Monane M, Matthias DM, Nagle BA, Kelly MA. Improving Prescribing Patterns for the Elderly Through an Online Drug Utilization Review InterventionA System Linking the Physician, Pharmacist, and Computer. JAMA. 1998;280(14):1249–1252. doi:10.1001/jama.280.14.1249
From the Departments of Medical Affairs (Drs Monane, Nagle, and Kelly) and Health and Utilization Management (Ms Matthias), Merck-Medco Managed Care, LLC, Montvale, NJ.
Context.— Pharmacotherapy is among the most powerful interventions to improve
health outcomes in the elderly. However, since some medications are less appropriate
for older patients, systems approaches to improving pharmacy care may be an
effective way to reduce inappropriate medication use.
Objective.— To determine whether a computerized drug utilization review (DUR) database
linked to a telepharmacy intervention can improve suboptimal medication use
in the elderly.
Design.— Population-based cohort design, April 1, 1996, through March 31, 1997.
Setting.— Ambulatory care.
Patients.— A total of 23269 patients aged 65 years and older throughout the United
States receiving prescription drug benefits from a large pharmaceutical benefits
manager during a 12-month period.
Intervention.— Evaluation of provider prescribing through a computerized online DUR
database using explicit criteria to identify potentially inappropriate drug
use in the elderly. Computer alerts triggered telephone calls to physicians
by pharmacists with training in geriatrics, whereby principles of geriatric
pharmacology were discussed along with therapeutic substitution options.
Main Outcome Measures.— Contact rate with physicians and change rate to suggested drug regimen.
Results.— A total of 43007 alerts were triggered. From a total of 43007 telepharmacy
calls generated by the alerts, we were able to reach 19368 physicians regarding
24266 alerts (56%). Rate of change to a more appropriate therapeutic agent
was 24% (5860), but ranged from 40% for long half-life benzodiazepines to
2% to 7% for drugs that theoretically were contraindicated by patients' self-reported
history. Except for rate of change of β-blockers in patients with chronic
obstructive pulmonary disease, all rates of change were significantly greater
than the expected baseline 2% rate of change.
Conclusions.— Using a system integrating computers, pharmacists, and physicians, our
large-scale intervention improved prescribing patterns and quality of care
and thus provides a population-based approach to advance geriatric clinical
pharmacology. Future research should focus on the demonstration
of improved health outcomes resulting from improved prescribing choices for
INDIVIDUALS aged 65 years and older constitute 12% of the US population;
however, they consume approximately 30% of prescribed medications.1 Older patients are prone to adverse drug events (ADEs)
because use of multiple medications regardless of age increases the chance
of ADEs2,3 and age-related physiologic
changes alter the pharmacokinetic and pharmacodynamic properties of many drugs.4 The incidence of ADEs in the elderly varies from 5%
to 35%, depending on the method used to define and measure the event.5 Adverse drug events may result in the need for additional
medications, disability, decreased quality of life and functioning, hospitalization,
Some medications are particularly prone to precipitating ADEs. Beers
et al6 used a Delphi survey with a panel of
experts in geriatrics to develop explicit criteria to identify medications
that should be avoided in older patients. The recommendation focused on drugs
that should be avoided, excessive dosing, and excessive duration of treatment.
A subset of these criteria was applied to the community-dwelling elderly in
the National Medical Expenditure Survey in 1987, which showed that nearly
25% of the elderly take at least 1 medication that should be avoided.7
Inappropriate prescribing in the elderly is often attributed to the
lack of training in geriatrics in medical and pharmacy education.8 An effective way to overcome this problem may be through
a concurrent drug utilization review (DUR) program. This type of utilization
management system is designed to send a warning to pharmacists when potentially
inappropriate drugs are prescribed.9,10
The warning provides an opportunity to educate the pharmacist and physician
through a discussion about the safety and effectiveness of a targeted medication
before it is dispensed.
The purpose of this study was to evaluate a program designed to decrease
the use of potentially inappropriate medications among the elderly through
a computer-based DUR intervention. The intervention included the rationale
for the alert, therapeutic alternatives, and withdrawal protocols if necessary,
and presented an opportunity to change the potentially inappropriate medication
before it was dispensed. Within this quality-of-care intervention, we measured
the change rate to a more appropriate medication in this population of elderly
All subjects were at least 65 years of age and receiving prescription
benefits from Merck-Medco Managed Care, LLC (MMMC), a prescription benefits
manager that provides medications through retail and mail pharmacy services
for approximately 51 million Americans. During the study period from April
1, 1996, through March 31, 1997, 2.3 million patients aged 65 years and older
filled at least 1 prescription through an MMMC mail-service pharmacy. In general,
patients use mail service more frequently to obtain maintenance medications
for chronic diseases. We report on all patients targeted through our computerized
DUR system with an actionable alert (alert triggering a conversation between
physician and pharmacist) completed in the 1-year surveillance period (N=23269).
Study Intervention. An independent medical advisory board, established by MMMC, consisting
of geriatric specialists in pharmacy, medicine, and nursing adopted the criteria
of Beers et al6 to identify the most dangerous
drugs for the elderly from a safety perspective. The MMMC Department of Medical
Affairs developed an integrated DUR education and intervention program centered
around a computerized online database aimed at decreasing the use of these
potentially unsafe drugs.
These senior-specific criteria covered 3 DUR categories: drug-age, maximum
daily dose, and drug-disease (Table 1).
The drug-age category defined drugs with pharmacokinetic,
pharmacodynamic, or ADE profiles known to be harmful in the elderly and for
which safer alternatives exist. Maximum daily dose alerts were limited to the short-acting benzodiazepines, for which specific
dosing recommendations exist for individuals older than 65 years. Drug-disease criteria defined drugs that should not be used in an older
patient in the presence of a specific condition that could be aggravated by
the drug. The disease history information was self-reported by the patient
during enrollment in the Partners for Healthy Aging Program,® a health
management program for the elderly designed and implemented by MMMC. These
senior-specific DUR criteria were then computerized and coded by the National
Drug Code to identify prescriptions requiring intervention.
Pharmacist training for the DUR program was conducted by a team of geriatric
pharmacy experts at all of the 13 MMMC mail-service pharmacies. These pharmacists
were instructed in both the pharmacy science around the DUR alerts, as well
as communication theory to conduct telephone one-to-one educational outreach
with physicians.11 If a potentially unsafe
medication was requested, the computer sent the pharmacist a warning. The
pharmacist subsequently attempted to call the physician to discuss the alert,
possible therapeutic alternatives, and applicable withdrawal recommendations.
The intervention outcomes included the following: (1) a discontinuation or
change in therapy, (2) no change in therapy, or (3) consideration of a change
in therapy at the next patient visit. Both the physician and the patient received
an explanatory confirmation letter in the mail if the original prescription
was changed. The prescription or changes were then filled and dispensed to
the patient through the MMMC mail-service channel. All relevant data pertaining
to the intervention were recorded electronically in MMMC DUR files.
Statistical Analysis. Frequency distributions as well as univariate and bivariate statistics
were computed to measure use of the targeted medications and the number of
physician contacts. The DUR change rate was determined as a function of the
number of interventions completed during the 1-year period of surveillance.
Specifically, the DUR change rate was equal to the percentage of events in
which calls to physicians were completed and recommended action was taken
(DUR change rate=[number of accepted recommendations/number of recommendations]×100).
The overall change rate was calculated for the 3 drug classes and separately
for drugs within the 3 groups. We also used z tests
to determine the significance of these DUR change rates from 2%, a level reported
in another comprehensive summary article to be the baseline rate of change
in prescribing over time.12,13
A total of 43007 alerts among 23269 elderly patients were triggered
during the study period across the 3 DUR categories. The median age of the
study population was 72 years (25%-75% interquartile range, 67-77 years);
24% were 80 years or older. Women constituted 62% of the population; 35% of
patients lived in the South, 31% in the Midwest, and 24% in the Northeast.
Patients received a median number of 8 unique prescriptions during the study
period. Approximately 25% of the patients completed medical history information,
reporting an average of 6 comorbidities. The most common reported comorbid
conditions included the following: (1) hypertension, (2) osteoarthritis, (3)
hypercholesterolemia, (4) peptic ulcer disease, and (5) angina.
The contact rate for reaching the targeted physicians of these patients
was 56% (24266/43007 alerts; average intervention time, 15 minutes), decreasing
the number of actionable alerts to 24266. A total of 19368 physicians were
contacted for the actionable alerts (average, 1.25 alerts per physician).
Most physicians were male (93%), between the ages of 40 and 60 years (66%);
40% were located in the South. Approximately 1 actionable alert was generated
per patient (24266 alerts per 23269 patients). The overall DUR change rate,
defined as the percentage of accepted recommendations divided by the total
number of recommendations, was 24% (N=5860/24266). Fifteen percent (3599/24266)
of alerts resulted in immediate change to a therapeutic alternative, and 9%
(2261/24266) resulted in a physician's indication to review the therapeutic
alternative at the patient's next visit. For the purpose of this analysis,
we present the DUR change rate as the sum of these 2 results above, since
more than 90% of these patients did not receive the targeted medication within
the next 6 months.
The data were first analyzed by type of DUR alert (Table 1). The intervention had the greatest impact on alerts in
the drug-age category (N=19362), resulting in a 24% rate of change. Although
the number of alerts was lower for the maximum daily dose category (N=4532),
this intervention resulted in a similar 25% rate of change. Interventions
involving the drug-disease category (N=372) resulted in a 5% rate of success.
The overall DUR change rate compared with an expected change rate of 2% was
significant at the P<.001 level. All change rates
were significant except for the β-adrenergic receptor antagonist-chronic
obstructive pulmonary disease drug-disease category.
There was marked variability in the change rate for specific DUR rules
within the 3 categories as described above. The drug-age alert resulted in
change rates of 17% to 40%: within this category, the long elimination half-life
benzodiazepine hypnotics had the highest rate of change at 40%. The maximum
daily dose category included only 1 class of drugs, the intermediate- or short-acting
benzodiazepines, and resulted in a 25% rate of success. There were 2 classes
of drugs included in the drug-disease category, with success rates ranging
from 2% to 7%. The reasons physicians gave for not changing were as follows:
agreed with intervention but it was not applicable to the patient (55%; 10123/18406),
disagreed with intervention (41%; 7546/18406), agreed with intervention but
it was inconvenient for the patient (2%; 368/18406), agreed with the intervention
but the patient did not (1%; 184/18406), or the physician terminated the call
without giving an explanation (1%; 184/18406).
This large-scale study examined a computerized online DUR database designed
to help reduce inappropriate prescribing and improve quality of care in an
elderly population. More than 43000 prescriptions in more than 23000 patients
were evaluated in this study. The telepharmacy intervention yielded a contact
rate with physicians greater than 50%, and the content of each intervention
call focused on quality-of-care messages based on best practices as determined
by the scientific literature and clinical guidelines. Messages linking the
computer, pharmacist, and physician led to improved quality of care with nearly
a quarter of all DUR alerts accepted by physicians, but varied from 40% for
long-acting benzodiazepines to 2% for use of β-adrenergic receptor antagonist
agents in patients with chronic obstructive pulmonary disease. Efforts to
provide alerts to physicians at the point of care are thought to provide ever
greater changes in prescribing and represent the next frontier for DUR intervention.
The use of prescription claims data offers major advantages in drug
surveillance, including the ability to document all health service use without
recall bias or incomplete drug history. Yet the limitations of claims-based
information must be recognized.14,15
This study used information on medications that were actually prescribed and
dispensed and does not include nonprescription drug use. No information is
recorded on use of medication prescribed and not dispensed in this analysis.
In addition, we assumed that patients took the medications as prescribed by
the physicians, which often is not the case.16,17
This evaluation may underestimate the extent of the problem and overestimate
the potential benefit of the intervention because it was based on mail-service
prescriptions and did not include retail pharmacies. Additionally, because
the response rate to the health questionnaire among patients was only 25%
and we had only limited clinical data beyond patients' self-report diagnoses
and prescription medication profile, this study may have underestimated the
number of medication alerts. Nonetheless, tracking of prescribing patterns
represents an intermediate outcome that can be easily monitored in a large
database and facilitates the identification and measurement of an important
indicator of quality of care.
As reported earlier, when estimating the impact on the quality of prescribing,
one should not compare the overall effect size of 24% with a theoretical 100%
change rate, but rather against the 2% baseline rate of change that occurs
in physicians' prescribing over time.13 Furthermore,
the drug alerts chosen for this computer-based intervention represent probable
quality-of-care indicators and may be more effective studying a population
than applying them in a case-by-case setting. Also, the drug criteria used
here are likely to represent the most clinically relevant issues in geriatric
pharmacy, thus small changes may have a major impact on clinical and economic
outcomes.18,19 The validity of
these drug criteria is based on the best available medical and pharmacy literature,
as established by consensus panels of experts or other methods to develop
The physician contact rate of 56% was less than optimal for improving
the overall care of older patients. While some physician feedback suggested
an unavailability or unwillingness to participate in our DUR intervention,
other physicians stated their appreciation for this patient-specific information.
Furthermore, if we considered all alerts generated and called on (N=43007)
vs the actionable alerts when a physician was successfully contacted (N=24266),
the overall change rate drops from 24% to 14%. We believe the actionable rate
is the more appropriate figure to use as it estimates the population exposed
to the intervention. The ability to merge computer online databases, pharmacist
intervention, and physician involvement will continue to be a major hurdle
given resource constraints as discussed above. Yet within the current environment
of cost and resource containment, our change rate may represent the best that
can be achieved, as well as a useful barometer for physician feedback concerning
Despite these limitations, the results of this study suggest that such
an intervention based on computer-generated DUR linked to systems that enhance
communication between pharmacists and physicians can be successful and can
improve prescribing of drug therapy for large patient populations. Our change
rates of 15% (immediate changes) and 9% (consider change later) are further
evidence of the ability of our program to both track prescriptions and intervene
on inappropriate medications before the medication reaches the patient. Furthermore,
in a follow-up analysis of physicians who stated they would "change medication
later," more than 90% of the targeted patients did not receive the inappropriate
medication in the following 6 months. While we did not recontact the physicians,
these data support the physicians' initial decisions in response to the DUR
What is the future of such an intervention program? The need for intervention
models targeted toward drugs with less than optimal risk-benefit profiles
in the elderly patient is evident, especially to reduce drug-drug interactions.
Second, the use of these drug-based quality indicators represents an intermediate
outcome as mentioned above, with the final end points that should be measured
being overall health status, functional status, and quality of life.21,22 Third, an assessment of the cost-effectiveness
of the intervention is needed. Fourth, the technology of communication must
be further examined, such as fax and Internet approaches. Finally, the ideal
scenario involves an intervention at the point of care, when the physician
and patient can discuss therapeutic options in the most logical place, the