Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

ORIGINAL INVESTIGATION

Discrepancies in the Use of Medications


Their Extent and Predictors in an Outpatient Practice
Susanna E. Bedell, MD; Samer Jabbour, MD, MPH; Robert Goldberg, PhD; Helene Glaser, RN;
Susan Gobble, MBA; Yinong Young-Xu, MA; Thomas B. Graboys, MD; Shmuel Ravid, MD

Background: Misuse of medications is a major cause result of patients taking medications that were not re-
of morbidity and mortality. Few studies have examined corded (n=278 [51%]); patients not taking a recorded
the frequency of, and factors associated with, discrepan- medication (n=158 [29%]); and differences in dosage
cies between what doctors prescribe and what patients (n=109 [20%]). Overall, discrepancies were randomly
take in actual practice. distributed among different drugs and discrepancy
types with no discernible pattern. On multivariate
Patients and Methods: Patients’ medication bottles analysis, patient age and number of recorded medica-
and their reported use of medications were compared with tions were the 2 most significant predictors of medica-
physicians’ records of outpatients seen between Novem- tion discrepancy.
ber 1997 and February 1998 in a private practice affili-
ated with an academic medical center in Boston, Mass. Conclusions: Discrepancies among recorded and re-
Three hundred twelve patients from the practices of 5 ported medications were common and involved all classes
cardiologists and 2 internists who were returning for their of medications, including cardiac and prescription drugs.
routine follow-up visits were included. Older age and polypharmacy were the most significant
correlates of discrepancy. The pervasiveness of discrep-
Main Outcome Measure: The presence of discrep- ancies can have significant health care implications, and
ancies based on comparing medication bottles with medi- action is urgently needed to address their causes. Such
cal records. action would likely have a positive impact on patient care.

Results: Discrepancies were present in 239 patients


(76%). The 545 discrepancies in these patients were the Arch Intern Med. 2000;160:2129-2134

I
N THIS ERA of polypharmacy, ex- more is known about adherence to medi-
tensive literature has docu- cations and less about discrepancy.16 The
mented the growing problem of present study was carried out in an out-
adverse drug reactions, misuse of patient practice setting to assess the mag-
medications, and significant cost nitude of the discrepancies between what
implications of drug-related morbidity and drugs are documented in the medical re-
mortality.1-6 While these problems affect cord and the medications that patients ac-
all segments of society, they are espe- tually take, to identify the types of dis-
cially prevalent among the elderly, a group crepancies, and to examine factors
that is especially vunerable because it com- associated with such discrepancies.
From Lown Cardiovascular
Center, Brookline, Mass
prises individuals who often have mul-
(Drs Bedell, Jabbour, Graboys, tiple medical conditions and therefore need RESULTS
and Ravid, Ms Glaser, and multiple medications.5,7-9 Errors and non-
Mr Young-Xu); The compliance in the use of medications in- The study sample included 312 white pa-
Department of Medicine, volve all types of drugs, including those tients; 48% were men and the mean age
Harvard Medical School that may be lifesaving, such as cardiac was 62 years. Table 1 describes the clini-
(Drs Bedell, Jabbour, Graboys, medications, and the resultant morbidity cal characteristics of the patient sample.
and Ravid), and the Division and mortality can be significant.10-15 Un- We stratified the sample according to the
of Cardiovascular Medicine, derstanding the magnitude and cause of specialty of the responsible physician in
Department of Medicine,
medication misuse is essential to devis- our office because the patient popula-
University of Massachusetts
Medical School (Dr Goldberg), ing adequate strategies to control this prob- tions seen by internists and cardiologists
Boston; and the Department lem. Understanding medication misuse is might be different.
of Medicine, Memorial Health especially important in the outpatient set- Patients seen by internists were sig-
Services, Long Beach, Calif ting, where there is opportunity to ad- nificantly younger. There was a nonsig-
(Ms Gobble). dress associated risk factors. Currently, nificant trend toward an increased num-

(REPRINTED) ARCH INTERN MED/ VOL 160, JULY 24, 2000 WWW.ARCHINTERNMED.COM
2129
Downloaded from www.archinternmed.com at American University of Beirut, on October 17, 2007
©2000 American Medical Association. All rights reserved.
PATIENTS AND METHODS not the cardiologists, at times verified the list of medica-
tions with the patient at the time of the office visit.
Information was abstracted from the medical record
STUDY SETTING about the patient’s sex, age, number of medications cur-
rently prescribed, person(s) responsible for the adminis-
The practice setting was the physician offices of 5 board- tration and supervision of the medications, whether other
certified cardiologists and 2 board-certified internists, all physicians participated in the patient’s care, number of years
of whom were affiliated with the same academic medical the patient had been with the physician office, and date of
center. All but 1 physician had practiced for more than 15 the patient’s last office visit.
years. In general, physicians saw their patients on an an- Between November 1, 1997, and February 28, 1998,
nual or as-necessary basis. On average, they spent 1 hour all patients scheduled for a visit with one of the physicians
with a new patient and 30 minutes with established pa- in the practice were called by a research assistant on the
tients. Medication changes by the primary care or cover- day before their appointment. Each day, the patients of a
ing physician were documented in the medical record. A different physician, assigned randomly, were interviewed
cardiovascular fellow responded to patient calls after of- so that there would be an equal opportunity to sample pa-
fice hours and was instructed to document in the charts tients from all physicians in the practice. Patients were asked
any recommended changes in the use of medications. to bring all their drugs (prescription and over-the-
counter) and medicated creams to the office visit. Ran-
DATA COLLECTION dom samples of patients were selected from the practices
of all physicians.
The medical record of each patient contained a list of the The research assistant (H.G. or S.G.) specifically
patient’s current medications, which was shared by all health asked patients to confirm that the medication bottles they
professionals involved in the patient’s care, both in the of- brought with them accurately reflected the name, dosage,
fice and in the hospital. This list was reviewed and up- and timing of the drugs taken at home. She noted the
dated at each office visit and became part of the medical labels on the medication bottles but recorded the patients’
note dictated on the day of the patient visit. It was also up- comments about what medications they actually took to
dated whenever prescriptions were renewed or added out- confirm the instructions on these labels. She compared
side the office visit. This has always been the established this information with the list of medications recorded in
process in our practice. The expectation was that the drug the chart. She also determined whether the patient took
list would contain information about the use of over-the- any additional medications that were not on the medica-
counter medications. An assistant to the 2 internists, but tion list and whether the patient was responsible for taking

ber of recorded medications in internal medicine patients had a longer association with our practice and a greater
when compared with cardiology patients (6.2 vs 5.5, number of medications listed on their medical records.
P=.07). This trend was most likely the result of greater Most discrepancies, 278 (51%), were attributable to
use of nonprescription drugs. The majority of patients patients taking medications that were not recorded. The
had established long-term relationships with their phy- rest of the discrepancies were attributable to patients not
sicians. Most patients were responsible for administer- taking a recorded medication (29%) or to differences in
ing their own medications, and the majority had other dosage (20%). The distribution of discrepancies accord-
physicians who also participated in their care. ing to medication type is shown in Table 3. While over-
the-counter medications were the single largest category,
MEDICATION DISCREPANCIES 61% of discrepancies involved prescription medications.
Discrepancies in cardiac medications according to
In 239 (76%) of the 312 patients, a total of 545 medi- subcategory and discrepancy type were noted (Table 4).
cation discrepancies were identified. Table 2 summa- Inconsistencies for nitrates were the most frequent, fol-
rizes patient characteristics according to the presence lowed by diuretics, angiotensin-converting enzyme in-
of discrepancy. hibitors, and b-blockers. Patients were as likely to have
Medication discrepancy occurred equally among men a discrepancy owing to a difference in dosage, taking an
and women. Overall, patients with discrepancy were sig- unrecorded medication, or not taking a recorded medi-
nificantly older. The percentage of discrepancies in dif- cation. Overall, discrepancies were randomly distrib-
ferent age groups was as follows: younger than 40 years, uted among different drugs and discrepancy types, with
47%; 40 through 49 years, 85%; 50 through 59 years, 73%; no discernible pattern noted.
and 60 years and older, 82%. Patients who had other phy-
sicians participate in their care were more likely to have PREDICTORS OF DISCREPANCY
discrepancies. This is not surprising, as these patients were
significantly older (mean age, 64 vs 49 years) and took We determined the association between any medication
more medications (mean number of medications, 6.0 vs discrepancy and demographic and clinical variables pre-
3.9) (P,.001 for both). Similarly, patients cared for by viously examined (Table 5). In univariate analyses, the
cardiologists were more likely to have medication dis- following covariates were associated with the presence
crepancies, as they too were significantly older (mean age, of any discrepancy: patient age, physician specialty, par-
67 vs 53 years; P,.001). Patients with discrepancies ticipation of another physician in patient care, years with

(REPRINTED) ARCH INTERN MED/ VOL 160, JULY 24, 2000 WWW.ARCHINTERNMED.COM
2130
Downloaded from www.archinternmed.com at American University of Beirut, on October 17, 2007
©2000 American Medical Association. All rights reserved.
his/her own medications. At the end of the structured patient was taking a medication not listed on the medical
interview, open-ended questions were asked to evaluate record or because he/she was not taking a documented
factors determining medication usage and to elicit pa- medication.
tients’ concerns and comments about their medications.
STATISTICAL ANALYSIS
DATA ANALYSIS
We used t tests for 2-group comparisons of continuous vari-
We defined medication discrepancy as the difference ables, and x2 analysis or the Fisher exact test of statistical
between the list of medications in the medical record (re- significance for comparisons of proportions. To identify mul-
ferred to as recorded medications) and what a patient actu- tivariate adjusted predictors of medication discrepancy, we
ally took based on medication bottles and on self-reports conducted 2 sets of analyses using different definitions of
to the trained research assistant (referred to as reported discrepancy. The first regression model, which we used to
medications). We categorized medications into 5 groups: examine factors associated with medication discrepancy,
(1) over-the-counter medications, including vitamins/ included discrepancies related to over-the-counter medi-
minerals, acetaminophen, decongestants, and gastrointes- cations and dosage differences, while the second analysis
tinal remedies such as antacids or histamine2 blockers; (2) excluded these discrepancies. The rationale behind these
anti-inflammatory medications, including aspirin; (3) psy- modeling approaches was to remove the effects of discrep-
choactive medications, including sleeping pills, antide- ancies of lesser clinical significance. We used logistic re-
pressants, and anxiolytics; (4) cardiac medications; and (5) gression to identify factors associated with discrepancy, while
other prescription drugs. Because of the importance of car- controlling for potentially confounding variables. Univari-
diac medications in our large cardiology practice, we fur- ate associations of independent covariates, such as age and
ther subdivided these medications into 8 groups: (1) sex, with medication discrepancy were initially deter-
b-blockers; (2) calcium channel blockers; (3) nitrates (ni- mined. Clinically relevant 2-way interactions were exam-
troglycerine and long-acting nitrates); (4) angiotensin- ined after the initial data categorization. Clinically rel-
converting enzyme inhibitors; (5) lipid-lowering drugs; (6) evant or statistically significant variables, including
diuretics; (7) warfarin sodium (Coumadin); and (8) oth- interaction terms, were entered into the final regression mod-
ers, including antihypertensive agents (other than those els. The most predictive and parsimonious models were se-
noted in categories 1, 2, 4, and 6). lected. Hosmer-Lemeshow goodness-of-fit testing was per-
For each medication group, we determined whether formed on selected regression models, and likelihood ratio
there was a discrepancy in the type or dosage of medica- testing was performed to compare different models. We used
tion. We noted whether the disparity occurred because the a significance level of .05. All P values were 2-sided.

the physician, and the number of recorded medications. 1. Desire for more information. Patients wanted
On multivariate analysis, patient age and the number of more details from their physicians about how the drug
recorded medications were the 2 most significant pre- they were prescribed would help their symptoms or how
dictors of discrepancy. As expected, we found a signifi- it would interact with other medications.
cant interaction between the number of medications on 2. Concerns about adverse effects. Adverse effects
the medical record and patient age, with older patients that were important to patients were often vague, such
taking more medications. We included this interaction as “feeling blah” or “feeling not myself.” Some patients
term in the regression models because we thought that worried that the dose of the medication they took was
it was of clinical significance and had an impact on the “too much.” Specific complaints most often focused on
assessment of the effect of other variables. There was evi- loss of libido or concern about liver toxicity.
dence for a nonsignificant trend toward increased dis- 3. Obstacles from convenience or cost. Conve-
crepancy when the patient was female, when the man- nience in taking medications and filling prescriptions was
aging physician was a cardiologist rather than an internist, more important to our patients than medication costs. Even
and when another physician participated in patient care. patients on a twice-daily medication regimen asked to sub-
Finally, we examined factors associated with medi- stitute it for a once-a-day medication. Patients wanted more
cation discrepancy, excluding over-the-counter medica- medications with each prescription to avoid the delay or
tions and dosage discrepancies, separately and jointly inconvenience of frequent refills. Some wondered whether
(Table 5). Age and number of recorded medications re- splitting a stronger pill would offer them savings.
mained significant predictors of discrepancy even though 4. Influence of multiple physicians. The majority of
the magnitude of effect for each of these variables changed. patients’ comments focused on the problems of having mul-
Participation of another physician was the most signifi- tiple physicians involved in their care. Many patients com-
cant predictor of this type of discrepancy. plained about lack of ready access to subspecialists and
that their primary care physician made medication changes
PATIENTS’ COMMENTS without consulting a specialist. One patient, for example,
said he had been “easing off” all his medications because
Comments from patients were informal, rather than quan- his primary care physician said that it would be fine to do
tified, but we did elicit meaningful responses about their so as long as he “felt all right.” Medication lists were of-
expressed concerns. We categorized patients’ feedback ten modified after the patients were discharged from the
about their medications into the following 4 areas: hospital, and patients were not always aware of the need

(REPRINTED) ARCH INTERN MED/ VOL 160, JULY 24, 2000 WWW.ARCHINTERNMED.COM
2131
Downloaded from www.archinternmed.com at American University of Beirut, on October 17, 2007
©2000 American Medical Association. All rights reserved.
and the basis of successful outcomes.20 While the extent
Table 1. Demographic Data by Physician Specialty* of medication discrepancies in our study was higher than
in previous reports, this difference likely reflected the me-
Internal ticulous effort given to correct identification of medica-
Cardiology Medicine Total
(n = 205) (n = 107) (N = 312) tions taken and the uniqueness of our study in using the
patients’ medication bottles rather than patients’ diaries,
Age, mean (± SD), y 67 (± 14) 53 (± 16) 62 (± 16)
Male, No. (%) 136 (66) 15 (14) 151 (48)
computer printouts, or pharmacy records to verify the pres-
Participation of another 184 (90) 87 (81) 271 (87) ence of discrepancy.21,22 Consistent with the report of Mon-
physician, No. (%) son and Bond,23 we found that the more drugs a patient
Patient responsible for 173 (84) 101 (94) 274 (88) takes, the more likely that there will be a discrepancy.21
medication, No. (%) The existing literature on medication use and mis-
Family responsible for 30 (15) 1 (1) 31 (10)
medication, No. (%)
use has primarily focused on one aspect of discrepancy,
Years with physician 7.0 2.2 5.3 namely patient compliance, which assesses the failure of
office, mean patients to adhere to prescribed medications. Our study
Time since last visit, 7.0 4.1 6.0 highlights the larger picture of discrepancy and extends the
mean, mo previous work of Wagner and Hogan22 demonstrating that
No. of recorded 5.5 6.2 5.8
what medications a patient takes does not depend on vo-
medications, mean
No. of reported 6.7 5.7 6.4 lition alone. Other factors, such as miscommunication
medications, mean among physicians or between physicians and patients, can
play an important role, as suggested by other reports.21,24
*Differences were significant at P,.05 for all variables except number of Our patients were sometimes following another physi-
recorded medications. cian’s orders, frequently outside our office practice, when
they failed to take prescribed medications or took addi-
tional nonrecorded drugs. The differences between the defi-
Table 2. Patient Characteristics According nitions of noncompliance and discrepancy notwithstand-
to Presence of Medication Discrepancy ing, existing data on high rates of noncompliance are
consistent with the present findings.25-29 A unique aspect
Discrepancy Discrepancy
Present Absent P
of our study was that we took into consideration the pa-
tients’ perspectives; similar to previous reports about medi-
Age, mean (± SD), y 64 (± 15) 56 (± 18) ,.001 cation compliance, we observed that concerns about the
Male, % 48 49 .51
Physician type, %
convenience of taking medications, filling prescriptions, and
Cardiologist 82 18 ,.001 adverse effects were most important for our patients.
Internist 65 35 ,.001 Another method to determine the medications pa-
Participation of another 80 56 ,.005 tients actually take is to evaluate computerized drug da-
physician, % tabases, such as pharmacy records.23 While this method
Years with physician office, mean 5.7 4.2 .05 has the advantage of improving efficiency in larger stud-
Time since last visit, mean, mo 5.7 6.5 .45
No. of recorded medications, mean 6.2 4.4 ,.001
ies and can serve as a surrogate for pill counting, it may
No. of reported medications, mean 7.0 4.4 ,.001 be impractical to use in an outpatient setting where drugs
are obtained from multiple sources. On the other hand,
self-reporting of drug intake may be subject to recollec-
tion bias but provides the clinician with ready access to
to inform their physician in our office of changes that were important information if meticulously performed.
made in their regimen outside the practice.
FACTORS ASSOCIATED WITH
COMMENT MEDICATION DISCREPANCY

This study demonstrates that there is considerable discrep- Older age and a higher number of recorded medications
ancy between recorded and reported medications in the ma- were strongly associated with medication discrepan-
jority of cases in our academic outpatient private practice. cies. Consistent with a previous report, our findings dem-
The discrepancies include all medications, prescription and onstrated that discrepancy was as likely to involve pre-
nonprescription, and were of different types, including dis- scription and potentially toxic medications as over-the-
crepancies in dosages, not taking recorded medications, and counter medications.21 Our finding that older age is
taking nonrecorded medications. One third of the discrep- associated with medication discrepancy seems plausible
ancies involved over-the-counter drugs or herbal thera- clinically, but, to our knowledge, prior studies have not
pies. Miscommunication about herbal therapies is rela- evaluated the influence of age on discrepancy.21,24,30 In
tively common because patients often self-prescribe without fact, within groups of patients older than 65 years, older
consulting or informing their physicians. Adverse effects age is associated with better compliance with antihyper-
from such therapies are not necessarily trivial.17-19 tensive therapy and with treatment of congestive heart
The majority of discrepancies occurred with pre- failure.26,31 Reports on compliance or likelihood of ad-
scription drugs and a full quarter with cardiac drugs. These verse drug reactions have not found an association with
findings are especially striking because medical therapy advancing age.7,32-34 We found nonsignificant trends to-
has been the foundation of medical care in our practice ward increased discrepancy in cases involving female pa-

(REPRINTED) ARCH INTERN MED/ VOL 160, JULY 24, 2000 WWW.ARCHINTERNMED.COM
2132
Downloaded from www.archinternmed.com at American University of Beirut, on October 17, 2007
©2000 American Medical Association. All rights reserved.
Table 3. Detailed Medication Discrepancies by Drug Type and by Subspecialty

Over-the-counter Prescription Sleep Cardiac Aspirin/Anti-inflammatory


All
Not Not in Not Not in Not Not in Not Not in Not Not in Types,
Dose Taking Record Total Dose Taking Record Total Dose Taking Record Total Dose Taking Record Total Dose Taking Record Total Total
Cardiology 2 20 112 134 19 17 50 86 10 5 20 35 46 33 39 118 10 7 13 30 403
Internal 2 15 22 39 9 31 13 53 3 15 2 20 4 6 6 16 4 9 1 14 142
medicine
Total 4 35 134 173 28 48 63 139 13 20 22 55 50 39 45 134 14 16 14 44 545

Table 4. Cardiac Medication Discrepancies by Type Table 5. Crude and Multivariate Predictors
of Medication Discrepancy*
No. (%)*
Discrepancy for
Not in Dose Recorded Any Discrepancy Prescription
Record Discrepancy Not Taking Total Medications
b-Blockers 3 (2) 10 (7) 4 (3) 17 (13) Crude OR Adjusted OR Only†, Adjusted
Calcium channel 1 (1) 3 (2) 4 (3) 8 (6) Factor (95% CI) (95% CI) OR (95% CI)
blockers Age, 10-y 1.36 (1.16-1.60) 1.60 (1.14-2.24) 1.47 (1.08-2.00)
Nitrates 10 (7) 6 (4) 9 (7) 25 (19) increment
ACE† inhibitors 2 (1) 11 (8) 6 (4) 19 (14) Sex 0.98 (0.58-1.65) 1.73 (0.83-3.58) 1.05 (0.58-1.93)
Lipid-lowering 5 (4) 6 (4) 5 (4) 16 (12) Physician being 0.40 (0.24-0.69) 0.44 (0.12-1.65) 0.61 (0.18-2.10)
Diuretics 5 (4) 10 (7) 8 (6) 23 (17) internist
Warfarin sodium 5 (4) 1 (1) 2 (1) 8 (6) Participation of 3.11 (1.51-6.40) 1.45 (0.63-3.38) 2.49 (1.14-5.44)
(Coumadin) another
Other 14 (10) 3 (2) 1 (1) 18 (13) physician
Total 45 (34) 50 (37) 39 (29) 134 (100) No. of recorded 1.21 (1.10-1.33) 2.28 (1.47-3.53) 1.74 (1.20-2.53)
medications
*Some percentages are not true totals because of rounding.
†ACE indicates angiotensin-converting enzyme. *OR indicates odds ratio; CI, confidence interval.
†Excluding over-the-counter medications and dose discrepancy.

tients, in cases involving patients managed by cardiolo- we identified reflected the realities of day-to-day use of the
gists as compared with internists, and in cases in which medical record in patient care. Also, if the charts were to
there was participation of another physician in patient be reviewed after the patient visit, physicians aware of the
care. A long-term patient-physician relationship did not study may have attempted to take a more accurate medi-
diminish the likelihood of medication discrepancy. cation history, thereby introducing bias. Finally, this study
We expected a lower number of discrepancies, con- did not assess the impact of medication discrepancy on
sidering that our patient population was well educated patients’ outcomes. However, studies on noncompliance
and of high socioeconomic status. To our knowledge, have clearly documented the association between medi-
there is no information in the published literature that cation misuse and adverse health outcomes, in both the
indicates how socioeconomic status may affect medica- outpatient and the inpatient settings.4,35-39 This medica-
tion discrepancy, and the relationship may be more com- tion misuse may have a major impact on outcomes for many
plicated than we expected. illnesses, including cardiovascular disease.11-13,15 For ex-
ample, discontinuing certain cardiac medications, such as
STUDY LIMITATIONS b-blockers, may be detrimental to patients with coronary
disease, triggering potentially life-threatening arrhyth-
There are several limitations to the present observational mias or myocardial infarction.13
study. First, while the study sample was representative of
our practice as a whole, our results might not be gener- RECOMMENDATIONS
alizable to other clinical settings, to patients enrolled in
health maintenance organizations, or to patients of low to There are several possible solutions to improve medica-
moderate socioeconomic status. Second, our data did not tion accuracy. Our findings suggest that a compulsive, spe-
allow us to separate discrepancies that were caused by im- cific, and systematic review of the patient’s medication
proper practice in our office from those resulting from a bottles should become a standard element in the patient’s
lack of communication from an outside physician or those care. Although this system may seem time consuming and
resulting from patients acting independently. Third, we cumbersome, it is unlikely to outweigh the cost of medi-
did not collect information on comorbid illnesses and cation misuse for patients with chronic or comorbid ill-
thereby did not assess their influence on discrepancy. nesses. It will ensure accuracy and identify any change in
Fourth, we may have found fewer discrepancies had we medications, whether initiated by an independent or non-
assessed medications in the medical record after a physi- compliant patient or by a physician who fails to commu-
cian visit rather than before. However, the discrepancies nicate his/her adjustment of medications. Critical review

(REPRINTED) ARCH INTERN MED/ VOL 160, JULY 24, 2000 WWW.ARCHINTERNMED.COM
2133
Downloaded from www.archinternmed.com at American University of Beirut, on October 17, 2007
©2000 American Medical Association. All rights reserved.
of the medication list should emphasize the simplest, most response of cholesterol on mortality in the Coronary Drug Project. N Engl J Med.
1980;303:1038-1041.
parsimonious prescribing regimen.40 Communication 11. Gallagher EJ, Viscoli CM, Horwitz RI. The relationship of treatment adherence to
among primary care physicians and subspecialists, such the risk of death after myocardial infarction in women. JAMA. 1993;270:742-744.
as cardiologists, clearly needs to improve to achieve greater 12. Horwitz RI, Viscoli CM, Berkman L, et al. Treatment adherence and risk of death
after a myocardial infarction. Lancet. 1990;336:542-545.
accuracy in medication use and instruction. 13. Psaty BM, Koepsell TD, Wagner EH, LoGerfo JP, Inui TS. The relative risk of in-
Other proposals that have been developed in the past cident coronary heart disease associated with recently stopping the use of b-block-
to enhance medication accuracy include a telecommuni- ers. JAMA. 1990;263:1653-1657.
14. Miller NH. Compliance with treatment regimens in chronic asymptomatic dis-
cations system for monitoring drugs,41,42 continuous elec- eases. Am J Med. 1997;102:43-49.
tronic monitoring of medication containers,43 follow-up by 15. McDermott MM, Schmitt B, Wallner E. Impact of medication nonadherence on
coronary heart disease outcomes. Arch Intern Med. 1997;137:1921-1929.
pharmacists,44 computerized prescribing,45 “one-write” non- 16. Gurwitz JH, Yeomans SM, Glynn RJ, Lewis BE, Levin RM, Avorn J. Patient non-
carbon prescription forms,46 or use of a standardized drug compliance in the managed care setting: the case of medical therapy for glau-
questionnaire.47 We hope to initiate a program of phar- coma. Med Care. 1998;36:357-369.
17. Pillans PI. Toxicity of herbal products. N Z Med J. 1995;108:469-471.
macy bar code labeling into our practice so that drugs can 18. McRae S. Elevated serum digoxin levels in a patient taking digoxin and Siberian
be mechanically recorded at the time of each visit and ginseng. CMAJ. 1996;155:293-295.
printed up for the physician to review. This system allows 19. Ginkgo biloba for dementia. Med Lett Drugs Ther. 1998;40:63-64.
20. Graboys TB, Blatt CM, Ravid S. Optimal medical therapy reduces referrals for
physicians to track changes in medications initiated by other invasive cardiovascular procedures. Am Coll Cardiol Curr J Rev. January/
physicians or during hospitalizations. Patients’ input should February 1997:81-84.
be carefully sought before adopting any solution to en- 21. Straka RJ, Fish JT, Benson SR, Suh JT. Patient self-reporting of compliance does
not correspond with electronic monitoring: an evaluation using isosorbide dini-
sure feasibility and relevance to patients’ preferences. trate as a model drug. Pharmacotherapy. 1997;17:126-132.
22. Wagner MM, Hogan WR. The accuracy of medication data in an outpatient elec-
tronic medical record. JAMA. 1996;3:234-244.
CONCLUSIONS 23. Monson RA, Bond CA. The accuracy of the medical record as an index of out-
patient drug therapy. JAMA. 1978;240:2182-2184.
Discrepancies among recorded and reported medica- 24. Cramer JA, Mattson RH, Prevey ML, Scheyer RD, Oullette VL. How often is medi-
cation taken as prescribed? JAMA. 1989;261:3273-3277.
tions were common in our study; they occurred in 75% 25. Monane M, Bohn RL, Gurwitz JH, Glynn RJ, Avorn J. Noncompliance with con-
of the patients. Discrepancies involved all classes of medi- gestive heart failure therapy in the elderly. Arch Intern Med. 1994;154:433-437.
cations, including cardiac and prescription drugs. Older 26. Rudd P, Tul V, Brown K, Davidson SM, Bostwick GJ. Hypertension continuation
adherence: natural history and role as an indicator condition. Arch Intern Med.
age and an increasing number of prescribed drugs were 1979;139:545-549.
the most significant correlates of discrepancy. The per- 27. Skaer TL, Sclar DA, Robison LM, et al. Effect of pharmaceutical formulation
vasiveness of medication discrepancy may have signifi- for antihypertensive therapy on health service utilization. Clin Ther. 1993;15:
715-725.
cant health care implications that deserve further study. 28. Dekker FW, Dieleman FE, Kaptein AA, Mulder JD. Compliance with pulmonary
Action is needed to address the variety of causes that may medication in general practice. Eur Respir J. 1993;6:886-890.
29. Price D, Cooke J, Singleton S, Feely M. Doctors’ unawareness of the drugs their
have an impact on discrepancy. Such action would likely patients are taking: a major cause of overprescribing? BMJ. 1986;292:99-100.
have a positive impact on patient care, patient-physi- 30. Monane M, Bohn RL, Gurwitz JH, Glynn RJ, Levin R, Avorn J. Compliance with
cian relationships, and long-term outcomes. antihypertensive therapy among elderly Medicaid enrollees: the roles of age, gen-
der, and race. Am J Public Health. 1996;86:1805-1808.
31. Horwitz RI, Horwitz SM. Adherence to treatment and health outcomes. Arch In-
Accepted for publication January 11, 2000. tern Med. 1993;153:1863-1868.
This study was supported in part by the Lown Cardio- 32. Col N, Fanale JE, Kronholm P. The role of medication noncompliance and ad-
verse drug reactions in hospitalizations of the elderly. Arch Intern Med. 1990;
vascular Research Foundation, Brookline, Mass, and by the 150:841-845.
Grimshaw-Gudewicz Charitable Foundation, Fall River, Mass. 33. Murphy J, Coster G. Issues in patient compliance. Drugs. 1997;54:797-800.
Reprints: Susanna E. Bedell, MD, Lown Cardiovascu- 34. Friedman GD, Collen MF, Harris LE, Van Brunt EE, Davis LS. Experience in moni-
toring drug reactions in outpatients: The Kaiser-Permanente Drug Monitoring
lar Center, 21 Longwood Ave, Brookline, MA 02446 (e-mail: System. JAMA. 1971;217:567-572.
Bambil@tiac.net). 35. Steel K, Gertman PM, Crescenzi C, Anderson J. Iatrogenic illness on a general
medical service at a university hospital. N Engl J Med. 1981;304:638-642.
36. Shapiro S, Slone D, Lewis GP, Jick H. Fatal drug reactions among medical in-
REFERENCES patients. JAMA. 1971;216:467-472.
37. Lazarou J, Pomeranz, BH, Corey PN. Incidence of adverse drug reactions in hos-
pitalized patients. JAMA. 1998;279:1200-1205.
1. Brennan TA, Leape LL, Laird N, et al. Incidence of adverse events and negli- 38. Bootman JL, Harrison DL, Cox E. The health care cost of drug-related morbidity
gence in hospitalized patients: results from the Harvard Medical Practice Study and mortality in nursing facilities. Arch Intern Med. 1997;157:2089-2096.
I. N Engl J Med. 1991;324:370-376. 39. Bedell SE, Deitz DC, Leeman D, Delbanco TL. Incidence and characteristics of
2. Bates DW, Cullen D, Laird N, et al. Incidence of adverse drug events and poten- preventable iatrogenic cardiac arrests. JAMA. 1991;265:2815-2820.
tial adverse drug events: implications for prevention. JAMA. 1995;274:29-34. 40. Sanson-Fisher RW, Clover K. Compliance in the treatment of hypertension: a need
3. Lesar TS, Briceland L, Stein D. Factors related to errors in medication prescrib- for action. Am J Hypertens. 1995;8:82S-88S.
ing. JAMA. 1997;277:312-317. 41. Friedman RH, Kazis LE, Jette A, et al. A telecommunications system for moni-
4. Classen DC, Pestotnik SL, Evans RS, Lloyd JF, Burke JP. Adverse drug events in toring and counseling patients with hypertension: impact on medication adher-
hospitalized patients: excess length of stay, extra costs, and attributable mor- ence and blood pressure control. Am J Hypertens. 1996;9:285-292.
tality. JAMA. 1997;277:301-306. 42. Wasson J, Gaudette C, Whaley F, et al. Telephone care as a substitute for rou-
5. Colley CA, Lucas LM. Polypharmacy: the cure becomes the disease. J Gen In- tine clinic follow-up. JAMA. 1992;267:1788-1793.
tern Med. 1993;8:278-283. 43. Kruse W, Koch-Gwinner P, Nikolaus T, Oster P, Schlierf G, Weber E. Measure-
6. Johnson JA, Bootman LJ. Drug-related morbidity and mortality: a cost-of- ment of drug compliance by continuous electronic monitoring: a pilot study in el-
illness model. Arch Intern Med. 1995;155:1949-1956. derly patients discharged from hospital. J Am Geriatr Soc. 1992;40:1151-1155.
7. Gurwitz JH, Avorn J. The ambiguous relation between aging and adverse drug 44. Muirhead G. Consenting adults. Drug Top. 1996;140:56.
reactions. Ann Intern Med. 1991;114:956-966. 45. Schiff GD, Rucker D. Computerized prescribing: building the electronic infra-
8. Monette J, Gurwitz JH, Avorn J. Epidemiology of adverse drug events in the nurs- structure for better medication usage. JAMA. 1998;279:1024-1030.
ing home setting. Drugs Aging. 1995;7:203-211. 46. Miller LG, Matson CC, Rogers JC. Improving prescription documentation in the
9. Classen DC, Pestotnik SL, Evens RS, Burke JP. Computerized surveillance of ad- ambulatory setting. Fam Pract Res J. 1992;12:421-429.
verse drug events in hospital patients. JAMA. 1991;266:2847-2851. 47. Colvin R. Prescription Drug Abuse: The Hidden Epidemic. Omaha, Neb: Addicus
10. Coronary Drug Project Research Group. Influence of adherence to treatment and Books Inc; 1998:21-25.

(REPRINTED) ARCH INTERN MED/ VOL 160, JULY 24, 2000 WWW.ARCHINTERNMED.COM
2134
Downloaded from www.archinternmed.com at American University of Beirut, on October 17, 2007
©2000 American Medical Association. All rights reserved.

You might also like