Inappropriate Medication Administration to the Acutely Ill Elderly: A Nationwide Emergency Department Study, 1992–2000


  • Jeffrey M. Caterino MD,

    1. From the *Department of Emergency Medicine, Allegheny General Hospital, Pittsburgh, PennsylvaniaDepartment of Emergency Medicine, Massachusetts General Hospital, Boston, MassachusettsChanning Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts.
    Search for more papers by this author
  • Jennifer A. Emond MS,

    1. From the *Department of Emergency Medicine, Allegheny General Hospital, Pittsburgh, PennsylvaniaDepartment of Emergency Medicine, Massachusetts General Hospital, Boston, MassachusettsChanning Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts.
    Search for more papers by this author
  • Carlos A. Camargo Jr MD, DrPH

    1. From the *Department of Emergency Medicine, Allegheny General Hospital, Pittsburgh, PennsylvaniaDepartment of Emergency Medicine, Massachusetts General Hospital, Boston, MassachusettsChanning Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts.
    Search for more papers by this author

  • The project was supported, in part, by an Emergency Medicine Foundation Center of Excellence Award (Dallas, TX). Abstract submitted to the Society for Academic Emergency Medicine Annual Meeting, May 2004, Orlando, FL.

Jeffrey M. Caterino, MD, Department of Emergency Medicine, Ohio State University, Columbus, OH. E-mail:


Objectives: To determine the national rate and trend of inappropriate medication administration to elderly emergency department (ED) patients. Secondary objectives were to identify risk factors for receiving an inappropriate medication and to determine whether administration is sometimes justified based on diagnosis.

Design: Retrospective analysis of ED visits in the 1992–2000 National Hospital Ambulatory Medical Care Survey. Inappropriate medications identified using Beers' 1997 explicit criteria.

Setting: EDs of U.S. noninstitutionalized general and short-stay hospitals.

Participants: ED survey patients aged 65 and older.

Measurements: Magnitude and rate of administration of 36 medications.

Results: Inappropriate medications were administered in an estimated 16.1 million (95% confidence interval (CI)=14.9–17.3 million) or 12.6% (95% CI=11.6–13.5%) of elderly ED visits from 1992 to 2000. The rate of inappropriate administration was unchanged throughout the study period (P=.40). Six drugs accounted for 70.8% of inappropriate administration: promethazine (22.2%), meperidine (18.0%), propoxyphene (17.2%), hydroxyzine (10.3%), diphenhydramine (7.1%), and diazepam (6.0%). In multivariate analysis, number of ED medications was the strongest predictor, with an odds ratio for two to three medications of 6.0 (95% CI=5.3–6.7) and for four to six medications of 8.1 (95% CI=7.2–9.2). Diagnoses indicating potentially appropriate uses of these medications were rarely present. For example, only 42.4% of patients receiving diphenhydramine and 7.4% receiving hydroxyzine were diagnosed with an allergic process.

Conclusion: Elderly ED patients are frequently administered inappropriate medications. Potentially appropriate uses of generally inappropriate drugs cannot account for such administrations. Inappropriate administration rates remain unchanged despite the 1997 publication of explicit criteria.

Adverse drug reactions in the elderly are of growing concern because of the increasing elderly population and the increasing numbers of medications prescribed to this group.1,2 Physiological changes of aging result in altered drug pharmacokinetics and pharmacodynamics, increasing the risk of adverse drug reactions. Polypharmacy and the presence of comorbid diseases are also common.1,3–5 As a result, the risk/benefit ratio for many drugs is unfavorable in the elderly. Administering high-risk, inappropriate medications to elderly patients increases the risk of drug side effects, morbidity, and mortality.6–18

A 1995 General Accounting Office report noted persistently high rates of high-risk medication administration to the elderly.19 In 1997, Beers proposed explicit criteria to identify inappropriate medication use in elderly patients. The criteria were based on a consensus of six nationally recognized experts using a modified Delphi consensus technique.20 The 1997 criteria were a modification of criteria initially published in 1991 and were intended to be broadly applicable to elderly populations.21 They included medications that should generally be avoided in the elderly because of an unfavorable risk/benefit ratio or to the availability of safer, equally effective alternatives. Additional medications were considered inappropriate at certain dosages, after certain duration of administration, or in the presence of certain comorbid conditions.

The 1991 and 1997 Beers criteria have been studied in office-based clinic patients, nursing home residents, and in the community-dwelling elderly,4,10,12,22–32 but there is little information on administration of inappropriate medications in the acutely ill elderly, for example in elderly patients in the emergency department (ED), other than a single study that applied the Beers criteria in one ED.18 To the authors' knowledge, no previous study has examined the national incidence of inappropriate medication administration in elderly ED patients.

This study used the National Hospital Ambulatory Medical Care Survey (NHAMCS). The primary aim was to determine the national rate of inappropriate medication administration to elderly ED patients, as defined using the Beers criteria, identifying changes in the rate of inappropriate administration from 1992 through 2000. The secondary aims were to identify risk factors for receiving an inappropriate medication and determine whether the administration of selected medications is sometimes justified based on the patient's diagnoses.


Survey Characteristics and Study Design

This observational study used a cross-sectional survey design. Data were obtained from the ED component of the 1992 through 2000 NHAMCS. The NHAMCS is a national probability sample survey of hospital outpatient and ED visits designed by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention. NHAMCS data are available for public use and contain no patient identifiers. The survey uses a four-stage probability design that randomly samples primary sampling units (a county, group of counties, town, or metropolitan statistical area), hospitals within sampling units, EDs within hospitals, and patient visits within EDs. Noninstitutional, nonfederal general and short-stay (<30 day) hospitals in the 50 states and District of Columbia are eligible.33,34 The plan and operation of NHAMCS are more fully described elsewhere.33–35 Participation has historically been high, with a 93% to 97% response rate and approximately 400 participating EDs each year. Approximately 25,000 ED patient-visit encounter forms are completed each year. Each patient-visit encounter is provided with a visit weight that can then be extrapolated to produce national estimates.35

Trained field representatives of the Bureau of the Census, who enroll the hospitals and train hospital staff, oversee survey data collection. Hospital staff perform all visit sampling and data collection over a 4-week period in each institution. A standardized data collection form is completed for each patient. Quality control during data collection consists of weekly field representative visits and review of records. Quality control during data processing is accomplished by reviewing a sample of visit records.33,34

For the present analysis, the study population included all ED patients aged 65 and older in the 1992 through 2000 NHAMCS. The hospital institutional review board approved the study protocol.

Survey Content and Data Abstraction

ED survey content included demographic information, insurance information, reasons for visit (maximum of 3), presence of an injury, diagnoses (maximum of 3), testing and procedures performed, medications administered in the ED (maximum of 5 through 1994, then maximum of 6), and disposition.33–35 Hospital data were also collected, such as location, metropolitan or nonmetropolitan area, and ownership type. Diagnoses were classified according to International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes.36 Medications were coded according to a unique scheme developed at the NCHS.37 Geographic region and metropolitan statistical area categories represent standardized geographic divisions defined by the U.S. Census Bureau.38

Potentially inappropriate medications for the elderly were identified using the 1997 Beers criteria.20 Thirty-six drugs considered generally inappropriate in the elderly were included in the study (Table 1). Dipyridamole was excluded from this list because of post-1997 evidence demonstrating potential benefits. Drugs considered inappropriate based on dosage, duration of administration, or the presence of comorbid conditions were not included because NHAMCS does not contain these data.20

Table 1. 1997 Beers Explicit Criteria for Medications Considered Generally Inappropriate in the Elderly
Drug ClassGeneric NameReason for Inclusion
  1. NSAIDs=nonsteroidal antiinflammatory drugs.

Long-acting benzodiazepines and other sedativesDiazepam, Chlordiazepoxide, Flurazepam, MeprobamateHave a long half-life in the elderly (often several days), producing prolonged sedation and increasing the risk of falls. Short- and intermediate-acting benzodiazepines are preferred.
Barbiturates (except phenobarbital)Amobarbital, Pentobarbital, SecobarbitalCause more side effects than most other sedative drugs in the elderly and are highly addictive.
AntidepressantsAmitriptyline, DoxepinStrong anticholinergic and sedating properties. Rarely the antidepressant of choice.
NarcoticsMeperidine, Propoxyphene, PentazocineMeperidine: Not an effective oral analgesic and has many disadvantages compared with other narcotic(s).
  Propoxyphene: Few advantages over acetaminophen, side effects of narcotic(s).
NSAIDSIndomethacin, PhenylbutazoneIndomethacin: Of all NSAIDS, produces the most central nervous system side effects. Phenylbutazone: May produce serious hematologic side effects.
AntiemeticTrimethobenzamideOne of the least effective anti-emetics, causes extrapyramidal side effects.
AntispasmodicsCyclobenzaprine, Methocarbamol, Carisoprodol, Oxybutynin, Chlorzoxazone, MetaxalonePoorly tolerated by the elderly, leading to anticholinergic side effects, sedation, and weakness; effectiveness at doses tolerated by the elderly is questionable.
AntiarrhythmicDisopyramideOf all anti-arrhythmic drugs, the most potent negative inotrope and may induce heart failure; also strongly anticholinergic.
AntihypertensiveMethyldopa, ReserpineMethyldopa: May cause bradycardia and exacerbate depression.
  Reserpine: Induces depression, impotence, sedation, and orthostasis.
HypoglycemicChlorpropamideCan cause prolonged, severe hypoglycemia and syndrome of inappropriate secretion of antidiuretic hormone.
Gastrointestinal antispasmodicsDicyclomine, Hyoscyamine, Propantheline, BelladonnaHighly anticholinergic and generally produce substantial toxic effects in the elderly. Additionally, their effectiveness is questionable. Best avoided in the elderly, especially for long-term use.
AntihistaminesDiphenhydramine, Chlorpheniramine, Hydroxyzine, Cyproheptadine, Promethazine, Tripelennamine, DexchlorpheniramineAll have potent anticholinergic properties. Preparations are available without antihistamines, and these are safer substitutes in the elderly. Diphenhydramine usually should not be used in the elderly. When used to treat allergic reactions, it should be used in the smallest possible dose and with great caution.

Public-use data files were obtained from the NCHS for 1992 through 2000. Each patient was determined to have been administered an inappropriate medication if their record contained one of the five-digit generic medication codes corresponding to an inappropriate medication in at least one of the six medication fields (specific medication codes available on request). Single-ingredient and combination drugs were considered.

Visit characteristics were identified for the elderly population overall and for those receiving an inappropriate medication. Visits before 1997 were coded as urgent/emergent or nonurgent. To keep analyses between earlier and later years consistent, visits that occurred after a change in coding in 1997 were coded as urgent/emergent if recorded as a wait time less than 15 minutes or 15 to 60 minutes and as nonurgent if recorded as more than 1 to 2 hours. Urgency and wait times were as determined by ED triage at each institution. ICD-9-CM diagnoses for selected medications were analyzed and population estimates derived for the most common diagnoses. Each diagnosis for a data set entry (maximum of 3 diagnoses) was included in the analysis. Similar ICD-9-CM codes were grouped to facilitate analysis.

Statistical Analysis

All analyses were performed using Stata 7.0 (Stata Corp., College Station, TX). National estimates of elderly visits and inappropriate medication administration were derived using the sampling weight provided for each patient visit encounter. Rates per 1,000 ED visits were calculated using projected NHAMCS estimates. Confidence intervals (CIs) were calculated according to a method recommended by NCHS using the relative standard error (SE) of each estimate.33–35

Standards previously identified by NCHS for acceptable reliability in survey estimates were adhered to. NCHS considers an estimate to be unreliable if it has a relative SE of more than 30%. In addition, estimates based on fewer than 30 records are considered inherently unreliable, regardless of their SE.33–35

Chi-square analysis was used to assess changes from 1992 through 2000 in the overall rate of medication administration and in the rates of individual, commonly administered medications. Multivariate logistic regression was used to evaluate independent predictors of inappropriate administration. Two-sided P-values less than .05 were considered statistically significant.


The 1992–2000 NHAMCS ED data set included 228,832 ED patient data records representing an estimated 866.5 million ED visits. Of these records, 33,395 were in patients aged 65 and older, representing an estimated 128 million visits (14.8%). Over this time, the yearly estimated number of overall ED visits steadily increased from 89.8 million to 108 million, and the yearly number of elderly ED visits steadily increased from 12.7 million to 16.2 million.

Medications were administered in 72% (95% CI=71–73%) of elderly ED visits. Inappropriate medications were administered during an estimated 16,073,000 visits (95% CI=14,868,000–17,279,000) or 12.6% (95% CI=11.6–13.5%) of all elderly visits over the study period. Twenty percent (95% CI=17–23%) of patients administered an inappropriate medication received more than one inappropriate drug. As a result, there were a total of 19,454,000 (95% CI=17,487,000–21,422,000) inappropriate drug administrations.

There was no statistically significant change over the study period in the rate of elderly patients receiving an inappropriate medication. From 1992 to 1994, 12.1% (95% CI=11.5–12.8; 4.7 million visits) of elderly patients received an inappropriate medication. From 1995 to 1997, the rate was 12.9% (95% CI=12.2–13.7; 5.5 million visits), and from 1998 to 2000 the rate was 12.6% (95% CI=11.8–13.3; 6.0 million visits).

Table 2 lists overall estimated administrations for individual medications. The six most commonly administered individual medications accounted for 70.8% of the total and included promethazine (22.2%), meperidine (18.0%), propoxyphene (17.2%), hydroxyzine (10.3%), diphenhydramine (7.1%), and diazepam (6.0%). Out of these six medications, only diazepam showed a statistically significant decrease in the rate of administration during the study period, from 10.0 per 1,000 visits (95% CI=8.6–11.3) in 1995–1997 to 6.9 per 1,000 visits (95% CI=5.8–7.9) in 1998–2000 (P=.04). Rates for the other commonly administered medications remained stable or increased slightly (data not shown) from 1992 through 2000.

Table 2. Specific Drug Mentions for Elderly Patients Receiving Inappropriate Medications in the Emergency Department, 1992–2000
Drug Type*Absolute n of Cases in Data SetEstimated Total Administrations in Thousands (95% CI)Proportion of Administered Inappropriate Medications %
  • *

    Estimates from drugs with <30 data set mentions or with relative standard error >30% are unreliable, and these drugs are omitted from the table.

  • Of 19,454,000 total inappropriate administrations.

Long-acting benzodiazepine
 Diazepam3001,161 (867–1,455)6.0
 Meperidine7823,505 (3,099–3,912)18.0
 Propoxyphene8023,336 (2,882–3,790)17.2
 Cyclobenzaprine107473 (318–629)2.4
 Methocarbamol39152 (68–235)0.8
 Oxybutynin42195 (116–274)1.0
Antihypertensive-Methyldopa47205 (111–299)1.1
Gastrointestinal antispasmodic
 Dicyclomine95323 (198–448)1.7
Hyoscyamine192704 (360–1,047)3.6
 Diphenhydramine3501,375 (1,106–1,644)7.1
 Hydroxyzine5492,002 (1,650–2,355)10.3
 Promethazine1,0354,327 (3,852–4,802)22.2

Results of the multivariate model are noted in Table 3. The strongest predictor of inappropriate medication administration was the number of medications administered in the ED. Other patient characteristics that were predictors of inappropriate medication use included age younger than 85, female sex, and white race. Predictive visit characteristics included region of the country other than the northeast, nonurgent visit, and discharge from the ED. Patients in government-run hospitals were less likely to receive an inappropriate medication. Ethnicity, health maintenance organization status, insurance status, and metropolitan/rural status did not predict inappropriate medication administration.

Table 3. Multivariate Predictors of Inappropriate Medication Administration in Elderly Emergency Department (ED) Patients, 1992–2000
CharacteristicRate per 1,000 ED Visits (95% CI)Odds Ratio95% CIP-value
  1. CI=confidence interval; MSA=metropolitan statistical area; HMO=health maintenance organization.

 ≥85102 (85–118)1Referent 
 75–84123 (110–136)1.21.04– 1.4.010
 65–74138 (125–152)1.31.2–1.5<.001
 Male99 (88–110)1Referent 
 Female143 (131–156)1.61.4–1.7<.001
 Black105 (87–123)1Referent 
 White129 (119–139)1.31.2–1.6<.001
 Other95 (48–143)1.10.8–1.6.519
 Non-Hispanic126 (116–136)1Referent 
 Hispanic114 (76–152)0.90.7–1.1.210
Insurance type
 Private133 (112– 154)1  
 Public125 (115–135)0.90.8–1.0.082
 Other134 (33–236)10.7–1.2.696
 Self-pay111 (51–171)0.90.7–1.3.670
HMO status
 Non-HMO member123 (60–185)1Referent 
 HMO member122 (102–141)0.90.8–1.1.330
Hospital ownership
 Voluntary nonprofit124 (113–134)1Referent 
 Government, nonfederal118 (102–134)0.80.7–0.9<.001
 Proprietary138 (113–163)10.9–1.1.970
MSA status
 Metropolitan area120 (109–130)1Referent 
 Non-MSA140 (125–155)1.21.0–1.2.178
U.S. geographical region
 Northeast93 (79–107)1Referent 
 Midwest124 (108–140)1.51.3–1.7<.001
 South149 (134–165)1.91.7–2.2<.001
 West123 (106–140)1.31.1–1.5.001
Visit urgency
 Urgent/emergent120 (109–130)1Referent 
 Nonurgent136 (122–149)1.31.1–1.3<.001
Number of medications received in ED
 0–143 (37–50)1Referent 
 2–3205 (185–226)6.05.3–6.7<.001
 4–6244 (219–270)8.17.2–9.2<.001
Admission status
 Admitted101 (89–114)1Referent 
 Discharged, other142 (130–154)1.61.5–1.8<.001

The most common ICD-9-CM diagnoses for six inappropriate medications are noted in Table 4. As an example of appropriate prescribing, the diagnosis of allergic reactions in patients receiving antihistamines was examined. Of those receiving diphenhydramine, only 42.4% were diagnosed with an allergic reaction. Of patients receiving hydroxyzine, only 7.4% were diagnosed with an allergic reaction.

Table 4. Most Common International Classification of Disease, Ninth Edition, Clinical Modification, Diagnoses for Selected Inappropriate Medications
ED Diagnosis (Not Exclusive)*Overall Estimated Administrations in Thousands (95% CI)
  • *

    Diagnoses are not exclusive (i.e., for each data set entry, each available diagnosis is counted).

  • Estimates without confidence intervals (CIs) are based on <30 data set entries or have relative standard error >30%.

  • Including allergic reaction, urticaria, anaphylaxis, contact dermatitis, skin rash, venom exposure, and drug reaction.

  • na=not applicable.

Diazepam1,161 (867–1455)
Respiratory failure/pulmonary edema438 (273–603)
Sprains/strains/fractures/dislocations160 (84–235)
Vertigo/presyncope/syncope101 (na)
Seizure100 (na)
Anxiety59 (na)
Diphenhydramine1,375 (1,106–1,644)
Allergic reaction583 (431–735)
Acute coronary syndrome/congestive heart failure132 (na)
Fractures and contusions93 (na)
Hydroxyzine2,002 (1,650–2,355)
Gastrointestinal disease/abdominal pain347 (191–503)
Back or neck strain213 (117–310)
Allergic reaction149 (85–213)
Headache/migraine headache119 (na)
Anxiety118 (na)
Meperidine3,505 (3,099–3,912)
Fracture or dislocation907 (664–1,151)
Gastrointestinal disease530 (388–673)
Abdominal pain, nausea, or vomiting499 (379–619)
Back or neck strain417 (303–532)
Extremity contusion/sprain/strain307 (208–405)
Genitourinary disease281 (170–392)
Promethazine4,327 (3,852–4,802)
Gastrointestinal disease819 (634–1,004)
Abdominal pain427 (326–529)
Genitourinary disease397 (258–534)
Nausea or vomiting329 (231–429)
Back or neck strain266 (177–355)
Vertigo/syncope/pre-syncope267 (147–387)
Propoxyphene3,336 (2,882–3,790)
Extremity contusion/sprain/strain800 (616–985)
Fracture or dislocation581 (370–793)
Back or neck strain426 (288–565)


This study provides the first national estimate of inappropriate medication administration to elderly ED patients. From 1992 through 2000, an estimated 12.6% of elderly ED patients received an inappropriate medication, representing an estimated 16,073,000 visits. This rate has remained constant despite the 1997 publication by Beers of explicit criteria intended to identify inappropriate medication administration in the elderly. The Beers criteria are the most widely accepted explicit criteria for inappropriate medication use.19–39 The portion of the criteria studied included medications that should be generally avoided in the elderly because of an unfavorable risk-benefit ratio or the availability of safer, effective alternatives.

Studies of elderly patients have estimated that adverse drug reactions cause up to 17% of admissions, as well as an annual 32,000 hip fractures.40,41 Inappropriate medication use contributes to these adverse drug reactions,6,11,13,15,18 and ED patients may be particularly vulnerable to adverse events because of patient acuity, physician unfamiliarity with the patient, and lack of follow-up. Studies in acutely ill hospitalized and ED patients have generally focused on drugs contraindicated because of drug-drug interactions or the presence of comorbidities.9,14,16,42–45 Unfortunately, although application of the Beers criteria has been widespread in office-based outpatient settings, it has rarely been used in the acutely ill elderly patient.

To the authors' knowledge, no study has applied the Beers criteria to determine the rate of inappropriate prescribing in inpatients, and only one author has applied it to an ED population.18 In a study in a single Chicago ED, 10.6% of patients were taking an inappropriate medication as outpatients, 3.6% received one in the ED, and 5.6% were prescribed one on discharge. The lower rate of inappropriate ED administration in this ED than in the national rate as found in the current study is primarily due to different patterns of drug administration. In that single ED, there were no administrations of propoxyphene or promethazine, whereas these drugs accounted for 39% of inappropriate medication administrations in the current study. Similar interfacility variation in prescribing has been noted in other venues, such as nursing homes.23

Studies in office-based outpatient populations have found varying rates of inappropriate medication use. One group of investigators conducted two studies using medications considered generally inappropriate by Beers.28,29 One study used the outpatient department portion of the 1994 NHAMCS, and the other used the 1992 National Ambulatory Medical Care Survey (NAMCS), a national survey of physician office visits. Prescription of inappropriate medications occurred in 2.9% and 5.0% of patient visits, respectively. Another group found a rate of inappropriate prescribing of 10.8% using the 1997 NAMCS but included in the analysis drugs identified by Beers as inappropriate in the presence of certain medical conditions.12

Studies using the entire Beers criteria and examining longer periods or institutionalized populations have generally found higher rates of inappropriate medication use. Over periods ranging from 1 month to 1 year, rates of inappropriate medication prescribing have ranged from 11.5% to 21.3%.4,10,30 Up to 24% of residents of long-term care or assisted living facilities and 40% of nursing home patients have received an inappropriate medication.23,32,43

The current study indicated a greater rate of inappropriate medication administration in ED patients than the previously mentioned outpatient studies, which also used national databases recording single patient visits.12,28,29 This finding is likely due to multiple factors, such as differences in ED patient characteristics, chief complaints, diagnoses, and acuity. For example, the most commonly administered inappropriate drugs in the ED differed from those in office-based outpatient populations. In both settings, propoxyphene, diazepam, diphenhydramine, and hydroxyzine were common. The most commonly administered ED drugs included meperidine and promethazine, drugs rarely used in other outpatient settings.12,27–30,32 Given the higher rate of inappropriate administration, ED physicians and other physicians caring for acutely ill patients should be particularly cautious in prescribing potentially inappropriate medications, especially because such patients may be more vulnerable to adverse drug effects.

The different rates of inappropriate prescribing may also reflect differences in ED physician training and practice patterns or unfamiliarity of ED physicians with the Beers criteria. Other than the previously mentioned study,18 the Beers criteria are not represented in the emergency medicine literature. Expansion of the Beers criteria into the emergency medicine setting has the potential to dramatically decrease inappropriate medication administration. Methods such as educational programs or computerized medication review have been effective in decreasing inappropriate prescribing.46–48 The lower overall rate in the ED study previously mentioned18 was primarily due to eliminating administration of only two medications, providing evidence that interventions targeted at a small number of medications could substantially decrease inappropriate ED prescribing. In one report, cooperation between the ED and pharmacy using a combination of education, care pathways, computer review, and formulary changes led to a 58% reduction in inappropriate prescribing of Beers criteria medications in hospitalized patients.48 Expansion of such programs into an ED setting, targeted at the relatively small number of medications that constitute the majority of inappropriate prescribing, is recommended. Additionally, given the high prevalence of inappropriate administration in this acutely ill population, efforts to quantify and reduce inappropriate medication administration in similar populations, such as inpatients, should be undertaken.

It might be argued that explicit reviews such as this overstate the overall rate of inappropriate medication administration because these medications may sometimes be used in clinically appropriate situations, but based on analysis of ICD-9-CM diagnoses, the use of medications in such circumstances cannot explain the high rates of inappropriate medication administration. Beers recognized that certain medications may be appropriate in rare, select circumstances.20 To address this concern, another study used a consensus panel of experts to identify potentially appropriate indications for Beers criteria medications.30 Of the most commonly administered medications, they concluded that meperidine should always be avoided in the elderly. Medications considered rarely appropriate included diazepam (for alcohol withdrawal or muscle spasm) and propoxyphene (in patients previously taking the drug who expressed a strong preference for continuation). In the current study, only an estimated 160,000 of the 1.2 million patient visits in which diazepam was administered were for sprain, strain, fracture, or dislocation. Although it was not possible to determine reasons for propoxyphene administration, it is unlikely that large numbers of the 3.3 million patients receiving the drug met the proposed criteria for acceptable administration. According to one study,30 drugs with some indications include antihistamines (for allergic reactions) and promethazine (specific indication unspecified), but only 42.4% of those receiving diphenhydramine and 7.4% of those receiving hydroxyzine were diagnosed with an allergic reaction. It is also likely that, in the majority of cases, the patients receiving promethazine could have been administered alternative antiemetics with fewer potential side effects.

Based on previous studies, it appears that factors underestimating inappropriate medication usage outweighed those overestimating it. Because of limitations in the survey instrument, drugs contraindicated at certain dosages or in the presence of certain comorbidities were not included. Potential drug-drug interactions or the appropriateness of medications approved after the 1997 publication of the Beers criteria were did not considered. In studies that have included drugs fulfilling these additional criteria, inappropriate prescription rates of up to 47% over 1 year have been reported.23,44,45 Another study reported that 23% of ED patients are at risk of drug-drug interactions and 11% are at risk of drug-disease interactions due to medications administered or prescribed in the ED.42 Studies of outpatients using implicit criteria have found that up to 88% of elderly patients have an incorrect dose, therapeutic duplication, or potential drug-drug interactions.49

Despite the fact that the majority of the Beers listed medications appear to be inappropriately prescribed, there are many factors that contribute to the selection of drugs for individual patients. It has been suggested that explicit criteria are “best used as a screening tool, rather than a definitive measure of quality of care or performance.”30 At a minimum, physicians should carefully assess the risk/benefit ratio in any elderly patient to whom they are prescribing any of these medications.

In the multivariate analysis, the greatest independent predictor of inappropriate medication administration was the number of medications administered, a result consistent with previous studies,4,6,12,18,26–30 but previous studies have shown mixed results on what effects, if any, other patient and hospital characteristics have on inappropriate medication administration.4,18,23,27–30 In the current study, many variables demonstrated an association with inappropriate prescribing, especially because the large sample size of the data set allowed detection of effects of small magnitude. The results of the multivariate analysis indicate that extra caution must be taken when administering multiple medications to the elderly. Other patient and visit characteristics may also predict increased risk. For example, patients who were aged 65 to 84, nonurgent, or discharged were actually more likely to receive an inappropriate medication. Insurance status was not a predictive criterion. Additionally, prescribing in hospitals of certain types or in certain areas of the country may warrant heightened awareness, although great interfacility variation in drug administration patterns is suspected. Rural areas did not have greater rates of inappropriate medication administration, whereas hospitals outside the northeast did.

Several potential limitations of the study deserve comment. First, no information was available concerning the occurrence of adverse drug events in patients receiving inappropriate medications. It is unclear whether adverse events due to medications administered in the ED are as frequent as in those taken chronically at home, but drugs given in the ED are often given as a prescription. One study demonstrated lower health-related quality-of-life scores on follow-up in patients who received an inappropriate medication in the ED and in patients who were discharged on one.18 Further study is needed to determine the magnitude of adverse drug events caused by inappropriate medication administration in the ED. Additionally, the identification of chronic inappropriate medication use while the patient is in the ED provides an opportunity for emergency physician intervention to decrease chronic use of these medications. For example, it has been noted that 10% of elderly ED patients were taking an inappropriate medication at home.18

Disadvantages of using secondary data include potential errors in data collection, analysis, and reporting, but the NHAMCS implemented a quality control process in the field and during data processing. The study's 95% CIs were calculated based on publicly available survey information, which does not include all survey characteristics, but the calculation is based on the variance estimation curves provided by NCHS, which take into account all survey characteristics. Additionally, NHAMCS data are statistically limited at small sample sizes or estimates, as noted in the Methods section, but the conclusions were based only on those results that met the NHAMCS standards for data reliability. During data collection, determinations of urgency were based on assessments made at triage. It is known that triage is not entirely accurate in determining patient urgency, but this data-point is representative of real-world ED functioning.

Some who prefer the use of implicit criteria applied on a patient-by-patient basis have challenged the use of explicit criteria,49 but the Beers criteria have been widely applied in the literature and in practice.25,39 The use of explicit criteria avoids possible biases that may occur with implicit reviews and allows the use of large national surveys such as NHAMCS. Previous studies using explicit criteria have generally retained or only slightly modified the Beers criteria. Medications that explicit criteria identify as inappropriate may be appropriate in certain circumstances because of diagnosis, previous tolerance of the drug, or failure of safer therapies.20,30 To address this concern, the most common diagnoses of certain medications were noted, and it was found that the majority were given in patients with diagnoses not clearly requiring that specific agent.

In December 2003, a group of investigators published a 2002 update to the 1997 Beers criteria.50 This update was based on new information and the release of new drugs. It resulted in the addition of multiple medications to the list of those considered generally inappropriate, including, ketorolac, amiodarone, clonidine, and long-term use of noncyclooxygenase-selective nonsteroidals. The use of such medications in elderly ED patients should also be rare, and future studies of inappropriate medication administration should include these medications. It was decided not to use the updated 2002 criteria for the current study. This allowed determination of the effect that publication of the 1997 Beers criteria had on ED drug administration patterns. As a result, it was possible to conclude that criteria publication alone is insufficient to change national ED drug administration patterns. Further interventions are needed and might include involvement of national physician or healthcare quality organizations to increase guideline knowledge and compliance.

In summary, an estimated 16 million or 12.6% of elderly ED patients received inappropriate medications in the ED between 1992 and 2000. The administration of medications in clinically acceptable circumstances cannot explain the high rate, which has persisted despite publication of the 1997 Beers criteria. Such administration has previously been associated with increased morbidity and mortality in the elderly. Despite multiple studies of elderly outpatients, data have previously been lacking on inappropriate medication use in the acutely ill elderly, whether in EDs or as inpatients. Further study should be directed at means of decreasing inappropriate medication administration and at determining the effect, in monetary and health-related terms, of the administration of such medications in these settings.

Wider dissemination of the Beers criteria in the emergency medicine community and hospital-wide institution of policies and programs to discourage inappropriate prescribing are encouraged. Additionally, the development of educational and point-of-care systems to decrease inappropriate medication administration in the ED should be encouraged. Such programs would ideally be partnerships between the clinical departments and the pharmacies consisting of initial educational efforts and formulary changes combined with ongoing computer review, compliance monitoring, and feedback. Such efforts may be accomplished at the level of the individual hospital, but support for such goals by national organizations concerned with geriatric medicine and emergency medicine issues would be expected to increase awareness of the need for such reforms and provide impetus for their implementation.