Association of Medicare Part D Benzodiazepine Coverage Expansion With Changes in Fall-Related Injuries and Overdoses Among Medicare Advantage Beneficiaries

Key Points español 中文 (chinese)

Question  Was the 2013 expansion of US Medicare Part D prescription drug coverage to include benzodiazepines associated with increased rates of fall-related injuries or overdoses among older adults?

Findings  This ecological study of more than 4.6 million Medicare Advantage beneficiaries found an increase in the rate of overdose after Part D coverage began to include benzodiazepines among adults aged 65 to 69 years and 80 years or older and an increase in the rate of fall-related injury in adults 80 years or older.

Meaning  Medicare’s expansion of benzodiazepine coverage to older adults may have been associated with an increased rate of overdose among those 69 years or younger and 80 years or older and of injury in those 80 years or older.

Importance  Benzodiazepines, which are associated with safety-related harms for older adults, were not covered when the US Medicare Part D prescription drug benefit began. Coverage was extended to benzodiazepines in 2013.

Objective  To examine whether the expansion of benzodiazepine coverage among Medicare Advantage (MA) beneficiaries was associated with increases in fall-related injuries or overdoses among older adults.

Design, Setting, and Participants  This ecological study used interrupted time-series with comparison-series analyses of MA claims data from 4 635 312 age-eligible MA beneficiaries and 940 629 commercially insured individuals (comparison group) stratified by age (65-69, 70-74, 75-79, and ≥80 years) to separately compare trends in fall-related injury and overdose before (January 1, 2010, to December 31, 2012) and after (January 1, 2013, to December 31, 2015) coverage expansion for benzodiazepines. Data analysis was performed from September 1, 2018, to August 31, 2019.

Exposures  Expansion of benzodiazepine coverage in Medicare Part D in 2013.

Main Outcomes and Measures  Monthly rate of fall-related injury and overdose.

Results  In 2012 (the year before the policy change), women constituted 57.5% of the MA group and 47.4% of the comparison group. A total of 25.8% of individuals in the MA group were aged 65 to 69 years, and 29.3% were 80 years or older (mean [SD], 75.1 [6.4] years); 56.7% of individuals in the comparison group were aged 65 to 69 years, and 15.1% were 80 years or older (mean [SD] age, 70.9 [6.5] years). In the MA group, 4 635 312 individuals contributed 156 754 749 person-months from 2010 through 2015; in the comparison group, 940 629 individuals contributed 25 104 534 person-months. After coverage of benzodiazepines began, the rate (ie, slope) of fall-related injury among MA beneficiaries increased from before to after coverage among all age groups. Compared with the comparison group, the increase in rate was statistically significant for those 80 years or older (rate changes for the MA vs comparison groups: 0.12 [95% CI, 0.07 to 0.17] vs −0.01 [95% CI, −0.11 to 0.10]; P = .04 for interaction). The overdose trend changed from decreasing to increasing among MA beneficiaries after coverage for all age groups, with a statistically significant increase compared with the comparison group among those aged 65 to 69 years (rate changes for the MA vs comparison groups: 0.23 [95% CI, 0.17 to 0.30] vs 0.02 [95% CI, −0.06 to 0.11]; P < .001 for interaction) and among those 80 years or older (rate changes for the MA vs comparison groups: 0.07 [95% CI, 0.00 to 0.14] vs −0.20 [95% CI, −0.35 to −0.05]; P = .002 for interaction). Results among MA beneficiaries were consistent when stratified by sex and when limited to those prescribed opioids.

Conclusions and Relevance  Medicare’s expansion of benzodiazepine coverage may have been associated with increases in the rates of overdose among adults ages 65 to 69 years and in the rates of overdose and fall-related injury among those 80 years or older.

Benzodiazepines are associated with a number of safety concerns among older adults, including increased risk of falls and hip fracture.^1,2 Although overdose receives less attention as a benzodiazepine-related adverse event, benzodiazepines are the second-most common medication class involved in pharmaceutical overdose deaths,³ and overdose deaths that involve benzodiazepines increased more than 6-fold from 1996 through 2014.⁴ More than 75% of benzodiazepine-related deaths involve opioids,³ and evidence continues to accumulate that use of benzodiazepines is associated with increased risk of opioid-related overdose and mortality.^5-8 Given that older adults experienced the largest absolute increases in opioid-related mortality between 2001 and 2016⁹ and also experience the highest rates of coprescribing of opioids and benzodiazepines,^10,11 benzodiazepine prescribing may be associated with increased opioid-related morbidity and mortality among older adults.

When the Medicare Part D prescription drug coverage program began in 2006, it did not include coverage for benzodiazepines, even though the class is widely prescribed to older adults.¹² The exclusion was effectively a legacy policy from Medicaid that allowed states to exclude 10 categories of medications from prescription drug coverage, including benzodiazepines.¹³ After Medicare Part D began, there was concern that the lack of benzodiazepine coverage was unduly limiting access to the widely prescribed medication class and causing a financial burden for those who paid out of pocket.^14,15 In response to these concerns, benzodiazepine coverage was added to Medicare Part D in 2013.¹⁶ However, lowering the barrier to use of benzodiazepines, which were considered inappropriate for most older adults at the time of the policy change,¹⁷ may have been associated with safety consequences for the millions of older US individuals enrolled in Medicare Part D plans.

In states that have implemented policies that limit benzodiazepine prescribing, fall-related adverse events among older adults have not decreased.^18,19 Although restricting benzodiazepine access has not been associated with injury reduction among older adults, less is known about the consequences associated with abruptly expanding benzodiazepine coverage; lowering the threshold for new prescriptions may have a different association with adverse outcomes than restricting coverage. Furthermore, no studies to our knowledge have examined the association of benzodiazepine coverage with overdose among older adults. In this analysis, we examined the association of fall-related injury and overdose with Medicare’s expansion of Part D coverage to include benzodiazepines. Given heightened public attention to opioid prescribing and the potential for harms when opioids and benzodiazepines are coprescribed,^5,8,20 we also examined outcomes limited to older adults prescribed opioids. We hypothesized that this nationwide expansion of benzodiazepine coverage would be associated with an increase in the rates of fall-related injury and overdose both overall and among those prescribed opioids.

For this ecological study, we used administrative health claims from Optum’s deidentified Clinformatics Data Mart Database derived from a large national health insurance company in the US. We created monthly measures of the outcomes using a denominator of all age-eligible (ie, ≥65 years of age) Medicare Advantage (MA) beneficiaries who were enrolled during that month. All older adults in the population had Medicare Part D prescription drug coverage as part of their MA plan. There were no other exclusion criteria. As a comparison group, we used adults 65 years or older in the database with commercial (ie, employer-sponsored) insurance; the prescription plans of these older adults would not have been subject to the benzodiazepine coverage change experienced by MA beneficiaries. The population-level numerators for the 2 analyses were the number of fall-related injury episodes that began in a given month and the number of overdoses. This study was considered not regulated by the Michigan Medicine Institutional Review Board, so informed consent was waived for this analysis of deidentified data. This study followed the Reporting of Studies Conducted Using Observational Routinely Collected Health Data (RECORD) extension of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.²¹

We conducted separate analyses for the 2 primary outcomes of interest: fall-related injury rates and overdose rates. We also examined changes in benzodiazepine prescribing after the policy change as an intermediary mechanism for the hypothesized changes in our outcomes of interest.

Episodes of fall-related injury were determined using an algorithm developed and validated using MA claims data.²² The algorithm captures single episodes of injury (ie, an ED visit, inpatient stay, and rehabilitation visits that followed a single fall event were counted together as a single episode) across the spectrum of severity from nonfracture injury to hip fracture. If a single injury episode crossed calendar months, it was attributed to the month when the episode began.

Overdoses were captured using International Classification of Diseases, Ninth Revision, Clinical Modification diagnoses from emergency department or acute inpatient encounters regardless of intent (codes 960-969 and E850-E858 [unintentional], E930-E949 [adverse drug effect], E950.0-E950.5 [self harm], and E980.0-E980.5 [undetermined] based on consensus recommendations from the Injury Surveillance Workgroup²³).

We obtained information on patient sex and age along with select additional demographic characteristics (eg, US Census division and race/ethnicity) from demographic data available in the Optum database. For descriptive purposes, we also determined the presence of select clinical conditions for which benzodiazepines may commonly be prescribed in clinical practice, including anxiety disorders, depression, dementia, and insomnia (eTable 1 in the Supplement). We stratified these descriptions by insurer type and precoverage and postcoverage period.

We conducted interrupted time-series analyses to compare the rates of the 2 injury types in the population before and after the Medicare Part D coverage expansion to include benzodiazepines in 2013. For all analyses, the precoverage period was January 1, 2010, to December 31, 2012; the postcoverage period was January 1, 2013, to December 31, 2015. Our models used monthly injury rates as the dependent variable and included a linear time-trend variable, an indicator for the postcoverage period (ie, an indicator for before and after Medicare policy change), and a term for the change in the linear time trend (ie, slope) between the precoverage and postcoverage periods. We did not include a term that tested for an immediate-level shift in outcomes (ie, a change in the model intercept as of the beginning of the postcoverage period in January 2013) because we hypothesized that any change in benzodiazepine prescribing and associated injuries would accumulate over time and not change immediately on January 1, 2013. We constructed analogous models in the comparison group (ie, commercial insurance beneficiaries). To test for the association of the Medicare benzodiazepine coverage policy expansion (ie, a difference in postcoverage rate change between MA beneficiaries and the comparison group), we fit a single model with interactions between group and each term in the interrupted time-series model, specifically reporting the significance of the interaction between groups (MA vs comparison) and the postcoverage rate shift.

Descriptive analyses revealed secular trends in the age distribution of the study population. To guard against confounding on this basis, we stratified analyses by age (65-69, 70-74, 75-79, and ≥80 years). Results of the primary analyses for both outcomes were similar between men and women within each age group; thus, we present primary findings stratified only by age. Models also included monthly fixed effects to account for seasonality of the outcomes. We did not detect any other temporal autocorrelation in the model errors. We used 2-tailed t tests to test the statistical significance (α = .05) of the association of benzodiazepine coverage expansion with injury (ie, the rate change between precoverage and postcoverage periods). In addition, given safety concerns about opioid and benzodiazepine coprescribing,²⁴ we completed additional analyses that limited the monthly MA population denominator to beneficiaries with 1 day or more of opioid supply in a given month to determine whether results limited to the population prescribed opioids were consistent with those for the overall MA population. Analysis was completed from September 1, 2018, to August 31, 2019.

In 2012 (the year before the policy change), women constituted 57.5% of the MA group and 47.4% of the comparison group. A total of 25.8% of individuals in the MA group were aged 65 to 69 years, and 29.3% were 80 years or older (mean [SD] age, 75.1 [6.4] years); 56.7% of individuals in the comparison group were aged 65 to 69 years, and 15.1% were 80 years or older (mean [SD] age, 70.9 [6.5] years). There were 4 635 312 individual patients (ie, age-eligible MA beneficiaries), who contributed 156 754 749 person-months from 2010 through 2015. In the comparison group, there were 940 629 individual patients, who contributed 25 104 534 person-months during the same period. Select population characteristics are given in Table 1. There was a marked increase in benzodiazepine claims among MA beneficiaries beginning in 2013; beneficiaries with 1 day or more of benzodiazepine coverage in a given month increased from approximately 0.5% in 2012 to 6% for much of 2013 (eFigure 1 and eFigure 2 in the Supplement).

The baseline rate (ie, the model-predicted rate in January 2010) of fall-related injury was 1.34 per 10 000 person-days (95% CI, 1.31-1.37 per 10 000 person-days) among those aged 65 to 69 years and increased for each age group; the rate was 3.20 per 10 000 person-days (95% CI, 3.15-3.24 per 10 000 person-days) among those 80 years or older (Table 2). Monthly event rates were similar among the comparison group. During the postcoverage period, the rate of fall-related injury increased for all MA age groups (Figure 1). Interaction tests for the rate change between the MA and comparison groups were not significant for the younger age groups (eg, 65-69, 70-74, and 75-79 years). However, for those 80 years or older, the increase in the rate of fall-related injury increased significantly among the MA group compared with the comparison group during the postcoverage period (rate changes for the MA vs comparison groups: 0.12 [95% CI, 0.07 to 0.17] vs −0.01 [95% CI, −0.11 to 0.10]; P = .04 for interaction). The MA analyses stratified by sex yielded similar results for men and women by age group (eTable 2 in the Supplement).

Among MA beneficiaries prescribed opioids, the fall rate was markedly higher than in the overall population; however, the precoverage and postcoverage changes were consistent (eTable 3 in the Supplement). Similar to the main analysis, the rate of fall-related injury changed after coverage; the precoverage rate was either negative or flat and shifted to an increasing trend in all age groups.

The baseline rate (ie, the model-predicted rate in January 2010) of overdose was 1.31 per 100 000 person-days (95% CI, 1.22-1.40 per 100 000 person-days) among those aged 65 to 69 years. The rate was highest among those aged 75 to 79 years (1.61 per 10 000 person-days; 95% CI, 1.52-1.70 per 100 000 person-days) (Table 3). The overdose rate was lower among the comparison group for all ages. During the postcoverage period (2013-2015), the rate of overdose among MA beneficiaries increased for all age groups (Figure 2). Compared with the comparison group, the increases in postcoverage overdose rates were statistically significant among those 65 to 69 years old and among those 80 years or older. For those aged 65 to 69 years, the postcoverage rate increased by 0.23 encounters per month (95% CI, 0.17-0.30) vs 0.02 (95% CI, −0.06 to 0.11; P < .001 for interaction) for the comparison group. For those 80 years or older, the postcoverage rate increased by 0.07 encounters per month (95% CI, 0.00-0.14 encounters per month) vs −0.20 (95% CI, −0.35 to −0.05; P = .002 for interaction) for the comparison group. The MA analyses stratified by sex yielded similar results for men and women by age group (eTable 4 in the Supplement).

As with the fall rate, the monthly rate of overdose was markedly higher among beneficiaries prescribed opioids than in the overall population (eTable 5 in the Supplement). Similar to the main analysis, the monthly event rate reversed direction from decreasing before coverage to increasing after coverage for those aged 65 to 69 years and 75 to 79 years, with a nonsignificant change among those aged 70 to 74 years.

In this analysis of 4.6 million MA beneficiaries, we found that, at the population level, expansion of Medicare Part D coverage to include benzodiazepines was associated with increases in benzodiazepine prescription claims for all adults 65 years or older. Our findings suggest that there may have been associated increases in the rate of injury among those aged 65 to 69 years and those 80 years or older. The oldest group experienced increasing rates of both fall-related injury and overdose, whereas among those aged 65 to 69 years, the trend of overdose increased. Findings were not different when the population was limited to older adults who were prescribed opioids.

Although the increase in benzodiazepine claims among MA beneficiaries was increased by 5.5% or more than 10-fold, there were some benzodiazepine claims during the precoverage period (ie, 2010-2012). This finding likely reflects that some MA plans could provide more generous prescription coverage than required by Medicare that included benzodiazepines.²⁵ The increase in claims for benzodiazepines likely overestimates the true increase in exposure among older adults because benzodiazepines are available in generic, inexpensive forms; some patients were likely self-paying for benzodiazepine prescriptions before the 2013 coverage change. In an analysis of MA beneficiaries in plans that included prescription drug coverage (including benzodiazepines) before Medicare Part D existed but then excluded benzodiazepines when Medicare Part D began, nearly 75% of MA beneficiaries continued use of benzodiazepines after exclusion by paying out of pocket.¹⁵ In addition, a recent analysis⁴ of benzodiazepine prescribing trends from 1996 to 2014 using the Medical Expenditure Panel Survey did not demonstrate an increase in prescription benzodiazepine use among respondents 65 years or older in 2013.⁴

We hypothesized that the expansion of coverage would be associated with both types of injury. However, the minimal effect sizes observed align with prior studies^18,19 of policies that restricted benzodiazepine access that did not demonstrate a decrease in injury. One such analysis¹⁹ that predated Medicare Part D examined hip fracture rates, comparing New York, which instituted a triplicate prescription policy that increased the administrative burden of benzodiazepine prescribing, with New Jersey, which had no such policy. Although there was an immediate 60% reduction in benzodiazepine prescribing in New York, the rate of hip fracture did not decrease compared with that in New Jersey. When Medicare Part D went into effect, some state Medicaid programs provided supplemental coverage of benzodiazepines, whereas other states did not. In an analysis¹⁸ of dually eligible nursing home residents, benzodiazepine prescribing varied with states’ supplemental coverage, but the risk of hip fracture was not lower in states that limited benzodiazepine coverage.¹⁸

Given these prior analyses, what may account for the limited associations with increased injury in the present analysis? The expansion of Medicare Part D may have had a limited association with overall prevalence of benzodiazepine use among older adults but still would have been associated with a reduced threshold for new use among benzodiazepines-naive patients, particularly those with limited resources. Because those who regularly use benzodiazepines develop physiologic tolerance with repeated exposure,²⁶ new users may be particularly susceptible to experiencing an injury when first exposed. A study by Hernandez et al²⁷ examining the risk of opioid-related overdose among concurrent opioid and benzodiazepine users suggests that the risk of overdose is highest in the first 90 days of concurrent use and then decreases with each successive time increment.²⁷ In addition, some practitioners and patients may have interpreted Medicare’s benzodiazepine coverage expansion as an indication that safety risks had been overstated, further lowering the prescribing threshold and leading to prescriptions that practitioners previously avoided because of safety concerns.

We did not anticipate that the change in the overdose rate associated with benzodiazepine coverage would be greater in relative terms than the change in fall-related injury. Safety warnings cautioning against benzodiazepine prescribing in older adults typically focus on the risk of falls.²⁸ Their association with overdose other than when prescribed with opioids receives less attention, and unlike for fall-related injury, we are unaware of prior analyses examining the association of benzodiazepine coverage policies with overdose. This lack of attention may be partially because overdose is less common among adults 65 years or older than among younger adults²⁹ and is therefore not perceived as a problem. However, a recent analysis⁹ found that the largest relative increases in opioid-related overdose from 2001 to 2016 were among adults aged 55 to 64 years followed by those 65 years or older.

Why might the associations with increased injury have been limited to those aged 65 to 69 years and those 80 years or older in the study population? It is possible that these age groups were vulnerable to the effects of policy change for different reasons. Adverse effects related to benzodiazepines are generally attributed to the role of benzodiazepine as a central nervous system depressant, and older adults are particularly sensitive to these effects.^30,31 An 80-year-old individual may be more sensitive to these effects than a 65-year-old individual. The 65- to 69-year age group includes members of the baby boom cohort, who started turning 65 years of age in 2011. Compared with prior cohorts of older adults (ie, the other age groups in this analysis), baby boomers have higher rates of alcohol use, nonmedical prescription opioid use, and marijuana use,^32-35 each of which might be associated with increased susceptibility to benzodiazepine-related injury from a new prescription. By occurring at the height of the opioid overdose crisis, the benzodiazepine coverage expansion potentially exacerbated the risks associated with prescription opioids.

We did not expect the overdose rate to have decreased during the precoverage period; across the US overall, overdose-related deaths among those 65 years or older increased slightly during that time.²⁹ Although favorable selection into MA plans has decreased,^36,37 evidence remains that patients with the most complex needs may exit MA,^38,39 and MA patients at the end of life are healthier, more educated, and have fewer functional limitations.⁴⁰ The rate also decreased among the comparison group except among those 80 years or older. The population of adults 65 years or older who continue to maintain employment with employer-sponsored insurance coverage may be healthier and are likely not representative of the general older adult population.

This study has limitations. First, the study was limited to MA beneficiaries and not the overall population with Medicare Part D coverage; thus, our findings may not generalize to the entire population of older adults. Second, claims data have limited ability to detect and appropriately classify both of the primary outcomes, which thus may have been limited to events serious enough to require medical attention and to those appropriately coded by practitioners. Third, we cannot account for the true prevalence of benzodiazepine use in the population because patients were likely paying for prescriptions out of pocket or receiving coverage through other sources (eg, Veterans Affairs or Medicaid). Fourth, we did not adjust for patient-level characteristics that may have been changing during this time and potentially contributed to increased falls or overdoses, although our findings remained when restricted to those prescribed opioids.

Compared with commercial coverage among similarly aged older adults, Medicare’s expansion of benzodiazepine coverage to older adults may have been associated with an increased rate of overdose among those 69 years or younger and 80 years or older and of injury in those 80 years or older. To our knowledge, this was the first analysis to consider the association of benzodiazepine coverage policy with the risk of overdose.

Accepted for Publication: February 2, 2020.

Published: April 3, 2020. doi:10.1001/jamanetworkopen.2020.2051

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Maust DT et al. JAMA Network Open.

Corresponding Author: Donovan T. Maust, MD, MS, Department of Psychiatry, University of Michigan, 2800 Plymouth Rd, NCRC 016-226W, Ann Arbor, MI 48109 (maustd@umich.edu).

Author Contributions: Dr Maust 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.

Concept and design: Maust, Lin, Goldstick, Haffajee, Bohnert.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Maust, Goldstick, Brownlee.

Critical revision of the manuscript for important intellectual content: Maust, Lin, Goldstick, Haffajee, Bohnert.

Statistical analysis: Goldstick.

Obtained funding: Maust, Bohnert.

Administrative, technical, or material support: Maust, Lin.

Supervision: Maust.

Conflict of Interest Disclosures: Dr Lin reported receiving grants from the Veterans Affairs’ Health Services Research and Development Service during the conduct of the study. Dr Bohnert reported receiving grants from the Centers for Disease Control and Prevention during the conduct of the study and grants from Michigan Department of Health and Human Services and the National Institutes of Health outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by grant R49 CE002099-05S1 from the Centers for Disease Control and Prevention.

Role of the Funder/Sponsor: The funding source 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 decision to submit the manuscript for publication.

Additional Contributions: Matthew Myers, MPH, Injury Prevention Center, University of Michigan, Ann Arbor, assisted with manuscript revision. Mr Myers was not compensated for this work.