Volume 63, Issue 3 p. 486-500
Clinical Investigations
Free Access

Prevalence of Potentially Inappropriate Medication Use in Older Adults Using the 2012 Beers Criteria

Amy J. Davidoff PhD

Corresponding Author

Amy J. Davidoff PhD

School of Public Health, Yale University, New Haven, Connecticut

Address correspondence to Amy J. Davidoff, Yale University School of Public Health, P.O. Box 208034, 60 College Street, New Haven, CT. E-mail: [email protected]Search for more papers by this author
G. Edward Miller PhD

G. Edward Miller PhD

Agency for Healthcare Research and Quality, Rockville, Maryland

Search for more papers by this author
Eric M. Sarpong PhD

Eric M. Sarpong PhD

Agency for Healthcare Research and Quality, Rockville, Maryland

Search for more papers by this author
Eunice Yang MA

Eunice Yang MA

School of Public Health, University of North Carolina, Chapel Hill, North Carolina

Search for more papers by this author
Nicole Brandt PharmD, MBA

Nicole Brandt PharmD, MBA

Geriatric Pharmacotherapy, Pharmacy Practice and Science, School of Pharmacy, University of Maryland Baltimore, Baltimore, Maryland

Search for more papers by this author
Donna M. Fick PhD, RN

Donna M. Fick PhD, RN

Department of Psychiatry, College of Nursing and College of Medicine, Pennsylvania State University, University Park, Pennsylvania

Search for more papers by this author
First published: 06 March 2015
Citations: 96

Abstract

Objectives

To use the most recently available population-based data to estimate potentially inappropriate medication (PIM) prevalence under the 2012 update of the Beers list of PIMs and to provide a benchmark from which to measure future changes.

Design

Retrospective cohort study using nationally representative data.

Setting

2006–2010 Medical Expenditure Panel Survey (MEPS).

Participants

Community-dwelling sample of U.S. older adults (N = 18,475).

Measurements

The updated Beers criteria were operationalized, generating a “broad” PIM definition that incorporated form, route, or dose restrictions where clearly specified and a “qualified” definition that applied specific exceptions where mentioned in the rationale associated with each drug category. Bivariate analyses described PIM prevalence, comparing the two operational definitions, and examined time trends.

Results

Of older adults with prescription medications, 42.6% had at least one medication fill that met the broad definition, with nonsteroidal anti-inflammatory drugs (NSAIDs) having the highest prevalence (10.9%). The rate declined from 45.5% in 2006–2007 to 40.8% in 2009–2010. The categories with the largest absolute decline were NSAIDs, selected sulfonylureas, and estrogens. PIM prevalence was 30.9% using the qualified definition.

Conclusion

Despite the overall high use of PIMs, there has been a decline observed in recent years. Future studies should test the effect of educational and clinical interventions on changes in PIM use and outcomes. The current study results can aid in targeting these interventions.

The Beers list of potentially inappropriate medications (PIMs) is an important indicator of medication prescribing quality and an important educational tool for clinicians. Initially developed in 1991 with a focus on medication use in nursing home residents,1 the list was updated and expanded to include all geriatric care settings in 1997 and again in 2003.2, 3 In 2012, an expert panel was convened in collaboration with the American Geriatrics Society to update the Beers criteria.4 The panel was charged with updating the Beers list and rating the quality of evidence that supported the panel's recommendations. To accomplish this, the panel systematically reviewed the literature, entertained public comment, and graded the published evidence during an open period, according to the Institute of Medicine standards. This approach ensured transparency and rigor. A modified Delphi method was used to achieve consensus on the panel's recommendations.

PIMs continue to be prescribed to older adults, despite evidence of poor outcomes and adverse events.5-9 Prevalence of PIMs has been assessed in a variety of studies, in different settings and subpopulations, including several U.S. population-based. For example, estimates based on the 1996 Medical Expenditure Panel Survey (MEPS), indicate that more than one-fifth of community-dwelling adults aged 65 and older received at least one of 33 PIMs.10 PIM exposure was more common in older adults with poor health and more prescriptions. Trends from 1987 to 1996 indicated a reduction in PIM use for some drug groups. A decrease in PIM exposures and a higher risk of PIM exposures associated with poor health was confirmed in a study using the 1995 and 1999 Medicare Current Beneficiary Survey.11 A more recent study, using the 2007 MEPS and operationalizing the same 33 PIM recommendations from prior studies, found a dramatic decline in PIM exposure for older adults from 21.3% in 1996 to 13.8% in 2007.12

With ongoing changes in medications available in the market, changing indications, and a heightened emphasis on medication safety, PIM use is a moving target, and estimates need to be updated regularly. Prior estimates of PIM exposure reflect earlier time periods and earlier versions of the Beers criteria. Furthermore, most prior estimates of PIM exposure selected drugs from the Beers list that should be avoided in all or most cases but did not apply specific qualifying criteria such as dose, duration, and reason for use. To help fill the gap in the literature and to provide a benchmark from which to observe future changes, the most recently available MEPS data were used to update PIM prevalence estimates that incorporate recent changes to the Beers criteria. The current study illustrates how operationalizing additional details of the Beers panel's recommendations alters the estimates of PIM prevalence and provides a more clinically relevant estimate of PIMs. The new system of evidence ratings in the Beers criteria was also used to distinguish PIM exposures supported by stronger evidence from those supported by weaker evidence.

Methods

Data and Cohort

Data from the 2006–2010 MEPS, an ongoing overlapping panel survey sponsored by the Agency for Healthcare Research and Quality (AHRQ) that collects detailed and nationally representative information on healthcare use and expenditures, insurance coverage, sources of payment, health status, and sociodemographic variables for the U.S. civilian, noninstitutionalized population, were used.13, 14 Each year, a new panel of households is sampled and interviewed in five survey rounds over 2.5 years to obtain annual data reflecting a 2-year reference period. In each interview round, the MEPS collects information about new prescription fills. Respondents commonly use medicine bottles and receipts when providing this information, hence respondents tend to report chronic condition medication use accurately.15 Additional details about medications, including quantity or days supplied, are obtained from dispensing pharmacies. The MEPS prescribed medication files are linked to the Multum Lexicon database (Cerner Multum, Inc., Denver, CO) that facilitated identification of relevant drugs. The study also used the MEPS condition files and full-year consolidated files, which contain information on individuals’ demographic, socioeconomic, and health characteristics. The sample was limited to adults aged 65 and older.

Criteria Selection

The 2012 update of the Beers criteria identified 38 categories of drugs that older adults should avoid, as well as criteria specific to older adults with selected chronic conditions and a smaller set of criteria associated with medications that should be used with caution. The current study focused on the first set of criteria, because they are applied most broadly, and scrutinized the criteria to determine which could be operationalized with greatest reliability using self-reported prescription drug and medical condition data reported in the MEPS. Thirty-six of the 38 categories were selected, excluding insulin dosed on a sliding scale, because the MEPS did not provide a mechanism to distinguish a fixed from a flexible dosing schedule. Mineral oil, which it was expected would be purchased over the counter and thus poorly documented in the MEPS, was also excluded.

Operational Definitions of PIM Use

Generating the operational definitions of PIM use involved a three-step process. In Step 1, the updated Beers’ criteria were used to identify relevant medication names or therapeutic classes and the specific restrictions or exceptions related to dose, route, duration, and medical condition, as well as the ratings of evidence quality and recommendation strength. This information was merged with the prescription medication files according to drug name (including combination products) or therapeutic class so that each medication fill record had the relevant criteria to assess whether it qualified as a PIM. In Step 2, the data elements in the MEPS medication file were used to compute the parameters (e.g., dose) needed to assess whether the fill met the operational definitions for PIM use. The MEPS prescribed medication files include information on drug name, therapeutic class, dose form, route of administration, strength, quantity, and reasons for use (medical condition for which the drug was used). Medication fills from 2009 and 2010 also included days supplied, which were used to calculate daily quantity (fill quantity/days supplied). The information on daily quantity from fills in 2009–2010 was used to logically impute days supplied values for each drug and quantity combination in the 2006–2008 data. Information on quantity, strength, and days supplied was used to calculate daily dose (quantity × strength/days). Finally, days supplied for each drug were summed across fills for each person to obtain annual days supplied, which was used to measure therapy duration. The reasons for use were based on self-report, and cited conditions were coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). The medication fills were searched for specific conditions mentioned in the Beers criteria. In sensitivity analyses, information about conditions of interest reported for each respondent that were not reported to be the reason for using a specific drug were linked and applied. This broader group of conditions may have been reported as the reason for using other healthcare services or because they caused lost work, school, or bed days. Detailed information about the operational definitions for each drug category is provided in Appendix Table A1.

To develop an indicator that a fill met the criteria for a PIM (Step 3), information from the Beers criteria (Step 1) was compared with the medication and person-level information on each medication record developed during Step 2. Two operational definitions were developed. A “broad” definition that assigned PIM status based on use of a specified drug, applying only criteria related to form, route, or dose restrictions where clearly specified. A “qualified” definition applied selected exceptions mentioned in the rationale associated with each drug category. These exceptions usually related to requirements for a minimum duration of therapy or the presence of a medical condition, making the qualified definition more restrictive. Person-level PIM exposure measures were generated by summing the medication fill level PIM measures within each of the 36 drug categories and then generating an indicator of PIM use that cut across the 36 categories.

Analytical Approach

The number and proportion of prescription medication fills that met the definition for PIM use and the number and proportion of older adults with PIM use overall and according to drug category were quantified. To characterize the difference between the two estimates, the proportion of individuals whose PIM status under the qualified measure was affected by duration and condition restrictions is reported. Sampling weights were used to generate nationally representative, average annual estimates overall for 2006 to 2010. To assess changes in prevalence over time, the person-level estimates for the periods 2006–2007 and 2009–2010 were compared. Estimated standard errors and t-tests of the significance of changes over time accounted for the complex design of the MEPS. Analyses were conducted using SAS version 9.2 (SAS Institute, Inc., Cary, NC) and Stata 12 (Stata Corp., College Station, TX).

Results

The study sample included 18,475 person-years, accounting for an annual average of 39.58 million older adults and 35.93 million older adults with at least one prescription medication; 52.1% (standard error (SE) 0.81%) were aged 65 to 74, 79.6% (SE 0.8%) were non-Hispanic white, and 57% (SE 0.41%) were female. Detailed information about the characteristics of the sample is provided in Appendix Table A2.

Table 1 presents PIM prevalence estimates. Of older adults with prescription medication use, 15.3 million (42.6%) had at least one prescription medication fill that met the broad definition for a PIM (left side of table), accounting for 106.0 million PIM fills. The prevalence of PIM use according to drug category ranged from a negligible quantity (e.g., chloral hydrate or ergot mesylates) to a high of 10.9% for nonsteroidal anti-inflammatory drugs (NSAIDs) and 9.3% for benzodiazepines. NSAIDs affected 25.7% of older adults with PIM fills and benzodiazepines affected 21.7%. The average number of prescription fills per person that met the broad definition for a PIM ranged from a low of 2.6 for nitrofurantoin to a high of 6.4 for selected sulfonylureas and tricyclic antidepressants.

Table 1. Potentially Inappropriate Medication (PIM) Receipt in Older Adults According to Detailed Categories, Broad and Qualified Definitions, 2006–2010
Prevalence of PIMS Broad Definition Qualified Definition
Persons with PIMs as a Percentage of Older Adults with Annual PIM Fills per Person/Category Persons with PIMs as a Percentage of Older Adults with Annual PIM Fills per Person/Category
Any Drug Usea Any PIMs Any Drug Usea Any PIMs
Any 42.6 100.0 6.9 30.9 100.0 6.9
According to category or subcategory
Anticholinergic
First-generation antihistamine 3.8 8.8 2.7 3.6 11.6 2.8
Antiparkinson 0.1 0.3 ns 0.1 0.4 ns
Antispasmodic 2.8 6.6 3.6 2.8 9.1 3.6
Antithrombotic
Dipyridamole rse rse ns rse rse ns
Ticlopidine rse rse ns rse rse ns
Anti-infective
Nitrofurantoin 1.3 3.0 2.6 0.6 1.9 ns
Cardiovascular
Alpha 1 blocker 4.3 10.0 5.3 3.7 12.1 5.3
Alpha agonist 2.2 5.2 6.1 2.2 7.2 6.1
Antiarrhythmic 2.4 5.7 6.1 1.0 3.4 5.1
Disopyramide rse rse ns rse rse ns
Dronedarone rse rse ns rse rse ns
Digoxin 1.8 4.2 6.0 1.8 5.8 6.0
Nifedipine, immediate release 0.2 0.4 ns 0.2 0.6 ns
Spironolactone 0.4 1.0 ns rse rse ns
Central nervous system
Tricyclic antidepressant 2.3 5.4 6.4 2.3 7.4 6.4
Antipsychotic 1.8 4.1 6.1 0.3 0.9 ns
Thioridazine, mesoridazine rse rse ns rse rse ns
Barbiturate 0.4 1.0 ns 0.4 1.5 ns
Benzodiazepine (all) 9.3 21.7 5.6 0.9 2.9 7.1
Short acting 6.7 15.7 5.6 0.6 1.8 6.9
Long acting 2.8 6.7 5.3 0.4 1.2 ns
Chloral hydrate 0.0 0.0 ns 0.0 0.0 ns
Meprobamate rse rse ns rse rse ns
Nonbenzodiazepine hypnotic 3.3 7.8 4.7 1.7 5.5 7.3
Ergot mesylate 0.0 0.0 ns 0.0 0.0 ns
Endocrine
Androgen 0.3 0.8 ns 0.2 0.7 ns
Desiccated thyroid 0.6 1.5 ns 0.6 2.1 ns
Estrogen with or without progestin 3.6 8.4 5.2 3.6 11.6 5.2
Growth hormone rse rse ns rse rse ns
Megestrol 0.3 0.8 ns 0.3 1.1 ns
Sulfonylurea 4.1 9.6 6.4 4.1 13.3 6.4
Gastrointestinal
Metoclopramide 1.0 2.4 5.4 1.0 3.3 5.4
Trimethobenzamide rse rse ns rse rse ns
Pain
Meperidine 0.1 0.2 ns 0.1 0.3 ns
Non-cyclooxygenase-selective nonsteroidal anti-inflammatory drug 10.9 25.7 4.0 4.7 15.2 6.9
Indomethacin, ketorolac 0.6 1.5 2.9 0.6 2.1 2.9
Pentazocine rse rse ns rse rse ns
Skeletal muscle relaxant 3.0 7.1 3.5 3.0 9.9 3.5
  • ns = insufficient sample to support reliable estimate; rse = relative standard error >0.3.
  • Source: Medical Expenditure Panel Survey Household Component, 2006–2010.
  • a Person purchased, or otherwise acquired, at least one outpatient prescription drug during the year.

The overall prevalence using the qualified definition was 30.9% of older adults with drug use. The most prevalent individual categories continued to include NSAIDs, although the rate dropped to 4.7%, and 4.1% used selected sulfonylureas. The proportion with potentially inappropriate benzodiazepine use was much lower under the qualified definition (0.9%).

Only 9.2% of individual prescription fills could be classified as PIMs using the broad definition, with 6.6% under the qualified definition. Distribution according to drug category is described in Appendix Table A3.

Figure 1 presents information about the distribution of additional criteria that were met under the qualified definition. Of older adults with any PIM under the qualified definition, 6.8% had a PIM that met specific dose criteria, 22.4% had a PIM that met duration criteria, and 19.5% had a PIM that met restrictions based on reasons for use; 67.2% had at least one fill for a drug that should always be avoided.

Details are in the caption following the image
Percentage of older adults receiving potentially inappropriate medications (PIMs) according to the type of qualifying criteria applied, qualified definition. Individuals may have PIM use in more than one category. Source: Medical Expenditure Panel Survey, Household Component, Consolidated files, 2006–2010.

Figure 2 reports the distribution of adults according to the quality of evidence used to characterize PIM status. Using the broad definition, 22% of the 42.6% of adults with a PIM had at least one medication for which the evidence quality was deemed to be high. Using the qualified definition, 14.0% of the 30.9% of adults with a PIM had at least one medication for which the evidence quality was deemed to be high. Almost all adults with PIMs had at least one drug category for which the recommendation was considered to be strong (data not shown).

Details are in the caption following the image
Percentage of older adults receiving potentially inappropriate medications (PIM) according to the quality of evidence, broad and qualified definitions. Individuals with multiple PIMs are categorized according the PIM with the highest level of evidence. Source: Medical Expenditure Panel Survey, Household Component, Consolidated files, 2006–2010.

Table 2 shows trends in the proportion of persons with PIM fills using the broad definition, comparing 2006–2007 with 2009–2010. Overall, the rate declined from 45.5% in 2006–2007 to 40.8% in 2009–2010, a 10.3% decrease from baseline (P < .001). The categories with the largest absolute decline were NSAIDs, selected sulfonylureas, and estrogens, whereas use of skeletal muscle relaxants increased during this period. Parallel results using the qualified definition are provided as Appendix Table A4.

Table 2. Change in Percentage of Older Adults with Potentially Inappropriate Medication (PIM) Receipt, Broad Definition, 2006–2007 Versus 2009–2010
PIM Category Percentage of Older Adults with a PIM Percentage Point Change Change, %
2006–2007 2009–2010
All 45.5 40.8 −4.7a −10.3
Anticholinergic
First-generation antihistamine 4.2 3.4 −0.8b −19.3
Antiparkinson rse rse
Antispasmodic 2.9 3.0 0.1 5.2
Antithrombotic
Dipyridamole rse rse
Ticlopidine rse rse
Anti-infective
Nitrofurantoin 1.5 1.3 −0.3 −18.0
Cardiovascular
Alpha 1 blocker 4.5 3.9 −0.6 −12.8
Alpha agonist 2.4 2.2 −0.3 −11.5
Antiarrhythmic 2.3 2.6 0.2 9.3
Disopyramide rse rse
Dronedarone 0.0 rse
Digoxin 2.0 1.4 −0.6b −30.3
Nifedipine, immediate release 0.3 rse
Spironolactone 0.4 0.6 0.2 58.3
Central nervous system
Tricyclic antidepressant 2.5 2.1 −0.4 −16.5
Antipsychotic 2.0 1.7 −0.3 −15.4
Thioridazine, mesoridazine rse 0.0
Barbiturate 0.5 0.4 −0.1 −23.2
Benzodiazepine (all) 9.5 9.0 −0.6 −5.9
Short acting 7.1 6.3 −0.8 −11.3
Long acting 2.8 2.9 0.2 6.0
Chloral hydrate 0.0 0.0
Meprobamate rse rse
Nonbenzodiazepine hypnotic 3.3 3.5 0.1 4.5
Ergot mesylate 0.0 0.0
Endocrine
Androgen 0.4 0.4 0.0 6.2
Desiccated thyroid 0.9 0.4 −0.5a −56.5
Estrogen with or without progestin 4.3 3.2 −1.1a −26.6
Growth hormone rse 0.0
Megestrol 0.4 0.3 −0.1 −16.5
Sulfonylurea 4.5 3.5 −1.0a −22.4
Gastrointestinal
Metoclopramide 1.1 0.9 −0.2 −14.0
Trimethobenzamide rse rse
Pain
Meperidine rse rse
Non-cyclooxygenase-selective nonsteroidal anti-inflammatory drug 11.7 10.4 −1.4b −11.7
Indomethacin, ketorolac 0.6 0.6 0.0 6.2
Pentazocine rse 0.0
Skeletal muscle relaxant 3.0 3.3 0.3 10.2
  • rse = relative standard error >0.3.
  • Source: Medical Expenditure Panel Survey, Household Component, 2006–2010.
  • a.05, b.10.

Discussion

PIM use has been examined over the past 3 decades using previously published Beers criteria, in different settings and subpopulations, and in the United States and internationally.4, 10-12 Given the changing landscape of available drugs, it is important to present updated information about the prevalence of PIM exposure. This is the first study to use nationally representative data for the U.S. community-based population to estimate prevalence of PIMs using the 2012 update to the Beers criteria. Two operational definitions, broad and qualified, were developed and applied to capture PIM use. Because the updated criteria included clinical caveats that are often important but hard to clearly define with administrative data, this approach is innovative and reveals important new information for research, education, and practice. The most compelling finding shows that 42% of older adults received PIMs but that a portion of them were not used for particularly long or were used for individuals who lacked specific medical conditions targeted in the Beers criteria as problematic or who had diagnoses for which use was justified. Even with these qualifications in the definition, 30.9% of community-dwelling older adults were prescribed PIMs, some of which are known to be associated with falls, delirium, declines in cognitive and physical functioning, and other potentially serious health outcomes.5, 6, 16 Although the proportion of older adults with PIM use is large, these prescriptions make up a much smaller proportion of total prescription fills, suggesting that most prescribing is not problematic with respect to the dimensions that the Beers criteria capture. The analysis does not address adherence to guidelines or cost of chosen therapy, dimensions that may also be relevant as quality or value indicators.

The updated 2012 American Geriatrics Society Beers criteria continue to highlight the use of PIMs in older adults. The updated criteria were applied to the 2006–2010 MEPS, which was the most-recent population-based data available when this study was conducted. These data predate the new guidelines and so do not reflect potential changes in prescribing that may result from their dissemination, although the evidence base used to update the criteria was developed over time, with some of the information available to clinicians during the study period. In an environment with evolving availability of new and old drugs and evidence of their effectiveness and safety, this study provides a useful snapshot and an important benchmark to assess the effect of the updated criteria over time.

This analysis suggests that PIM use is decreasing, but with the addition of new medication categories such as nonbenzodiazepine sedatives, continued intervention and surveillance are needed.4 Furthermore, PIM use has been operationalized as a marker of quality prescribing through various metrics and indicators.17-19 Therefore, it is helpful to target critical drug categories that have the highest prevalence in this study, namely first-generation antihistamines, antispasmodics, nonselective alpha1 blockers, nonbenzodiazepine hypnotics, estrogens, selected sulfonylureas, NSAIDs and skeletal-muscle relaxants. Benzodiazepine use was also highly prevalent using the broad definition, although much less common when applying the qualified definition. Despite the continued prevalence in some categories, it is encouraging that the use of agents such as selected sulfonylureas and digoxin, which have limited efficacy and adverse effects in older adults, appears to be declining.

These estimates of PIM exposure are similar to estimates from selected subgroups of community-based older adults enrolled in managed care organizations (40.7%)6 or receiving home care (38%)20 yet substantially higher than those of other general community-based population estimates, for example, the estimated 13.8% of older adults with PIM use in 2007.12 One study12 applied the same criteria published in an earlier study10 to compare trends over time. Although this approach may document discontinuation of older drugs, it does not incorporate newly available drugs or drugs for which there is new evidence of harm and will underestimate the extent of problematic drug prescribing. Another important difference is that other studies operationalized only the subset of criteria in which the drug was to be avoided in all cases. The current results indicate that, using the qualified definition, 67% of all adults with PIMs had used PIMs that should be avoided universally.

There are limitations to this analysis. Some of the Beers criteria could not be fully implemented; for example, the MEPS lacks information on the exact timing of medication use, so it was not possible to assess concurrent medication use or the timing of medication use relative to condition diagnosis. The MEPS drug use data may be subject to underreporting. A recent comparison of drug use reported in the MEPS by Medicare beneficiaries and Part D claims suggested that the MEPS underreports medication use for acute conditions, albeit to a lesser extent for chronic use.21 Underreporting, therefore, is likely to have affected the results for nitrofurantoin but should be less important for the majority of other PIMs, which are primarily used to treat chronic conditions. Subsequent improvements to procedures for editing drug quantity in the MEPS were applied to these data and are expected to reduce underreporting.15, 22 An additional limitation is that the reason for use of each drug was based on self-report. Older adults taking many medications, especially off label, may not be aware of the correct indication. As a result, certain diagnoses may be underreported as the reason for using a drug. For example, antipsychotics are contraindicated for treatment of behavioral problems of dementia. Very few prescription records were found in which the specific ICD-9-CM code was consistent with that criterion. The qualified criteria were broadened in two ways, first by including antipsychotics used for dementia without requiring the behavioral problems and then by including antipsychotic use if the older adult reported a diagnosis of dementia associated with any healthcare services or a disability, but the capture rate was still quite low. As a result of the potential underreporting of dementia, there is a large difference between PIM prevalence estimates using the broad definition (that did not require a diagnosis of dementia) and the qualified definition associated with antipsychotics. These were considered to be upper and lower bound estimates; where the true estimate lies is unclear.

Despite these caveats, this study has several important strengths, including the large sample size and transparent methods of determining PIM use using the updated Beers criteria. This is also the first study to operationalize the clinical definition of PIMS in a broad and qualified manner. This study has clear practice implications, and the results illustrate that interventions to decrease PIM use are still greatly needed. Several interventions have been found to decrease PIM use if implemented before the point of ordering.23 Several studies have used interruptive alerts when PIMs are prescribed to recommend alternative medications or nondrug approaches,24 lower dose and frequency, or no medication.25, 26 These types of alerts are most effective because they require an action from the provider or prescriber of the PIM before proceeding.

PIM exposure is a critical element of quality. Because prescription drug availability changes over time, it is important to update the criteria used and to add new drugs or eliminate drugs that are no longer on the market continually. Characteristics of older adults who are likely to experience PIMs, selected characteristics of providers, and the association between PIM receipt and other dimensions of quality are being examined in a related study. Future studies should test interventions to decrease PIM use and evaluate the effect on clinical and individual outcomes. To have the largest return on investment, interventions to reduce PIMs should also focus on the most-common PIMs, such as NSAIDs and short-acting benzodiazepines. Ongoing studies using the updated criteria are important to advancing the quality initiative for the prevention of adverse medication events in older adults.

Acknowledgments

The views expressed in this article are those of the authors, and no official endorsement by the Agency for Healthcare Research and Quality, Department of Health and Human Services is intended or should be inferred.

Conflict of Interest: Drs. Davidoff, Miller, Sarpong and Brandt and Ms. Yang received no external financial support for the research and/or authorship of this article. Dr. Fick receives partial support for her time from National Institute of Nursing Research Grant R01 NR011042.

Author Contributions: Davidoff, Brandt, Fick: study concept and design. Miller, Sarpong, Yang: data analysis. Davidoff, Sarpong, Miller: drafting of the article. Brandt, Fick, Yang: critical revisions for intellectual content. All authors: approval of final version.

Sponsor's Role: None.

    Appendix

    Table A1. Operationalizing Broad and Qualified Definitions of Potentially Inappropriate Medication Use in the Medical Expenditure Panel Survey (MEPS), 2006–2010
    Category or Subcategory Broad Definition Qualified Definition
    Conditions and Exceptions Detailed Notes Conditions and Exceptions Detailed Notes
    Anticholinergic (excluding TCA)
    First-generation antihistamine Avoid all, except for single-agent diphenhydramine with route of administration other than oral Administration route for diphenhydramine almost always oral Oral diphenhydramine excepted (not a PIM) if duration <1 month For single-agent diphenhydramine, duration and exempted use ≤1 month assessed as proxy for acute use to treat severe allergic reactions
    Antiparkinson agent Avoid all, except benztropine with route of administration other than oral Same as broad definition
    Antispasmodic Avoid all, no exceptions Same as broad definition
    Antithrombotic
    Dipyridamole Avoid only oral short-acting version Avoid if dose form is not extended-release tablet or capsule or delayed-release capsule Same as broad definition
    Ticlopidine Avoid all, no exceptions Same as broad definition
    Anti-infective
    Nitrofurantoin Avoid all, no exceptions Avoid for long-term use, avoid in individuals with creatinine clearance <60 mL/min Long-term use defined as >30 days, measured by days supplied; could not operationalize creatinine clearance measure; instead searched for any reported health service use to treat chronic kidney disease (ICD-9-CM codes 403, 404, 585) as proxy for poor creatinine clearance
    Cardiovascular
    Alpha 1 blocker Avoid doxazosin, prazosin, and terazosin Avoid for treatment of hypertension and most other uses; exception if drug is used to increase urinary flow in men with BPH Excepted if BPH (ICD-9-CM code 600) reported as condition alpha 1 blockers were intended to treat
    Alpha agonist, central Avoid all, no exceptions Avoid clonidine as first-line antihypertensive Could not operationalize therapy lines in MEPS
    Antiarrhythmic drug (Class 1a, 1c, III) Avoid all, no exceptions Avoid use for atrial fibrillation; all other conditions excepted Categorized as PIM if atrial fibrillation (ICD-9-CM codes 427.3*, 427.89, or 427.9) reported as condition antiarrhythmic drugs intended to treat
    Disopyramide Avoid all, no exceptions Same as broad definition
    Dronedarone Avoid all, no exceptions Avoid if individuals has atrial fibrillation or CHF Searched for any health services use to treat atrial fibrillation (ICD-9-CM codes 427.31, 427.32, 427.89, 427.9) or CHF (ICD-9-CM code 428)
    Digoxin Avoid if dose >125 μg/d Assigned if digoxin strength is 250 μg Same as broad definition
    Nifedipine, immediate release Avoid if not extended release Exception if dose form is extended-release tablet or capsule or delayed-release capsule Same as broad definition
    Spironolactone Avoid if dose >25 mg/d Assigned if spironolactone strength is 50 or 100 mg Avoid if dose >25 mg/d and if individual has CHF or creatinine clearance <30 mL/min Could not operationalize creatinine clearance criterion; instead used CKD diagnosis; assigned if spironolactone strength is 50 or 100 mg and if individual has any reported health service use to treat CHF (ICD-9-CM code 428) or CKD (ICD-9-CM codes 403, 404, 585)
    Central nervous system
    TCA Avoid if doxepin >6 mg/d; avoid all others Same as broad definition
    Antipsychotic (conventional and atypical) Avoid all, no exceptions Avoid use for behavioral problems of dementia Limited use of codes for behavioral problems of dementia, so used broader set of dementia codes (ICD-9-CM codes 290, 294, 331) reported as condition antipsychotics were intended to treat; because of concerns about underreporting of dementia as a reason for drug use, assigned as PIM if individual reported any health service use to treat dementia
    Thioridazine, mesoridazine Avoid all, no exceptions Same as broad definition
    Barbiturate Avoid all, no exceptions Same as broad definition
    Benzodiazepine—short acting Avoid all, no exceptions Avoid if used to treat delirium, insomnia, or agitation Categorized as PIM if delirium (ICD-9 ICD-9-CM codes 290, 780.1), or insomnia (ICD-9 ICD-9-CM codes 780.5, 780.9, 327) reported as condition benzodiazepines intended to treat; because of concern about underreporting of dementia (underlying reason for agitation diagnosis) as reason for drug use, assigned as PIM if individual reported any health service use for dementia
    Benzodiazepine—long acting Avoid all, no exceptions
    Chloral hydrate Avoid all, no exceptions Same as broad definition
    Meprobamate Avoid all, no exceptions Same as broad definition
    Nonbenzodiazepine hypnotic Avoid all, no exceptions Avoid chronic use (>90 days supplied during year) Duration of therapy measured based on days supplied
    Ergot mesylate: isoxsuprine Avoid all, no exceptions Same as broad definition
    Endocrine
    Androgen Avoid all, no exceptions Avoid except for use for hypogonadism Excepted if hypogonadism (ICD-9-CM code 257) was reported as condition androgens intended to treat
    Desiccated thyroid Avoid all, no exceptions Same as broad definition
    Estrogen with or without progestin Avoid oral or transdermal, except vaginal Same as broad definition
    Growth hormone Avoid all, no exceptions Same as broad definition
    Insulin, sliding scale Not implemented MEPS did not provide a mechanism to distinguish fixed from flexible dosing schedule Same as broad definition
    Megestrol Avoid all, no exceptions Same as broad definition
    Sulfonylurea Avoid chlorpropamide and glyburide, no exceptions Same as broad definition
    Gastrointestinal
    Metoclopramide Avoid all, no exceptions Avoid except for use for gastroparesis Unable to implement exception because of underreporting of gastroparesis
    Mineral oil, oral Not implemented Over-the-counter purchases poorly documented in MEPS Not implemented
    Trimethobenzamide Avoid all, no exceptions Same as broad definition
    Pain
    Meperidine Avoid all, no exceptions Same as broad definition
    Aspirin Avoid if daily dose >325 mg Assigned if aspirin strength was 600 or 770 mg Avoid chronic NSAID use Assigned if aspirin strength was 600 mg or 770 mg and days supplied >90
    Non-cyclooxygenase-selective NSAID, oral Avoid all, no exceptions Avoid chronic NSAID use Assigned if days supplied >90
    Indomethacin Avoid all, no exceptions Same as broad definition
    Ketorolac, oral and parenteral Avoid all, no exceptions Same as broad definition
    Pentazocine Avoid all, no exceptions Same as broad definition
    Skeletal muscle relaxants Avoid all, no exceptions Same as broad definition
    • TCA = tricyclic antidepressant; BPH = benign prostatic hypertrophy; CHF = congestive heart failure; ICD-9 = International Classification of Diseases, Ninth Revision; CKD = chronic kidney disease; NSAID = nonsteroidal anti-inflammatory drug; PIM = potentially inappropriate medication; MEPS = medical expenditure panel survey.
    Table A2. Characteristics of U.S. Community-Dwelling Older Adults, 2006–2010
    Characteristic Total Number of Older Adults Percentage Distribution of Older Adults (SE)
    Unweighted Average Annual Total (1,000) (SE)
    Age
    65–74 9,912 20,603 (558) 52.1 (0.8)
    75–84 6,241 13,738 (429) 34.7 (0.6)
    ≥85 2,322 5,241 (238) 13.2 (0.5)
    Race and ethnicity
    White, non-Hispanic 11,748 31,507 (875) 79.6 (0.8)
    Black, non-Hispanic 3,063 3,386 (170) 8.6 (0.4)
    Hispanic 2,326 2,784 (174) 7.0 (0.5)
    Other 1,338 1,906 (214) 4.8 (0.5)
    Sex
    Male 7,897 17,066 (456) 43.1 (0.4)
    Female 10,578 22,517 (549) 56.9 (0.4)
    Marital status
    Married 9,654 21,359 (655) 54.0 (0.8)
    Formerly married 8,063 16,776 (450) 42.4 (0.7)
    Never married 757 1,446 (97) 3.7 (0.2)
    Education
    <High school 5,822 9,375 (277) 23.7 (0.6)
    High school graduate 5,905 13,719 (433) 34.7 (0.7)
    Some college (1–3 years) 3,003 7,218 (264) 18.2 (0.5)
    College graduate (4 years) 1,906 4,730 (231) 12.0 (0.5)
    Postgraduate (≥5 years) 1,533 4,053 (203) 10.2 (0.4)
    Federal poverty level, %
    <100 (poor) 2,926 3,902 (146) 9.9 (0.3)
    100–199 (low income) 4,847 10,068 (288) 25.4 (0.5)
    200–399 (middle income) 5,411 11,674 (364) 29.5 (0.6)
    ≥400 (high income) 5,291 13,938 (471) 35.2 (0.7)
    Supplemental insurance
    Private group with drug coverage 6,287 15,057 (465) 38.0 (0.7)
    Private nongroup with drug coverage 2,894 3,953 (152) 10.0 (0.4)
    Medicaid with drug coverage 1,648 4,080 (179) 10.3 (0.4)
    Medicare managed care with drug coverage 3,775 8,055 (361) 20.3 (0.7)
    No medical supplement, drug coverage 1,669 3,535 (189) 8.9 (0.4)
    Private or public supplemental without drug coverage 694 1,778 (130) 4.5 (0.3)
    No medical or drug supplemental coverage 1,508 3,124 (158) 7.9 (0.4)
    General health status
    Excellent to very good 4,203 10,456 (350) 26.4 (0.5)
    Good 6,516 14,460 (420) 36.5 (0.5)
    Fair to poor 7,581 14,295 (377) 36.1 (0.6)
    BMI, kg/m2
    <25.0 6,040 13,105 (392) 33.1 (0.6)
    25.0–29.9 6,572 14,181 (420) 35.8 (0.5)
    ≥30.0 4,647 9,675 (292) 24.4 (0.5)
    ADL limitations
    Yes 2,000 3,720 (190) 9.4 (0.4)
    No 15,867 34,346 (857) 86.8 (0.4)
    IADL limitations
    Yes 3,556 6,884 (251) 17.4 (0.5)
    No 14,311 31,181 (809) 78.8 (0.5)
    Census region
    Northeast 3,067 7,864 (355) 19.9 (0.8)
    Midwest 3,911 8,789 (461) 22.2 (1.0)
    South 7,231 14,634 (566) 37.0 (1.1)
    West 4,266 8,295 (360) 21.0 (0.8)
    Metropolitan Statistical Area status
    Yes 14,629 31,907 (870) 80.6 (1.40)
    No 3,846 7,675 (610) 19.4 (1.40)
    Has usual source of care?
    Yes 16,350 35,259 (875) 89.1 (0.33)
    No 1,342 2,513 (124) 6.3 (0.29)
    More likely than others to take risks
    No or uncertain 13,509 28,918 (703) 73.1 (0.44)
    Yes 2,687 5,828 (204) 14.7 (0.35)
    Smoke
    No 14,851 32,002 (810) 80.8 (0.41)
    Yes 1,670 3,367 (130) 8.5 (0.30)
    • SE = standard error.
    • Source: Medical Expenditure Panel Survey Household Component, Consolidated Files 2006–2010.
    • Because of rounding and missing values, percentages do not always sum to 100%.
    • Education, health status, body mass index (BMI), activities of daily living (ADLs), instrumental activities of daily living (IADLs), usual source of care, more likely to take risks, and smoking status had some missing values.
    Table A3. Potentially Inappropriate Prescription Medication Fills Among Older Adults, by Detailed Categories, 2006–2010
    PIM Category Broad Definition Qualified Definition
    PIM Fills as Percentage of PIM Fills as Percentage of
    All Drug Purchases All PIM Fills All Drug Purchases All PIM Fills
    All 9.2 100.0 6.6 100.0
    Anticholinergic
    First-generation antihistamine 0.3 3.5 0.3 4.7
    Antiparkinson 0.0 0.3 0.0 0.4
    Antispasmodic 0.3 3.4 0.3 4.8
    Antithrombotic
    Dipyridamole rse rse 0.0 rse
    Ticlopidine rse rse 0.0 rse
    Anti-infective
    Nitrofurantoin 0.1 1.2 0.1 1.2
    Cardiovascular
    Alpha 1 blocker 0.7 7.7 0.6 9.3
    Alpha agonist 0.4 4.6 0.4 6.4
    Antiarrhythmic 0.5 5.0 0.2 2.5
    Disopyramide rse rse rse rse
    Dronedarone rse rse rse rse
    Digoxin 0.3 3.6 0.3 5.0
    Nifedipine, immediate release 0.0 0.3 0.0 0.4
    Spironolactone 0.1 0.8 0.0 rse
    Central nervous system
    Tricyclic antidepressant 0.5 5.0 0.5 6.9
    Antipsychotic 0.3 3.6 0.0 0.6
    Thioridazine, mesoridazine rse rse rse rse
    Barbiturate 0.1 0.7 0.1 0.9
    Benzodiazepine (all) 1.6 17.7 0.2 3.0
    Short acting 1.2 12.6 0.1 1.8
    Long acting 0.5 5.1 0.1 1.1
    Chloral hydrate 0.0 0.0 0.0 0.0
    Meprobamate rse rse rse rse
    Nonbenzodiazepine hypnotic 0.5 5.3 0.4 5.9
    Ergot mesylate 0.0 0.0 0.0 0.0
    Endocrine
    Androgen 0.0 0.5 0.0 0.4
    Desiccated thyroid 0.1 1.3 0.1 1.8
    Estrogen with or without progestin 0.6 6.3 0.6 8.8
    Growth hormone rse rse 0.0 rse
    Megestrol 0.0 0.4 0.0 0.5
    Sulfonylurea 0.8 8.8 0.8 12.3
    Gastrointestinal
    Metoclopramide 0.2 1.9 0.2 2.6
    Trimethobenzamide rse rse 0.0 rse
    Pain
    Meperidine rse rse 0.0 rse
    Non-cyclooxygenase-selective nonsteroidal anti-inflammatory drug 1.4 14.7 1.0 15.3
    Indomethacin, ketorolac 0.1 0.6 0.1 0.9
    Pentazocine rse rse 0.0 rse
    Skeletal muscle relaxant 0.3 3.6 0.3 5.1
    • rse = relative standard error >0.3; PIM = potentially inappropriate medication.
    • Source: Medical Expenditure Panel Survey Household Component, 2006–2010.
    Table A4. Change in Percent of Older Adults with Potentially Inappropriate Medication (PIM) Receipt, Qualified Definition, 2006/07 Versus 2009/10
    PIM Category Percentage of Older Adults with a PIM Percentage Point Change Percentage Change
    2006–2007 2009–2010
    Any 33.2 29.4 −3.8a −11.5
    Anticholinergic
    First-generation antihistamine 4.0 3.2 −0.8b −19.3
    Antiparkinson rse rse 0.0
    Antispasmodic 2.9 3.0 0.1 5.2
    Antithrombotic
    Dipyridamole rse rse
    Ticlopidine rse rse
    Anti-infective
    Nitrofurantoin 0.7 0.5 −0.2 −29.5
    Cardiovascular
    Alpha 1 blocker 3.8 3.5 −0.3 −7.6
    Alpha agonist 2.4 2.2 −0.3 −11.5
    Antiarrhythmic 1.0 1.2 0.2 19.9
    Disopyramide rse rse
    Dronedarone 0.0 0.0 0.0
    Digoxin 2.0 1.4 −0.6b −30.3
    Nifedipine, immediate release 0.3 rse
    Spironolactone 0.0 rse
    Central nervous system
    Tricyclic antidepressant 2.5 2.1 −0.4 −16.5
    Antipsychotic 0.2 0.3 0.1 36.4
    Thioridazine, mesoridazine rse 0.0
    Barbiturate 0.5 0.4 −0.1 −23.2
    Benzodiazepine (all) 0.9 0.9 0.0 2.9
    Short acting 0.7 0.6 −0.1 −15.3
    Long acting 0.3 0.4 0.1 30.1
    Chloral hydrate 0.0 0.0 0.0
    Meprobamate rse rse
    Nonbenzodiazepine hypnotic 1.5 2.0 0.5 30.0
    Ergot mesylate 0.0 0.0 0.0
    Endocrine
    Androgen rse 0.2
    Desiccated thyroid 0.9 0.4 −0.5a −56.5
    Estrogen with or without progestin 4.3 3.2 −1.1a −26.6
    Growth hormone rse 0.0
    Megestrol 0.4 0.3 −0.1 −16.5
    Sulfonylurea 4.5 3.5 −1.0a −22.4
    Gastrointestinal
    Metoclopramide 1.1 0.9 −0.2 −14.0
    Trimethobenzamide rse rse
    Pain
    Meperidine rse rse
    Non-cyclooxygenase-selective nonsteroidal anti-inflammatory drug 4.9 4.7 −0.2 −3.8
    Indomethacin, ketorolac 0.6 0.6 0.0 6.2
    Pentazocine rse 0.0
    Skeletal muscle relaxant 3.0 3.3 0.3 10.2
    • rse = relative standard error >0.3.
    • Source: Medical Expenditure Panel Survey Household Component, Consolidated Files 2006–2010.
    • a.05, b.10.