reimbursed pharmacies were not classified by the RxRisk-V. This may have decreased the prevalence of the chronic conditions in our population if veterans received medications that are not covered by either VA or Medicare benefits. Also, RxRisk-V assigns disease categories based on treatment for conditions, and plan choice may impact treatment for conditions. Because the VA serves a unique population, which is predominately male and older and has low socioeconomic status, our results demonstrating differences in patient sociodemographic, clinical, and medication use by plan enrollment status may not be generalizable to other settings or populations. Although the data are from 2009, the findings inform current policy discussions because they highlight the importance of coordinating multiple system use, which is a major issue in the current US healthcare system. Finally, although our list of HRMs was specific to the elderly, it did not include all medications with safety concerns (eg, QTc interval prolongation with quinolones or bleeding with warfarin). Therefore, our results regarding the prevalence of HRMs should be considered a conservative estimate of the use of medications with safety concerns in an elderly population.


Medication reconciliation within health systems can identify medication discrepancies and reduce potential harm.38 Most electronic health records (EHRs), including VA’s health information system, allow providers to add services or medications received outside their health system. However, this process relies on providers asking patients for this information and documenting it in the EHR as non-VA medication orders. According to findings of a previous study, more than 38% of veterans who obtained non-VA medications did not discuss these medications with VA physicians.39 Studies outside the VA system have also found that doctor–patient communication about drugs was suboptimal.40

Our results highlight that MA enrollees were more likely to be dual or Medicare-only pharmacy users compared with the FFS sector enrollees. These veterans acquired many drug classes, such as opioid analgesics, more frequently from Part D–reimbursed pharmacies than from VA pharmacies. Providers both inside and outside of the VA caring for veterans should consider that substantial information about the medication profiles of their patients might be unavailable in their healthcare systems’ EHRs, and they should ask patients for this information. 


The authors wish to thank Fran Cunningham, PharmD, director, and Muriel Burk, PharmD, clinical pharmacy specialist, at the VA Center for Medication Safety (Hines, IL) for reviewing earlier drafts of the manuscript.
Print | AJMC Printing...