Geographic Variation in the Prevalence of High-Risk Medication Use Among Medicare Part D Beneficiaries by Hospital Referral Region

BACKGROUND: Understanding geographic patterns of high-risk medication (HRM) prescribed and dispensed among older adults may help the Centers for Medicare & Medicaid Services and their partners develop and tailor prevention strategies. OBJECTIVE: To compare the geographic variation in the prevalence of HRM use among Medicare Part D beneficiaries from 2011 to 2013, for Medicare Advantage Prescription Drug (MA-PD) plans and stand-alone Prescription Drug Plans (PDPs). METHODS: This retrospective study used the data of a 5% national Medicare sample (2011-2013). Beneficiaries were included in the study if they were aged ≥ 65 years, continuously enrolled in MA-PDs or PDPs (~1.3 million each year), and had ≥ 2 prescriptions for the same HRM (e.g., amitriptyline) prescribed and dispensed during the year based on the Pharmacy Quality Alliance’s (PQA) quality measures for HRM use. Multivariable logistic regression was used to estimate adjusted annual HRM use rates (i.e., adjusted predictions, average marginal predictions, or model-adjusted risk) across 306 Dartmouth Atlas of Health Care hospital referral regions (HRRs), controlling for sociodemographic, health-status, and access-to-care factors. RESULTS: Among eligible beneficiaries each year (1,161,076 in 2011, 1,237,653 in 2012, and 1,402,861 in 2013), nearly 40% were enrolled in MA-PD plans, whereas the remaining 60% were in PDP plans. The adjusted prevalence of HRM use significantly decreased among Medicare beneficiaries enrolled in MA-PD (13.1%-8.4%, P < 0.001) and PDP (16.2%-12.2%, P < 0.001) plans from 2011 to 2013. For MA-PD and PDP beneficiaries, HRM users were more likely to be (all P < 0.001) the following: female (MA-PD: 70.4% vs. 59.9%; PDP: 72.8% vs. 62.5%); White (MA-PD: 84.6% vs. 81.4%; PDP: 86.6% vs. 85.3%); with low-income subsidy or dual eligibility for Medicaid (MA-PD: 22.3% vs. 16.6%; PDP: 29.2% vs. 23.3%); and disabled (MA-PD: 15.6% vs. 8.7%; PDP: 15.4% vs. 8.5%) compared with non-HRM users in 2013. In 2013, significant geographic variation existed, with the ratios of 75th-25th percentiles of HRM use rates across HRRs as 1.42 for MA-PDs and 1.31 for PDPs. For MA-PDs, the top 5 HRRs with the highest HRM use rates in 2013 were Casper, WY (20.4%), Waco, TX (16.7%), Lubbock, TX (15.7%), Santa Barbara, CA (15.2%), and Temple, TX (15.1%); for PDPs, they were Lawton, OK (18.8%), Alexandria, LA (18.8%), Lake Charles, LA (18.6%), Oklahoma City, OK (18.0%), and Slidell, LA (18.0%). CONCLUSIONS: Substantial geographic variation exists in the prevalence of HRM use among older adults in Medicare, regardless of prescription drug plan. Areas with high prevalence of HRM use may benefit from targeted interventions (e.g., medication therapy management monitoring or alternative medication substitutions) to prevent potential adverse consequences.


R E S E A R C H
I nappropriate medication use is more common among older adults (varying from 12.3% to 42.6% based on different definitions) compared with younger individuals because of aging along with multiple comorbid conditions. [1][2][3][4][5][6][7] Inappropriate medication use may contribute to changes and reduction in hepatic metabolism and renal clearance among older adults and thus enhance drug toxicity. 8 Inappropriate medication use related to adverse health outcomes causes a substantial clinical and economic burden, with an estimated annual cost of $7.2 billion in the United States. [9][10][11] Several organizations and stakeholders have developed guidelines and policies to prevent inappropriate medication use among older adults. For example, the Beers Criteria, first • Older adults have a higher risk of adverse drug events than younger adults because of age-related drug metabolism changes and their multiple comorbidities. • Adverse drug events because of high-risk medication (HRM) use, such as falls with benzodiazepine use, among older adults are associated with a substantial clinical and economic burden in the United States. • Previous reports suggested geographic variation of HRM use, based on the Healthcare Effectiveness Data and Information Set (HEDIS), in beneficiaries enrolled in stand-alone Prescription Drug Plans (PDPs).

What is already known about this subject
• This study was the first to examine geographic variation of HRMs prescribed and dispensed patterns for Medicare Advantage Prescription Drug (MA-PD) plans and stand-alone PDPs using the Pharmacy Quality Alliance's HRM measure, which is included in the Centers for Medicare & Medicaid Services' star ratings. • Our findings showed that substantial geographic variation in HRM use exists among Medicare Part D enrollees regardless of Part D plan types and areas at high-risk of HRM use differed between MA-PD plans and PDPs. • Plan sponsors and health care systems may strategize and prioritize targeted interventions for the areas with high HRM use prevalence to prevent potential adverse health outcomes.
The objective of this study was to examine HRM use in MA-PD and PDP beneficiaries across HRRs using the PQA measure from 2011 to 2013, adjusting for demographic, health status, and access to care factors. Understanding geographic patterns of HRM use among older adults by type of Part D plans may help CMS and their partners develop and tailor prevention strategies.

■■ Methods Study Design and Data Sources
This retrospective cross-sectional study used Medicare Master Beneficiary Summary Files (MBSF) and Part D Prescription Drug Event (PDE) files for a 5% nationally representative sample of Medicare beneficiaries from 2011 to 2013 (~3.2 million unique beneficiaries). MBSF included sociodemographic and eligibility information. 27 PDE files contained information regarding the prescriptions filled and reimbursed by Part D plans including medication names, National Drug Code (NDC) numbers, quantity, prescription fill date, days of supply, and prescriber's national provider identifier. 28 We obtained regional and access-to-care factors from the publicly available Area Health Resource File (AHRF) for the year 2013-2014 through the Medicare beneficiaries' social administration codes. AHRF includes more than 6,000 regional variables regarding health facilities, health professions, measures of resource scarcity, health status, economic activity, training programs for health care professionals, and socioeconomic and environmental characteristics. Variables from AHRF data included in this study were rural/urban continuum (metropolitan vs. nonmetropolitan), primary care shortage area (partial, full, or none), and mental health care shortage area (partial, full, or none). The University of Arizona Institutional Review Board approved this study. The reporting of this study complied with the STROBE guidelines. 33

Study Sample Each Year
Supplementary Figure 1 (available in online article) outlines the selection of the study sample for each calendar year (2011,2012,2013). Beneficiaries aged ≥ 65 years who were continuously enrolled in either a Medicare Part D MA-PD plan or a PDP for the entire 12 months were included in the final study sample. Beneficiaries who were not U.S. residents, died, had end-stage renal disease, did not fill a prescription, or switched between MA-PD plans and PDPs during the year were excluded.

Dependent Variable: High-Risk Medication Use
Using the PQA's HRM measure released in 2013, we first identified claims containing HRMs across 9 therapeutic classes, including anticholinergics (excluding tricyclic antidepressants), anti-infectives, anti-thrombotics, cardiovascular medications, central nervous system medications, endocrine medications, published in 1991, [12][13][14] has been widely adopted by health care organizations to reduce potentially inappropriate medication use. In 2013, the Pharmacy Quality Alliance (PQA) adapted subsections of the Beers Criteria and developed a measure to identify older adults with high-risk medications (HRM) use when alternative medications were available. [15][16][17] In 2015, the Centers for Medicare & Medicaid Services (CMS) included the PQA's HRM measure in the star rating system for Medicare Part D plans. [18][19][20] Meanwhile, health plans also sought to decrease HRM use through different strategies such as prior authorization, formulary selection, educational mailings to prescribers, and performing comprehensive medication reviews as part of medication therapy management (MTM) services. 18,21 These approaches may be effective to some extent; however, strategies to more accurately and effectively identify high-risk individuals or areas are needed to better guide these interventions.
Previously, CMS has reported geographic variation in HRM use at the state level in the whole Medicare Part D population. 22 However, use of health services and prescription medications varies substantially within the same state, thus, hospital referral regions (HRRs) may better reflect health services resources and utilization in a particular geographical area. HRRs are commonly used regional units developed by the Dartmouth Atlas of Health Care to examine geographic variation in a variety of topics such as health care spending and chronic disease management, especially in Medicare services. [23][24][25] HRRs are intended to approximate regional health care markets for tertiary medical care. 23 Each HRR contains at least 1 hospital that performs major cardiovascular procedures and neurosurgery. 23 Previously, Zhang et al. (2010) evaluated HRM use among stand-alone Prescription Drug Plan (PDP) beneficiaries at the HRR level, using the Health Effectiveness Data Information Set (HEDIS) measure. [26][27][28] However, the HEDIS HRM list was not as thorough as the PQA HRM list. At the time of this study, the HEDIS HRM only required Medicare beneficiaries to have received 1 HRM, 29 whereas the PQA HRM measure required beneficiaries to have 2 or more prescription fills for the same HRM during the measurement period. 30 The PQA measure therefore offers an opportunity for a more conservative/targeted analysis than using the HEDIS measure.
Furthermore, over 30% of Medicare Part D beneficiaries are enrolled in Medicare Advantage Prescription Drug (MA-PD) plans. There are important differences between the structure of MA-PD plans and PDPs, in particular MA-PD plans have greater incentives to lower health care costs than PDPs through the provision of quality bonus payments, so MA-PD plans may use more aggressive strategies to monitor HRM use. 31 Thus, it is important to assess the magnitude of HRM use in MA-PD plans and PDPs. 32

Geographic Variation in the Prevalence of High-Risk Medication Use Among Medicare Part D Beneficiaries by Hospital Referral Region
gastrointestinal medications, pain medications, and skeletal and muscle relaxants (Supplementary Table 1, available in online article). Beneficiaries who had at least 2 prescription fills for the same HRM during the year met the PQA's HRM use criteria (i.e., HRM users). Using beneficiary-level data, we calculated crude prevalence as the proportion of beneficiaries with HRM use among beneficiaries in MA-PD plans versus PDPs each year. 34

Independent Variable: Hospital Referral Regions
The independent variable was the 306 Dartmouth Atlas of Health Care HRRs, which were aggregated from 3,436 hospital service areas based on where the largest proportion of major cardiovascular and neurosurgery procedures were performed. Each HRR has at least 1 town or city where major cardiovascular surgical procedures and neurosurgery are performed. 35 For each year, we assigned each beneficiary to an HRR based on their residential ZIP code that was obtained from the MBSF of the year. 36

Covariates
Sociodemographic characteristics included age, sex, race/ ethnicity (White, Black, Hispanic, and Others), having lowincome subsidy (LIS) or dual eligibility (DE) for Medicaid, with disability status based on the original reason for Medicare entitlement. Health status factors included beneficiaries' comorbidities measured by the modified RxRisk-V score, and numbers of unique pharmacies and unique providers visited during the year.
Although many calculations of comorbidity scores require using medical claims data, it is worth noting that the HRM measure included in the CMS star rating system was calculated based on prescription drug claims alone because some health plans only have access to prescription drug claims data. Therefore, we used prescription drug claims alone to calculate the modified RxRisk-V score (excluding HRM medications) as a proxy of beneficiary comorbidity complexity that is applicable to all Part D health plans. 37,38 The RxRisk-V, developed as an all-age risk adjustment tool, is a prescription-based risk adjustment score that uses ambulatory pharmacy data to identify 42 chronic disease categories. RxRisk-V has been used to predict total health care costs and mortality. 37,[39][40][41][42] A previous study showed that the RxRisk-V model performed similarly to the Ambulatory Clinical Group, a diagnostic-based model. 42

Statistical Analysis
The annual crude prevalence of HRM use was calculated by dividing the number of beneficiaries with at least 2 prescription fills for the same HRM by the total number of beneficiaries for each HRR each year by MA-PD plan versus PDP.
Our analytical approaches consisted of 2 steps. Separate models were created for MA-PD and PDP beneficiaries. First, multivariable logistic regression was used to estimate adjusted annual prevalence of HRM use (also called average marginal predictions or model-adjusted risk) for each HRR, controlling for demographics, health status, and regional/access-to-case factors. Second, HRRs were ranked based on the adjusted prevalence of HRM use from highest to lowest. We calculated the ratios of 75th to 25th percentile of adjusted HRM use rates; a ratio greater than 1 suggests geographic variation. 43 All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). Sociodemographic, health status, and regional or accessto-care factors were similar across years. The 2013 data are presented in Table 1

■■ Discussion
This study found that prevalence of HRM use, as defined by the CMS/PQA, significantly decreased by over 25% from 2011 to 2013 in MA-PD and PDP beneficiaries aged ≥ 65 years, although it still remains common (approximately 9% in MA-PD plans and 12% in PDPs annually). In addition, our findings suggested disparities because of geographic differences remain a concern regardless of health plan type or year. After adjusting for sociodemographic, health status, and regional or access-tocare factors, HRR-level geographic variation persisted across 3 years. The majority of areas with high HRM prevalence for MA-PD and PDP beneficiaries were located in the southern United States.
Consistent with previous studies, [44][45][46] our findings showed that the PDP beneficiaries with higher HRM prevalence were concentrated in the southern regions of the United States, while MA-PD beneficiaries with higher HRM prevalence clustered in the southern and western regions. Multifaceted factors may have contributed to the observed geographic variations in the prevalence of HRM use, especially in southern regions. Residents in the southern regions of the United States were more likely to be Black or Hispanic, dual-eligible for Medicaid, and to have higher prevalence of multiple chronic diseases and polypharmacy.
A recent report found that the areas with highest priced and health-risk adjusted Medicare spending per capita were mainly located in the southern regions. 47 In addition, these areas tended to have fewer physicians available for access for care per resident and a larger need of hospital beds, surgical centers, and post-acute health care professionals. 47 Another study identified

Percentage of Beneficiaries with Use of at Least 2 High-Risk Medications Across Hospital Referral Regions: MA-PD Beneficiaries, 2011-2013
Note: The high-risk medication use rates were adjusted for age, sex, race, Part D low-income subsidy/dual eligibility status, disability status, regional characteristics (e.g., rural/urban continuum and access-to-care factors), and clinical complexity (e.g., modified RxRisk index).

Limitations
Several limitations of this study are noteworthy to discuss. First, as with any study using claims data, we were not able to measure sociobehavioral information, the need or clinical rationale for HRM use for some beneficiaries, and physician and patient preferences in our study. Similarly, unmeasured confounders cannot be ruled out. Second, this study relied on the assumption that dispensed medications were consumed as prescribed. Third, regions with small sample sizes of Medicare beneficiaries may lead to unstable estimates of HRM use. Fourth, the clinical complexity may have been underestimated because the RxRisk score was calculated based on prescription claims data alone. For example, some patients with diabetes may only need lifestyle modifications instead of medication treatment. Finally, our study used data from 2011 to 2013 that we had access to. The study findings may not be generalizable to more recent years.

■■ Conclusions
Substantial geographic variation in HRM use exists among older adults in Medicare, regardless of type of prescription that the characteristics of hospitals with low service quality and high costs included those that were small and located in the southern regions compared with the hospitals with highquality services and low cost. 48 Other factors contributing to the geographic variation in the HRM use included variations in physician-prescribing patterns, health plan formularies, and patient and prescriber preference. 49 Future research is warranted for further investigation of the interaction between different Part D plan types, area characteristics, and patient health outcomes.
Identifying regions with higher prevalence of HRM use may better guide the current intervention programs such as pharmacist-provided MTM services to focus on these regions. MTM typically constitutes a comprehensive medication review (CMR) to identify medication-related problems, including HRM use among older adults. 50 Several studies have demonstrated the value of MTM services in improving health outcomes in a variety of chronic conditions and settings. [51][52][53] Although 1 study found that MA-PD providers with high CMR completion rates had lower HRM use among their beneficiaries, the association between CMR completion and HRM use was not observed among PDP beneficiaries. 21 Leveraging community efforts to promote education related to the risk of HRM use in these high-risk regions may also reduce HRM use and associated consequences.
The strengths of this study included the following: (a) use of the PQA/CMS definition for HRM use that is currently