Appropriateness of Medication Selection for Older Persons in an Urban Academic Emergency Department

1232 ELDER DRUGS Chin et al. • INAPPROPRIATE MEDICATIONS FOR ELDERS

Appropriateness of Medication Selection for OlderPersons in an Urban Academic Emergency Department

MARSHALL H. CHIN, MD, MPH, LINDA C. WANG, JD, LEI JIN, MA,ROBERT MULLIKEN, MD, JAMES WALTER, MD, DEON COX HAYLEY, DO,

THEODORE G. KARRISON, PHD, MICHAEL P. NERNEY, BA,ANNETTE MILLER, RN, PETER D. FRIEDMANN, MD, MPH

Abstract. Objectives: To determine the frequencyof potentially inappropriate medication selection forolder persons presenting to the ED, the most commonproblematic drugs, risk factors for suboptimal medi-cation selection, and whether use of these medica-tions is associated with worse outcomes. Methods:

The authors performed a prospective cohort study of898 patients 65 years or older who presented to anurban academic ED in 1995 and 1996. Seventy-ninepercent of the patients were African-American and43% did not graduate from high school. Potentiallyinappropriate medications and adverse drug–diseaseinteractions were identified using the 1997 Beers ex-plicit criteria for elders. During the three months af-ter the initial visit, revisits to the ED or hospital,death, and changes in health-related quality of lifewere analyzed as measured by validated questionsadapted from the Medical Outcomes Study. Results:

Upon presentation, 10.6% of the patients were takinga potentially inappropriate medication, 3.6% weregiven one in the ED, and 5.6% were prescribed oneupon discharge from the ED. The most frequently

prescribed potentially inappropriate medications inthe ED were diphenhydramine, indomethacin, me-peridine, and cyclobenzaprine. Emergency physiciansadded potentially inappropriate medications most of-ten to patients with discharge diagnoses of musculo-skeletal disorder, back pain, gout, and allergy or ur-ticaria. Potentially adverse drug–disease interactionswere relatively uncommon at presentation (5.2%), inthe ED (0.6%), and on discharge from the ED (1.2%).Potentially inappropriate medications and adversedrug–disease interactions prescribed in the ED werenot associated with higher rates of revisit to the ED,hospitalization, or death, but were correlated withworse physical function and pain. However, confi-dence intervals were wide for analyses of revisits anddeath. Conclusions: Suboptimal medication selectionwas fairly common and was associated with worse pa-tient-reported health-related quality of life. Key

words: emergency services; aging; drugs; pharmacol-ogy; outcomes; quality of care. ACADEMIC EMER-GENCY MEDICINE 1999; 6:1232–1242

ADVERSE drug reactions are common and ex-pensive. For example, a recent meta-analysis

estimated that 6.7% of hospitalized patients haveserious adverse drug reactions.1 In the outpatientsetting, rates of adverse drug reactions haveranged from 1.7% to 50.6%.2 Costs associated withadverse drug events in a typical tertiary care hos-pital have been approximated to be $1.1–5.6 mil-lion per year,3,4 and drug-related morbidity andmortality have been estimated to cost $76.6 billion

annually in the ambulatory setting.5 Therefore, cli-nicians, administrators, and policymakers have be-come interested in what system factors contributeto adverse drug reactions and how provider prac-tices around medication prescribing can be im-proved.6,7

Several factors place older persons at especiallyhigh risk for iatrogenic complications of medicationuse. The normal aging process is associated withphysiologic changes that can affect drug excretion,

From the Section of General Internal Medicine (MHC, LJ,DCH, MPN, PDF), Section of Emergency Medicine (RM, JW,AM), Department of Medicine, Department of Health Studies(TGK), Pritzker School of Medicine (LCW), University of Chi-cago, Chicago, IL. Mr Nerney is currently a medical student atthe Chicago College of Osteopathic Medicine, Midwestern Uni-versity, Downers Grove, IL. Dr. Friedmann is currently withthe Division of General Internal Medicine, Rhode Island Hos-pital, Brown University, Providence, RI.Received May 13, 1999; revision received August 3, 1999; ac-cepted August 20, 1999. Presented in part at the annual meet-ing of the Midwest Society of General Internal Medicine, Chi-cago, IL, September 1998.

Supported by grants from the Chicago Community Trust andRetirement Research Foundation. Dr. Chin is supported by Na-tional Institute on Aging Geriatric Academic Program Award5-K12-AG00488. Dr. Friedmann is supported by MentoredClinical Scientist Career Development Award 1K08-DA00320from the National Institute on Drug Abuse. Ms. Wang was sup-ported by the 1997 University of Chicago Pritzker School ofMedicine Summer Research Program.Address for correspondence and reprints: Marshall H. Chin,MD, MPH, University of Chicago, Section of General InternalMedicine, 5841 South Maryland Avenue, MC 2007, Chicago, IL60637. Fax: 773-834-2238; e-mail: [email protected]

ACADEMIC EMERGENCY MEDICINE • December 1999, Volume 6, Number 12 1233

metabolism, concentration, and distribution (phar-macokinetics), as well as drug action at their re-ceptor sites (pharmacodynamics).8 Even at normaldrug concentrations, many organs have alteredsensitivity to drugs, and medications that youngadults tolerate may be toxic to older persons. Ad-ditionally, many older persons take multiple drugsfor a variety of concomitant but unrelated diseases.At any one time, older persons take an average of4.5 medications, and have an average of 5 to 10diseases.9 As drugs are often not disease-specific intheir action, the treatment of one medical condi-tion may worsen another. For example, diuretictherapy for hypertension or congestive heart fail-ure may exacerbate urinary incontinence. Thus,physicians must be aware of potentially inappro-priate prescriptions for the elder population.

In recent years, a number of investigators haveattempted to devise quality of care standards forthe use of medications in older persons in an at-tempt to reduce the incidence of adverse drug re-actions. These criteria are based primarily onexpert opinion and consensus panels. A number ofstudies have examined the incidence of adversedrug events in elders,10,11 as well as the incidenceof inappropriate medications in the nursing homeand community settings.12–17 However, few inves-tigators have attempted to validate drug appropri-ateness criteria for older persons by linking pro-cesses of care to outcome.18,19 Moreover, existingstudies do not adjust adequately for confoundingfactors,18 or else use appropriateness criteria thatmay be overly sensitive.19 We are also unaware ofany studies of the appropriateness of geriatricmedication prescription in the ED.

Therefore, we aimed to study the 1997 Beerscriteria, the most recent and perhaps most widelyaccepted criteria for inappropriate geriatric medi-cation selection,20,21 among older persons present-ing to an ED. The ED is a particularly importantsetting for the study of medication use among el-ders, as patients are seen quickly, the emphasis isgenerally on the chief complaint even though thepatient may have multiple problems, and often thepatient is unknown to the physician. Also, we cananalyze the medications the patient was receivingin the outpatient setting prior to the visit, the med-ications prescribed in the ED, and any relationshipto subsequent outcome during a time of great po-tential vulnerability of the patient. In addition,geriatrics issues are generally not emphasized inthe training of emergency physicians (EPs),22 andthe ED is a largely unexplored segment in the con-tinuum of the care of older persons with respect tomedication. Indeed, the Geriatric Emergency Med-icine Task Force of the Society for Academic Emer-gency Medicine has started to address these issuesby creating a conceptual model of care for the elder

patient in which the emergency visit is viewedwithin broader medical, psychosocial, cognitive,and functional contexts.23 Our specific aims wereto determine the frequency of potentially inappro-priate medication selection for older persons in theED, the most common problematic drugs, the riskfactors for suboptimal medication selection, andthe relationship of potentially inappropriate med-ications to revisits and worse health status.

METHODS

Study Design. We performed a prospective co-hort study that included longitudinal patient sur-veys and chart review of the index visit to the ED.The study, including verbal informed consent, wasapproved by the University of Chicago Institu-tional Review Board.

Study Setting and Population. We prospec-tively identified all eligible patients (adults aged65 years or older) who presented to the ED at theUniversity of Chicago Hospitals, an urban aca-demic medical center, from October 1995 to June1996.24 In a pilot study, relatively few older pa-tients visited our ED after midnight. Therefore, weenrolled patients between the hours of 8:00 AM and12:00 midnight. Patients were included if they con-sented to be interviewed and completed the survey,and if, on chart review, the patient’s medications(upon presentation, given in the ED, given at dis-charge) were listed exactly.

Study Protocol. From the questionnaire in theED, we collected demographic data, including age,gender, marital status, race, and education, andwhether a patient thought that he or she could usemore help with everyday tasks. We also obtainedthe following health status and clinical data: Mini-Mental State Examination score (telephone ver-sion)25; deficiencies in six activities of daily living(ADLs)26; number of hospital admissions in thepreceding six months; Medical Outcomes Study(MOS) health-related quality of life scores onhealth perception, physical function, mentalhealth, social function, and pain, as adapted byBaker et al. for the ED setting (0–100 scale; 0worst, 100 best)27,28; and known health conditions.For the ADL and MOS questions, we asked aboutthe patient’s status around the time the patientfirst came to the ED and during the month priorto the acute illness. If patients were too ill orscored less than 17 points on the Mini-MentalState Examination, we attempted to administerthe survey to a proxy. At two weeks and again atthree months after discharge from the ED or hos-pital, trained research assistants administered ad-ditional surveys that contained the ADL and MOS


TABLE 1. Demographic and Clinical Characteristics of theStudy Population (n = 898)

DemographicAge (years) 76.3 6 7.9*

Gender—female 63%

Marital status—single/divorced/widowed 66%

RaceAfrican-AmericanWhiteOther

79%19%2%

Education<High school$High school

43%57%

Could use more help 35%

ClinicalAt presentation

Number of medications used 3.6 6 2.7*

In EDNumber of medications added 1.3 6 1.3*

Number of diseases 1.8 6 1.3*

Activities of daily living deficiencies 1.7 6 2.0*

Hospital admissions in preceding sixmonths 0.5 6 1.1*

Health-related quality of life scores†Health perceptionPhysical functionMental healthSocial functionPain

30.7 6 28.2*34.1 6 28.2*73.2 6 26.9*44.5 6 43.6*50.5 6 42.6*

*Mean 6 SD.†Health-related quality of life scores = Medical OutcomesStudy health-related quality of life scores one month prior tothe ED visit as adapted by Baker et al. for the ED setting (0–100 scale; 0 worst, 100 best).28

questions and inquired about revisits to the EDand hospital. At one year, the research assistantsasked about revisits. The surveys also asked howfrequently the patient missed taking prescribedmedication on a five-point scale, ranging from‘‘none of the time’’ to ‘‘all of the time.’’ From EDcharts, we gathered the presenting complaint,number of diseases, and medications prescribedupon admission, in the ED, and upon discharge.

Measurements. We used the 1997 Beers criteriaof inappropriate medications for the ambulatory el-der population (Appendixes A and B).20 These ex-plicit criteria were developed by a modified Delphiconsensus method with a panel of six nationallyrecognized experts in geriatrics, clinical pharma-cology, pharmacoepidemiology, clinical pharmacy,and psychopharmacology.

Potentially Inappropriate Medication Prescrip-

tions. The Beers criteria define drugs that shouldgenerally be avoided because they are ineffective

or pose unnecessarily high risk for older persons(Appendix A), and drugs that are appropriate touse in older persons only at certain doses, frequen-cies, or duration of therapies. We excluded fromour analysis those drugs defined by the Beers cri-teria as inappropriate only at certain doses (lora-zepam, oxazepam, alprazolam, temazepam, zolpi-dem, triazolam, digoxin, and iron supplements)since we did not have dose information in our da-tabase.

Potentially Adverse Drug–Disease Interac-

tions. The Beers criteria also include drugs thatshould not be used in older persons known to havespecific medical conditions. These criteria for po-tentially adverse drug–disease combinations cover15 conditions and diseases (Appendix B). For ex-ample, beta-blockers used in a patient with asthmacould worsen respiratory function. We identifiedpatients with these conditions at presentation tothe ED using data on initial complaints, past med-ical history, and known health conditions, and inthe ED using the same information plus ECG dataand discharge diagnoses.

We excluded the corticosteroid–diabetes inter-action, since we did not collect data on either du-ration of therapy (inappropriate only if corticoste-roid recently started) or types of preparations (lesslikely to be inappropriate if topical steroids used).We also excluded the aspirin–ulcer interactionsince we did not collect data on dosage. The Beerscriteria define this combination as inappropriateonly at aspirin doses higher than 325 mg.

Data Analysis. We recorded potentially inappro-priate medications and adverse drug–disease in-teractions. We examined the medications the pa-tients were taking upon presentation to the EDand the medications prescribed in and upon dis-charge from the ED. For each time period, we de-termined correlates of prescription of potentiallyinappropriate medications or adverse drug–dis-ease interactions. Sociodemographic candidatevariables included age, gender, race, education,marital status, and need for help. Clinical vari-ables included number of medications, number ofdiseases, baseline deficiencies in ADLs, number ofprior admissions in the previous six months, andrequiring a proxy for survey responses. In addi-tion, baseline MOS health-related quality of lifescores were candidate variables for a potentiallyinappropriate prescription upon presentation tothe ED, and health-related quality of life scores atthe time of the ED visit were covariates for a po-tentially inappropriate prescription in and upondischarge from the ED. Age was categorized into65–84 years and 85 years or more, and educationinto less than high school and high school graduate


TABLE 2. Numbers of Potentially Inappropriate MedicationPrescriptions*

InappropriateMedications

AtPresentation

Added by EmergencyPhysician

InED

Upon Dischargefrom ED†

AmitriptylineMethyldopaDiazepamDiphenhydramineIndomethacinOxybutyninPropoxyphene and

combination products

2014111098

8

195

167

2DipyridamoleHyoscyamineTiclopidineDicyclomineChlorpropamideHydroxyzineCyproheptadineDisopyramideFlurazepam

763221111

3

2

2

11

1

PromethazinePropanthelineReserpineMeperidineCyclobenzaprineMethohexitalTrimethobenzamide

111

751

3

1

TOTAL 107 35 23

*Beers criteria for inappropriate geriatric medication selec-tion.20 Some patients had more than one potentially inappro-priate prescription.†Nonadmitted patients.

or higher. In bivariate analyses, we used the chi-square test for categorical variables and the Wil-coxon rank sum test for continuous variables. Weused stepwise logistic regression for multivariateanalyses, with entry criterion p # 0.20 and staycriterion p # 0.05.

We then determined whether the medicationappropriateness criteria were correlated with re-visits to the ED or hospital, or death, as well asthe adapted MOS health-related quality of lifescores for health perception, physical function,mental health, pain, and social function during thethree months following the index visit. Death wasa secondary outcome. The primary independentvariable was whether the patient had been pre-scribed a potentially inappropriate medication oradverse drug–disease interaction. Separately, weanalyzed potentially inappropriate prescriptionsprior to the ED visit (comparison group was all pa-tients without potentially inappropriate medica-tions prior to the ED visit), in the ED (comparisongroup was all other patients except those receivingpotentially inappropriate medications at dischargefrom the ED), and upon discharge from the ED innonadmitted patients (comparison group was allother nonadmitted patients except those receivingpotentially inappropriate medications in the ED).

For the revisit analysis, our outcome was thenumber of days from discharge from the ED orhospital to the first revisit to the ED or hos-pital, or death. In bivariate analyses, we usedKaplan-Meier curves and the log-rank test.29,30 Inmultivariate analyses, we used Cox proportionalhazards regression with a forward stepwise ap-proach.31 The candidate variables were patient’sage, gender, race, education, marital status,whether the patient lives alone, type of residence,number of hospital admissions in the previous sixmonths, number of diseases, deficiencies in ADLsat baseline, and whether the patient needs morehelp. We checked the Cox models’ goodness of fitby examining log (-log) plots of the estimated eventrates and Schoenfeld residuals.32,33

In the analysis of health-related quality of life,we used Zeger and Liang’s generalized estimatingequations approach for repeated measures.34 Thismethod incorporates the baseline, ED, two-weekfollow-up, and three-month follow-up health statusscores in the regression model, and then calculateswhat covariates are independently associated withthe health-related quality of life over the entiretime period. We performed backwards eliminationwith the same covariates listed above. In analyzingthe effect of potentially inappropriate medicationsadded in or upon discharge from the ED, we usedthe baseline health-related quality of life scores ascovariates to adjust further for casemix differ-ences. We checked the goodness of fit of the gen-

eralized estimating equations models by examin-ing residual plots.

RESULTS

Study Population. One thousand eight hundredseven patients were registered in the ED duringthe study period. Of the 1,807 patients presentingto the ED, 1,205 (67%) consented to be in the studyand 981 (81%) completed a questionnaire. We re-viewed the ED charts of 941 (96%) patients, andwe excluded 43 patients whose charts indicatedthat the patient’s medications were ‘‘unknown,’’ orelse showed the class, as opposed to the name, ofa drug, either upon presentation, in the ED, orupon discharge. Thus, we included 898 subjects inthis study. Approximately half of the patients weredischarged from the ED to home, and half wereadmitted to the hospital.

The average age was 76.3 years, 63% were fe-male, 66% were single, divorced, or widowed, 79%were African-American, and 43% had less than ahigh school education (Table 1). The mean numberof diseases recorded per patient was 1.8. Aboutone-third reported they could have used more help,


TABLE 3. Numbers of Potentially Adverse Drug–DiseaseInteractions*

Diseases and DrugsAt

Presentation

Added by EmergencyPhysician

InED

Upon Dischargefrom ED†

AsthmaAtenololMetoprololTimolol

231

1

Blood-clotting disorders,on anticoagulation

AspirinDiclofenacIndomethacinNabumetoneNaproxen

71211

4

1 1

Chronic obstructive pul-monary disease

AlprazolamDiazepamLorazepamAtenololPindololTimolol

122111

DiabetesAtenololBetaxololMetoprololNadololPropranololTimolol

716122

Peptic ulcer diseaseDiclofenacIbuprofenIndomethacinPotassium chloride

1122

3

Peripheral vasculardisease

AtenololMetoprolol

31

TOTAL 55 5 5

*Beers criteria for inappropriate geriatric medication selec-tion.20 Some patients had more than one potentially inappro-priate prescription.†Non-admitted patients.

and 57% had at least one ADL deficiency. The av-erage number of medications used at the time ofpresentation to the ED was 3.6, with a range from0 to 19. More than two-thirds of the patients weregiven at least one medication in the ED. Most pa-tients reported taking the medications when re-surveyed at the two-week and three-month follow-ups.

Potentially Inappropriate Medication Prescrip-

tions. Overall, 133 of 898 patients (14.8%) had a

total of 165 potentially inappropriate medicationprescriptions (Table 2). At presentation, 95 of 898patients (10.6%) had a total of 107 potentially in-appropriate drug prescriptions. In the ED, 32 of898 patients (3.6%) were given a total of 35 poten-tially inappropriate medication prescriptions.Upon discharge from the ED, 23 (5.6%) of the 418nonadmitted patients were given 23 potentially in-appropriate medications. The five drugs most re-sponsible for potentially inappropriate prescrip-tions at presentation were amitriptyline,methyldopa, diazepam, diphenhydramine, and in-domethacin. In the ED, the four most commonlyprescribed potentially inappropriate medicationswere diphenhydramine, meperidine, indometha-cin, and cyclobenzaprine. The most common dis-charge diagnoses of patients for whom EPs addedpotentially inappropriate medications were mus-culoskeletal disorder (17%), back pain (15%), gout(15%), and allergy or urticaria (15%).

Potentially Adverse Drug–Disease Interac-

tions. Overall, 56 of 898 patients (6.2%) had atotal of 65 potentially adverse drug–disease inter-actions (Table 3). At presentation, 47 of 898 pa-tients (5.2%) had a total of 55 potentially adversedrug–disease interactions. In the ED, 5 of 898 pa-tients (0.6%) had a total of 5 potentially adversedrug–disease interactions. Upon discharge fromthe ED, 5 of 418 nonadmitted patients (1.2%) had5 potentially adverse drug–disease interactions.The medications most commonly responsible forthese possible interactions were beta-blockers,nonsteroidal anti-inflammatory drugs, aspirin, andsedative-hypnotics. Potentially adverse drug–dis-ease interactions were most common in patientswith diabetes, coagulopathy, and chronic obstruc-tive pulmonary disease.

Bivariate and Multivariate Correlates of Poten-

tially Inappropriate Medications and Adverse

Drug–Disease Interactions. Bivariate corre-lates (p # 0.05) of potentially inappropriate medi-cation prescriptions or adverse drug–disease in-teractions at presentation were the number ofdrugs used, the number of diseases, and worsebaseline health perception, physical function, men-tal health, social function, and pain. In the ED, thecorrelates were the number of drugs added and notrequiring a proxy for survey responses (Table 4).Age, gender, marital status, race, education, needfor more help with everyday tasks, deficiencies inADL, and number of hospital admissions in thepreceding six months were not significantly corre-lated. Multivariate correlates of those events atpresentation were the number of drugs used (ad-justed odds ratio per additional drug = 1.4, 95% CI= 1.3 to 1.5) and worse baseline pain (adjusted


TABLE 4. Bivariate Correlates of Potentially Inappropriate Medication Prescriptions or Adverse Drug–Disease Interactions*

Patients with PotentiallyInappropriate Prescriptions

or Interactions

Patients without PotentiallyInappropriate Prescriptions

or Interactions p-value

At presentation (n = 131) (n = 767)Number of drugs usedNumber of diseasesHealth perception‡Physical function‡Mental health‡Social function‡Pain‡

5.9 6 2.8†2.5 6 1.3†

40.8 6 26.6†47.5 6 28.6†78.1 6 23.5†60.0 6 40.3†60.3 6 37.8†

3.2 6 2.5†1.7 6 1.3†

46.6 6 29.1†54.4 6 28.7†82.5 6 21.7†68.7 6 38.5†71.1 6 36.6†

0.00010.00010.040.010.030.010.001

In EDNumber of drugs addedNot having proxy

(n = 36)2.6 6 1.6†

88.9%

(n = 850)1.3 6 1.3†

64.6%0.00010.01

*Beers criteria for inappropriate geriatric medication selection.20

†Mean 6 SD.‡Medical Outcomes Study health-related quality of life scores one month prior to the ED visit as adapted by Baker et al. forthe ED setting (0–100 scale; 0 worst, 100 best).28

odds ratio per 10 points = 1.07, 95% CI = 1.01 to1.13). In the ED, the independent correlates werethe number of drugs added (adjusted odds ratio peradded drug = 1.8, 95% CI = 1.5 to 2.2) and notneeding a proxy for survey responses (adjustedodds ratio = 4.6, 95% CI = 1.3 to 15.6).

Relationship of Inappropriate Medication Cri-

teria to Subsequent Revisits or Death, Death,

and Health-related Quality of Life. Survey fol-low-up rates at two weeks, three months, and oneyear were 80%, 78%, and 82%, respectively. We cal-culated three-month outcomes for the 868 patientsfor whom we had follow-up data. Within threemonths of the initial visit, 167 (19%) patients hadan ED revisit, 165 (19%) had a hospital admission,and 222 (26%) had either an ED revisit or hospitaladmission. Eighty-five (10%) patients died, and282 (33%) had either an ED revisit or hospitali-zation or death. Of those patients with a revisit tothe ED or a subsequent hospitalization or deathwithin three months, 22% had previously been pre-scribed a potentially inappropriate medication oradverse drug–disease interaction. However, the in-appropriate medication criteria were not indepen-dently associated with subsequent revisits ordeath, death, nor whether the patient respondedto the three-month follow-up survey, although con-fidence intervals were fairly broad (Table 5). Log(-log) plots of the estimated event rates and Schoen-feld residuals indicated that the proportional haz-ards models generally fit reasonably well. The ex-ception occurred for the analysis of revisit to theED or hospital, or death, within 90 days for pa-tients prescribed a potentially inappropriate med-ication in the ED. In this analysis, the standarderrors were large, but little evidence existed for aneffect of the covariate.

Prescription of a potentially inappropriate med-ication or adverse drug–disease interaction priorto the ED visit was associated with worse physicalfunction (23.5 points) and pain (27.3 points) dur-ing the four-month period around the emergencyillness (Table 5). Prescription of a potentially in-appropriate medication or adverse drug–diseaseinteraction in the ED was correlated with lowerphysical function (210.7 points) and pain (211.3points), and upon discharge from the ED wascorrelated with lower physical function (212.7points). To place these differences in perspective,normative data from the MOS Short Form-36 com-paring patients with and without comorbiditiesshowed that a six-point difference in physical func-tion score was similar to having either angina,back pain, or musculoskeletal complaints, and anine-point difference in pain score was comparableto hip impairment.35 Plots of the models’ residualsvs fitted variables did not reveal any systematicdeviation from randomness.

DISCUSSION

At presentation to the ED, 10.6% of our study pop-ulation were taking a potentially inappropriatemedication. A smaller number of patients wereprescribed a potentially inappropriate medicationin or upon discharge from the ED. Relatively fewpatients had potentially adverse drug–disease in-teractions. We were unable to demonstrate an as-sociation between the appropriateness criteria andsubsequent revisit to the ED or hospital, or death,but the prescription of potentially inappropriatedrugs was correlated with lower health-relatedquality of life scores.

Our outpatient incidence rate of 10.6% for po-


TABLE 5. Relationship of Inappropriate Medication Criteria (Potentially Inappropriate Medication or Adverse Drug–DiseaseInteraction) to Revisits, Death, and Health-related Quality of Life (n = 868)*

Outcomes

Prescribed Priorto ED Visit(n = 131)

Prescribedin the ED(n = 36)

Prescribed uponDischarge from the ED†

(n = 26)

Revisit to the ED or hospital, or death—adjusted hazard ratio (95% CI)‡ 1.31 (0.94, 1.83) 1.57 (0.87, 2.82) 0.62 (0.22, 1.70)

Death—adjusted hazard ratio (95% CI)‡ 1.04 (0.54, 2.00) 0.94 (0.23, 3.92) 3.31 (0.38, 28.72)

Health-related quality of life score—beta-coefficient (95% CI)§

Health perception 23.4 (26.9, 0.2) 22.6 (28.9, 3.7) 21.2 (211.2, 8.9)Physical function 23.5 (26.9, 20.1)} 210.7 (217.1, 24.4)} 212.7 (220.5, 24.8)}Mental health 21.8 (25.2, 1.7) 22.5 (28.6, 3.7) 20.8 (26.9, 5.3)Social function 0.2 (24.8, 5.1) 20.7 (211.6, 10.3) 211.9 (224.1, 0.4)Pain 27.3 (212.5, 22.0)} 211.3 (220.1, 22.6)} 26.7 (219.6, 6.1)

*Beers criteria for inappropriate geriatric medication selection.20

†Nonadmitted patients.‡Cox proportional hazards regression of number of days to outcome.§Health-related quality of life scores = Medical Outcomes Study health-related quality of life scores as adapted by Baker et al.for the ED setting (0–100 scale; 0 worst, 100 best).28 The multivariable models use the generalized estimating equations approachfor repeated measures.}p # 0.05.

tentially inappropriate medications is generallysimilar to other studies’ rates of inappropriatemedication use in elders. Stuck et al. had modifiedthe original 1991 Beers criteria and found that14% of community-residing persons aged 75 yearsand older in Santa Monica, California, were takingan inappropriate medication.16 With a subset of theoriginal Beers criteria limited to 23 drugs thatshould be avoided regardless of dosage or durationof use, Willcox et al. analyzed the 1987 NationalMedical Expenditure Survey and found that 23.5%of community-dwelling persons aged 65 years orolder received at least one inappropriate medica-tion.15 Using 20 medications from the originalBeers criteria as the standard, Futterman et al.determined that 11.5% of patients in a 1994 NewYork Medicare HMO database were taking an in-appropriate medication, and 17.5% of the patientsin the 1992 Medicare Current Beneficiary Surveywere prescribed one of the medications.17

Using implicit criteria from a validated Medi-cation Inappropriateness Index of ten rating ques-tions,36 Schmader et al. studied 208 geriatric out-patients and found that 14% of the 16,440 ratingsfor 1,644 medications were inappropriate.37 Liptonet al. applied implicit criteria incorporating inap-propriate choice of therapy, dosage, schedule,drug–drug interactions, therapeutic duplication,and allergy, and determined that more than 80%of older inpatients discharged on three or moremedications had at least one prescribing problem.38

We are aware of no other study of potentiallyinappropriate medications prescribed in the ED. Inour study, 3.6% of patients were prescribed an in-

appropriate medication in the ED, and 5.6% upondischarge from the ED. While these are lower ratesthan in the various community or inpatient stud-ies, the incidences are still significant given thelarge numbers of older persons presenting to theED. Moreover, major improvements in care mightresult from even brief training. The potentially in-appropriate medications prescribed in the ED werefrequently associated with musculoskeletal paindiagnoses and gout. These inappropriate medica-tions included indomethacin, meperidine, and cy-clobenzaprine. Thus, education of EPs might em-phasize the proper treatment of these commonpresenting complaints.7 Treatment of older pa-tients with chronic pain by a variety of physiciansis currently inadequate.39

At presentation to the ED, only 5.2% of ourstudy population was at risk for potentially ad-verse drug–disease interactions. Even fewer pa-tients were at risk for potentially adverse drug–disease interactions in (0.6%) or upon dischargefrom (1.2%) the ED. Persons with diabetes, coag-ulopathy, or chronic obstructive pulmonary diseasewere the most common patients who were pre-scribed a potentially adverse drug–disease inter-action, and beta-blockers, nonsteroidal anti-in-flammatory drugs, aspirin, and sedative-hypnoticswere the most frequent potentially inappropriateagents.

Goldberg et al. reported that 21.5% of theirstudy population, which consisted of patients at-tending the EDs of a community hospital and anurban academic hospital, had potentially adversedrug–disease interactions. Of these potential in-


teractions, 34% were related to treatment in theED.40 That study, however, was limited to a high-risk population of all patients taking three or moremedications, and patients 50 years or older takingat least two drugs. Accordingly, our lower incidenceof potentially adverse drug–disease interactionsfrom medications added in the ED may be due inpart to inclusion of some patients with less co-morbidity.

Prescription of multiple drugs was the mostconsistent independent predictor of a potentiallyinappropriate medication or adverse drug–diseaseinteraction. Investigators have found polyphar-macy to be a frequent correlate of adverse drugreactions.2

We are aware of two published studies of geri-atric patients that have linked inappropriate med-ication selection to worse outcomes. However, thestudies were limited by small sample size, or elseused especially sensitive criteria for appropriate-ness and adverse outcome. Schmader et al. studied208 older veterans receiving care at an outpatientclinic who took at least five drugs.18 In bivariateanalyses, they found that patients who had morehospital admissions and unscheduled ED or am-bulatory care visits over the subsequent 12 monthshad worse scores on a Medication AppropriatenessIndex.36 An insufficient number of outcomes ex-isted to perform multivariable analyses. Thus, noadjustment was made for confounding factors suchas comorbidity, severity of illness, or number ofmedications. Lindley et al. examined 416 admis-sions of older patients to a hospital and determinedthat 27% of the patients taking medications had151 adverse drug reactions.19 Of these reactions,half were from medications that were absolutelycontraindicated or else unnecessary. The face va-lidity of the appropriateness criteria was not clear.For example, 96 of the 151 adverse drug reactionswere from diuretics. Of these, 87 were asympto-matic biochemical abnormalities such as hypona-tremia and hypokalemia.

LIMITATIONS AND FUTURE QUESTIONS

We note some limitations to our study. It may notbe possible to generalize our results beyond our ur-ban academic medical center, nor to the patientswho decided not to enroll in the study. The major-ity of our study population was African-American,and many patients had less than a high school ed-ucation. In addition, we were unable to include pa-tients who presented to the ED between midnightand 8 AM, who may have been sicker. Because bothmedications and diseases in the ED questionnaireand charts were self-reported, we could have un-derestimated potentially inappropriate medication

prescriptions and adverse drug–disease interac-tions. In addition, self-reported adherence to med-ications upon follow-up may be inaccurate. Also,we may not have been able to adjust fully forcasemix differences in our multivariable analysescomparing patients with and without potentiallyinappropriate medications. It is possible that ouroutcome measures were too crude to detect somemeaningful problems with the use of these medi-cations, such as transient changes in mental statusthat could be dangerous, and we lacked adequatestatistical power for rarer outcomes such as mor-tality. In addition, we did not study adverse drug–drug interactions or possible underuse of appro-priate medications.41–44

Appropriateness criteria such as those of Beershave strengths and limitations as quality of carestandards for medication use, and these need to beacknowledged more clearly.45–50 Explicit processcriteria are not measures of actual clinical out-come,51 and the standards assume that some pa-tients will not tolerate the drugs. For example,some patients may have tolerated a potentially in-appropriate medication for years and are thusproperly continued on the drug. Or, it may be ac-ceptable to give patients certain potentially inap-propriate drugs if they are observed closely in theED, administered in low doses, or prescribed for ashort time period, such as diphenhydramine for anacute allergic reaction. These points are particu-larly important to consider when adapting criteriafor the appropriateness of medication selection tothe ED setting. In addition, wise clinical judgmentcould lead to the choice of potentially inappropri-ate medications for some patients. For example, di-abetic patients with ischemic heart disease or ahistory of myocardial infarction might benefit frombeta-blockers. Some healthy older persons may bephysiologically comparable to younger persons. Or,sometimes side effects are desirable such as ami-triptyline’s anticholinergic effects in a patient withParkinson’s disease. Still, for many of these medi-cations, safer substitutes for older persons existthat should be the preferred drugs. While theBeers criteria were not developed for older personsin the ED setting, they are nevertheless the mostcomprehensive, generally applicable, and up-to-date criteria now available to identify potentiallyinappropriate medication prescriptions and ad-verse drug–disease interactions in ambulatoryolder persons. They may be particularly useful asscreening tools for suboptimal care that could iden-tify patients whose medication regimens should beassessed in more detail. A variety of interventions,including computerized prescribing,52 provider ed-ucation,53 audit of charts with feedback to physi-cians,54 and use of a clinical pharmacist,55 have im-proved provider behavior.7 Future work needs to


adapt current geriatric medication appropriate-ness criteria to the ED setting, and to test whetheradherence to such guidelines leads to better healthoutcomes.

CONCLUSIONS

Suboptimal medication selection was relativelycommon both before and during a patient’s EDvisit. Quality of care standards for medication usein elders, such as the Beers criteria, need to bevalidated in more detailed studies. However, theguidelines generally have good face validity, andour study found that prescription of potentially in-appropriate medications was associated withworse health-related quality of life. Accordingly,primary care physicians and EPs should exercisecaution when prescribing for all older personssince the potential consequences of inappropriateprescriptions are severe.

References

1. Lazarou J, Pomeranz BH, Corey PN. Incidence of adversedrug reactions in hospitalized patients: a meta-analysis of pro-spective studies. JAMA. 1998; 279:1200–5.2. Nolan L, O’Malley K. Prescribing for the elderly. Part 1:Sensitivity of the elderly to adverse drug reactions. J Am Ger-iatr Soc. 1988; 36:142–9.3. Classen DC, Pestotnik SL, Evans RS, Lloyd JF, Burke JP.Adverse drug events in hospitalized patients: excess length ofstay, extra costs, and attributable mortality. JAMA. 1997; 277:301–6.4. Bates DW, Spell N, Cullen DJ, et al. The costs of adversedrug events in hospitalized patients. JAMA. 1997; 277:307–11.5. Johnson JA, Bootman JL. Drug-related morbidity and mor-tality: a cost-of-illness model. Arch Intern Med. 1995; 155:1949–56.6. Leape LL, Bates DW, Cullen DJ, et al. Systems analysis ofadverse drug events. JAMA. 1995; 274:35–43.7. Soumerai SB, McLaughlin TJ, Avorn J. Improving drug pre-scribing in primary care: a critical analysis of the experimentalliterature. Milbank Q. 1989; 67:268–317.8. Chutka DS, Evans JM, Fleming KC, Mikkelson KG. Drugprescribing for elderly patients. Mayo Clin Proc. 1995; 70:685–93.9. Beers MH, Ouslander JG. Risk factors in geriatric drug pre-scribing: a practical guide to avoiding problems. Drugs. 1989;37:105–12.10. Hanlon JT, Schmader KE, Koronkowski MJ, et al. Adversedrug events in high risk older outpatients. J Am Geriatr Soc.1997; 45:945–8.11. Chrischilles EA, Segar ET, Wallace RB. Self-reported ad-verse drug reactions and related resource use: a study of com-munity-dwelling persons 65 years of age and older. Ann InternMed. 1992; 117:634–40.12. Beers MH, Ouslander JG, Rollingher I, et al. Explicit cri-teria for determining inappropriate medication use in nursinghome residents. Arch Intern Med. 1991; 151:1825–32.13. Beers MH, Ouslander JG, Fingold SF, et al. Inappropriatemedication prescribing in skilled-nursing facilities. Ann InternMed. 1992; 117:684–9.14. Spore DL, Mor V, Larrat P, et al. Inappropriate drug pre-scriptions for elderly residents of board and care facilities. AmJ Public Health. 1997; 87:404–9.15. Willcox SM, Himmelstein DU, Woolhandler S. Inappropri-ate drug prescribing for the community-dwelling elderly.JAMA. 1994; 272:292–6.16. Stuck AE, Beers MH, Steiner A, et al. Inappropriate med-

ication use in community-residing older persons. Arch InternMed. 1994; 154:2195–200.17. Futterman R, Fillit H, Roglieri JL. Use of ineffective orunsafe medications among members of a Medicare HMO com-pared to individuals in a Medicare fee-for-service program. AmJ Man Care. 1997; 3:569–75.18. Schmader KE, Hanlon JT, Landsman P, et al. Inappropri-ate prescribing and health outcomes in elderly veteran outpa-tients. Ann Pharmacother. 1997; 31:529–33.19. Lindley CM, Tully MP, Paramsothy V, Tallis RC. Inappro-priate medication is a major cause of adverse drug reactionsin elderly patients. Age Ageing. 1992; 21:294–300.20. Beers MH. Explicit criteria for determining potentially in-appropriate medication use by the elderly: an update. Arch In-tern Med. 1997; 157:1531–6.21. Gurwitz JH. Suboptimal medication use in the elderly: thetip of the iceberg. JAMA. 1994; 272:316–7.22. Sanders AB. Care of the elderly in emergency depart-ments: conclusions and recommendations. Ann Emerg Med.1992; 21:830–4.23. Sanders AB (ed). Emergency Care of the Elder Person. St.Louis, MO: Beverly Cracom Publications, 1996.24. Chin MH, Jin L, Karrison TG, et al. Older patients’ health-related quality of life around an episode of emergency illness.Ann Emerg Med. 1999; 34:595–603.25. Roccaforte WH, Burke WJ, Bayer BL, Wengel SP. Valida-tion of a telephone version of the Mini-Mental State Exami-nation. J Am Geriatr Soc. 1992; 40:697–702.26. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW.The index of ADL: a standardized measure of biological andpsychosocial function. JAMA. 1963; 185:914–9.27. Ware JE Jr, Sherbourne CD. The MOS 36-item short-formhealth survey (SF-36). 1. Conceptual framework and item se-lection. Med Care. 1992; 30:473–83.28. Baker DW, Stevens CD, Brook RH. Patients who leave apublic hospital emergency department without being seen bya physician: causes and consequences. JAMA. 1991; 266:1085–90.29. Kaplan EL, Meier P. Nonparametric estimation from in-complete observations. J Am Stat Assoc. 1958; 53:457–81.30. Peto R, Pike MC, Armitage P, et al. Design and analysisof randomized clinical trials requiring prolonged observation ofeach patient. Part 2. Analysis and examples. Br J Cancer. 1977;35:1–39.31. Cox DR. Regression models and life tables (with discus-sion). J R Stat Soc B. 1972; 34:187–220.32. Kay R. Proportional hazard regression models and theanalysis of censored survival data. Appl Stat. 1977; 26:227–37.33. Schoenfeld D. Partial residuals for the proportional haz-ards regression model. Biometrika. 1982; 69:239–41.34. Zeger SL, Liang KY. Longitudinal data analysis for dis-crete and continuous outcomes. Biometrics. 1986; 42:121–30.35. Ware JE Jr, Kosinski M, Keller SD. SF-12: How to Scorethe SF-12 Physical and Mental Health Summary Scales, Sec-ond Edition. Boston, MA: The Health Institute, New EnglandMedical Center, 1995, p 51.36. Hanlon JT, Schmader K, Samsa G, et al. Method for as-sessing the appropriateness of drug therapy. J Clin Epidemiol.1992; 45:1045–51.37. Schmader K, Hanlon JT, Weinberger M, et al. Appropri-ateness of medication prescribing in ambulatory elderly pa-tients. J Am Geriatr Soc. 1994; 42:1241–7.38. Lipton HL, Bero LA, Bird JA, McPhee SJ. The impact ofclinical pharmacists’ consultations on physicians’ geriatricdrug prescribing: a randomized controlled trial. Med Care.1992; 30:646–58.39. Bernabei R, Gambassi G, Lapane K, et al. Management ofpain in elderly patients with cancer. JAMA. 1998; 279:1877–82.40. Goldberg RM, Mabee J, Chan L, Wong S. Drug–drug anddrug–disease interactions in the ED: analysis of a high-riskpopulation. Am J Emerg Med. 1996; 14:447–50.41. Beers MH, Storrie M, Lee G. Potential adverse drug in-teractions in the emergency room: an issue in the quality ofcare. Ann Intern Med. 1990; 112:61–4.42. Herr RD, Caravati EM, Tyler LS, Iorg E, Linscott MS. Pro-spective evaluation of adverse drug interactions in the emer-gency department. Ann Emerg Med. 1992; 21:1331–6.


43. Karas S Jr. The potential for drug interactions. Ann EmergMed. 1981; 10:627–30.44. Chin MH, Friedmann PD, Cassel CK, Lang RM. Differ-ences in generalist and specialist physicians’ knowledge anduse of angiotensin-converting enzyme inhibitors for congestiveheart failure. J Gen Intern Med. 1997; 12:523–30.45. Brook RH, Kamberg CJ, Mayer-Oakes A, Beers MH,Raube K, Steiner A. Appropriateness of acute medical care forthe elderly: an analysis of the literature. Health Policy. 1990;14:225–42.46. Buetow SA, Sibbald B, Cantrill JA, Halliwell S. Appropri-ateness in health care: application to prescribing. Soc Sci Med.1997; 45:261–71.47. Naylor CD. What is appropriate care? N Engl J Med. 1998;338:1918–20.48. Miller AE Jr. Mismedicating the elderly—says who? J FamPract. 1995; 41:331–6.49. Hoehn JB, Krinksy D, Beary JF, et al. Inappropriate drugprescribing [letters]. JAMA. 1995; 273:455–7.

50. Naylor CD. Grey zones of clinical practice: some limits toevidence-based medicine. Lancet. 1995; 345:840–2.51. Kahn KL, Rogers WH, Rubenstein LV, et al. Measuringquality of care with explicit process criteria before and afterimplementation of the DRG-based prospective payment sys-tem. JAMA. 1990; 264:1969–73.52. Schiff GD, Rucker TD. Computerized prescribing: buildingthe electronic infrastructure for better medication usage.JAMA. 1998; 279:1024–9.53. Avorn J, Soumerai SB, Everitt DE, et al. A randomizedtrial of a program to reduce the use of psychoactive drugs innursing homes. N Engl J Med. 1992; 327:168–73.54. Hamdy RC, Moore SW, Whalen K, et al. Reducing poly-pharmacy in extended care. South Med J. 1995; 88:534–8.55. Hanlon JT, Weinberger M, Samsa GP, et al. A randomized,controlled trial of a clinical pharmacist intervention to improveinappropriate prescribing in elderly outpatients with poly-pharmacy. Am J Med. 1996; 100:428–37.

APPENDIX A

1997 Beers Criteria for Inappropriate Medication Prescriptions

PainPropoxyphene and combination productsIndomethacinPhenylbutazonePentazocineMeperidine

AntiemeticTrimethobenzamide

Muscle relaxant/antispasmodicMethocarbamolCarisoprodolOxybutyninChlorzoxazoneMetaxaloneCyclobenzaprine

BenzodiazepinesFlurazepamLorazepam >3 mg qdOxazepam >60 mg qdAlprazolam >2 mg qdTemazepam >15 mg qdZolpidem >5 mg qdTriazolam >0.25 mg qdChlordiazepoxideDiazepam

AnxiolyticMeprobamate

AntidepressantAmitriptylineDoxepin

CardiacDigoxin >0.125 mg except for atrial arrhythmiasDipyridamole

AntihypertensiveMethyldopaReserpine

MiscellaneousDisopyramideChlorpropamideAnticholinergic gastrointestinal antispasmodic drugsAnticholinergic antihistaminesDiphenhydramineErgot mesyloidsCyclospamolIron supplements >325 mgAll barbiturates except phenobarbitolTiclopidine

(Appendix B follows on the next page.)


APPENDIX B

1997 Beers Criteria for Potentially Inappropriate Drug–Disease Interactions

Disease or Condition Drug

Heart failure DisopyramideDrugs with high sodium content

Diabetes Beta-blockersCorticosteroids (recently started)

Hypertension Diet pillsAmphetamine

Chronic obstructive pulmonary disease Beta-blockersSedative-hypnotics

Asthma Beta-blockers

Peptic ulcer disease Non-steroidal anti-inflammatory drugsAspirin (>325 mg)Potassium supplements

Seizures or epilepsy Clozapine, thorazine, thioridazine,ChlorprothixeneMetoclopramide

Peripheral vascular disease Beta-blockers

Blood-clotting disorders, on anticoagulation AspirinNon-steroidal anti-inflammatory drugsDipyridamole, ticlopidine

Benign prostatic hypertrophy Anticholinergic antihistaminesGastrointestinal antispasmodic drugsMuscle relaxantsOpioid drugs (including propoxyphene)Flavoxate, oxybutyninBethanecholAnticholinergic antidepressant drugs

Incontinence Alpha-blockers

Constipation Anticholinergic drugsOpioid drugsTricyclic antidepressant drugs

Syncope or falls Beta-blockersLong-acting benzodiazepine drugs

Arrhythmias Tricyclic antidepressant drugs

Insomnia DecongestantsTheophyllineDesipramine, selective serotonin reuptake inhibitors, methylphenidate,

monoamine oxidase inhibitorsBeta-agonists

Appropriateness of Medication Selection for Older Persons in an Urban Academic Emergency Department

Documents

Transcript of Appropriateness of Medication Selection for Older Persons in an Urban Academic Emergency Department