Fischer MA, Avorn J. Economic Implications of Evidence-Based Prescribing for HypertensionCan Better Care Cost Less?. JAMA. 2004;291(15):1850-1856. doi:10.1001/jama.291.15.1850
Author Affiliations: Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Mass.
Context Deviation from evidence-based guidelines in hypertension treatment is
common, but its economic impact has not been rigorously studied. Suboptimal
prescribing patterns contribute to the high cost of medications for elderly
patients as well as the difficulty in providing affordable prescription drug
benefits for older Americans.
Objective To calculate the potential savings from the perspective of health care
payers that would result from increased adherence to evidence-based recommendations
for managing hypertension in patients older than 65 years.
Design Comparative analysis of medications prescribed vs potential regimens
suggested by evidence-based guidelines tailored to each patient's medical
history, with calculation of the costs of both the actual and the evidence-based
Setting and Patients A total of 133 624 patients being treated for hypertension during
2001 who were enrolled in a large state pharmaceutical assistance program
that provides prescription drug insurance for elderly persons.
Main Outcome Measure Cost difference between medications actually prescribed and regimens
suggested by evidence-based guidelines.
Results The patients studied filled more than 2.05 million prescriptions for
antihypertensive medications in 2001, at an annual program cost of $48.5 million
($363 per patient). We identified 815 316 prescriptions (40%) for which
an alternative regimen appeared more appropriate according to evidence-based
recommendations. Such changes would have reduced the costs to payers in 2001
by $11.6 million (nearly a quarter of program spending on antihypertensive
medications), as well as being more clinically appropriate overall. Replacement
of calcium channel blockers resulted in the largest potential savings. Use
of pricing limits similar to those in the Medicaid program would have resulted
in even larger potential savings of $20.5 million (42% of program costs).
Conclusions Adherence to evidence-based prescribing guidelines for hypertension
could result in substantial savings in prescription costs for elderly patients
with hypertension that would amount to savings of about $1.2 billion nationally.
Identification of similar areas in which prescribing can be improved will
be critical for the affordability of prescription drug benefit programs.
Costs of medications for chronic conditions continue to escalate, particularly
for the elderly.1,2 The Congressional
Budget Office projected that in 2003, Medicare beneficiaries would spend an
average of $2440 per person on prescription drugs.3 Although
some older patients have supplemental insurance to cover the cost of prescription
drugs,1 the proportion without coverage has
increased in recent years.4 Vulnerable older
patients may not fill prescriptions because of their lack of insurance coverage.5,6 Legislation to create a prescription
drug benefit for Medicare recipients was recently passed,7 but
substantial co-payments will remain for many patients, and the cost of the
benefit will likely be much higher than originally anticipated.8 Identification
of areas in which clinical practice can be improved while lowering costs to
patients and prescription drug plans will be critical to ensure the affordability
of medications for older patients.
More than 60% of Americans aged 65 years and older have hypertension.9 Estimates of the cost of its treatment have ranged
from $7 billion to $15.5 billion per year.10- 13 Therapeutic
options have expanded tremendously in the past 2 decades, and guidelines have
changed as well.14 In the recently published
Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial
(ALLHAT),15 more than 33 000 patients
with hypertension were randomly assigned to receive amlodipine, lisinopril,
or chlorthalidone; the thiazide-type diuretic was shown to be most effective
in controlling systolic blood pressure as well as in preventing heart failure
and stroke; it was also the least costly. The seventh report of the Joint
National Committee on Prevention, Detection, Evaluation, and Treatment of
High Blood Pressure (JNC VII) recommended thiazide-type diuretics as first-line
agents for uncomplicated hypertension in most patients.14 Although
some trials offer evidence for angiotensin-converting enzyme (ACE) inhibitors,16 a comprehensive meta-analysis emphasized strong evidence
for thiazides as the preferred agents for the majority of patients with this
condition.17 Specific comorbidities (eg, diabetic
nephropathy or coronary artery disease) constitute indications for other agents
(ACE inhibitors and β-blockers, respectively), although even in these
conditions, guidelines state that thiazide-type diuretics could still be considered
We investigated the economic impact that would result from better adherence
to evidence-based therapeutic guidelines. To do this, we analyzed medication
use patterns in more than 130 000 hypertensive patients in a state drug
assistance program for elderly patients. We evaluated every antihypertensive
regimen in light of the clinical history of each patient and then estimated
the potential cost savings to the health care delivery system that could have
been realized through adherence to evidence-based recommendations.
Pennsylvania's Pharmaceutical Assistance Contract for the Elderly (PACE)
program provides prescription drug coverage for a large population of that
state's elderly patients who are not indigent enough to qualify for Medicaid
but have annual incomes of less than $14 000 for single persons or $17 200
for married persons.18 The program covered
all drug costs, minus a $6 co-payment for each prescription, for 201 000
beneficiaries in 2001. Prior publications have studied medication use in the
PACE program.19,20 Medication
claims were available from 1994 through 2001. Data on filled prescriptions
were linked to Medicare part A and B files containing additional data on all
diagnoses and comorbidities, both outpatient and inpatient, recorded for these
patients through the end of 2000. All personal identifiers were transformed
to anonymous untraceable coded study numbers to protect patient confidentiality.
Institutional review board approval was obtained for review of the claims
records. Data analyses were performed using SAS software.21 Prior
research has validated the accuracy of computerized claims data for determining
medication exposure.22- 26 The
PACE program receives rebates averaging about 16% from drug manufacturers;
this was considered in all cost analyses.
We identified every patient who filled a prescription for antihypertensive
medications during 2001 and who had been an active user of the PACE program
for at least 1 year prior to his/her first antihypertensive prescription.
To avoid considering antihypertensive drugs prescribed for other indications,
we required that study patients have a recorded diagnosis of hypertension.
We recorded data on all other PACE prescriptions filled by these patients
since 1994, as well as all diagnoses recorded in all outpatient and inpatient
encounters or reflected in medications used (eg, insulin for diabetes) during
the period 1994-2001. In this way, we identified the presence of diabetes
mellitus, congestive heart failure (CHF), history of myocardial infarction
(MI), asthma, chronic obstructive pulmonary disease (COPD), angina, nephropathy,
or benign prostatic hypertrophy.
In keeping with current guidelines, we considered thiazides to be appropriate
first-line hypertension therapy for patients without specific contraindications
or indications for another drug. The JNC VII report includes specific indications
for certain classes of antihypertensives,14 which
we incorporated. For patients with CHF or diabetes and nephropathy, we considered
ACE inhibitors to be indicated as first-line therapy.14,27- 30 For
patients with a history of ischemic heart disease, we considered β-blockers
as first-line therapy,14,31 except
in patients with diagnoses of asthma, COPD, or CHF. Angiotensin receptor blockers
were considered acceptable first-line agents in patients with diabetes and
nephropathy.32,33 We followed
JNC VII in allowing α-blockers as potentially favorable treatment for
hypertensive patients with prostatic disease,14 although
some evidence suggests that this may be a problematic strategy.34,35 For
patients with hypertension and angina, we classified calcium channel blockers
as an acceptable therapeutic choice based on JNC-VII guidelines,14 although
other recommendations suggest use of β-blockers for most such patients.36
For each patient, we considered all comorbid conditions in assigning
alternative antihypertensive medications when appropriate. If there were any
current or prior prescriptions for a given medication, we assumed that the
drug class was currently being used or had been tried previously and was abandoned
due to ineffectiveness or adverse effects, and we did not assign it as a substitute.
Medications were replaced on a prescription-for-prescription basis, assuming
that compliance with the substitute regimen would be the same as compliance
with the prescriptions actually filled, in keeping with recent data.15Figure 1 summarizes
the basic logic used in assigning this replacement therapy. For patients without
a significant comorbidity and with no history of thiazide use, we assigned
thiazide as replacement therapy; we did the same for additional prescriptions
for patients with comorbid conditions. For any patient with a relevant comorbidity
who had no history of using the indicated first-line agent, we assigned that
agent as a replacement. For patients receiving combination agents, we used
the same logic but continued use of combination drugs. For example, if a patient
with CHF was using a calcium channel blocker–ACE inhibitor combination,
we assumed replacement with a thiazide–ACE inhibitor combination.
We calculated the average per-day cost of each medication class to be
substituted, as measured by the average amount PACE paid to pharmacies. The
daily cost of thiazides or β-blockers reflected use of generic hydrochlorothiazide
or atenolol, respectively. The average cost of ACE inhibitors was that for
lisinopril, the least expensive and most often used ACE inhibitor during the
study year. For combination agents, we calculated the average per-day cost
of the most commonly used combination product in a given category. The cost
per day was averaged over all doses of a given medication. We then recalculated
the cost of each patient's alternative regimen, using evidence-based substitutions
as outlined herein and in Figure 1.
For both the actual and replacement costs, we subtracted 16% to account for
To incorporate other health care costs of hypertension treatment, we
also included the laboratory costs (calculated from laboratory claims data
in the study population) associated with these replacements. For thiazides,
we assumed that potassium would be checked twice yearly37;
for ACE inhibitors, we assumed that both potassium and creatinine would be
checked twice yearly.37 Patients taking thiazide
diuretics often need potassium replacement. We calculated the average cost
of generic long-acting potassium tablets or capsules based on more than 75 000
prescriptions covered by PACE in 2001. Using findings from the Systolic Hypertension
in the Elderly Program (SHEP) trial, we assumed that 30% of patients taking
thiazides would require potassium supplements38,39 and
added this cost to the cost of thiazide therapy.
We then calculated the amount paid by the PACE program in the study
year for each patient's actual regimen and compared it with the total amount
that the evidence-based regimen would have cost the health care system for
In Medicaid, the prices of generically available drugs are fixed at
a lower level by "maximum allowable cost" (MAC) pricing.18 In
secondary analyses, we evaluated the additional impact of using MAC-level
pricing for ACE inhibitors. In other analyses, we also varied the percentage
of patients taking thiazide diuretics who would require potassium supplements.
Of approximately 201 000 PACE participants, 133 624 (66%)
had a diagnosis of hypertension and filled 2 051 281 antihypertensive
drug prescriptions during the study year, an average of 15.4 prescriptions
per patient. Table 1 shows the
basic demographic characteristics of this population and the frequency of
major comorbid conditions. Table 1 also
shows the earliest year for which PACE data were available for these patients.
Nearly two thirds of patients (n = 85 551) had a history in PACE extending
to 1994; an additional 32 322 (24.2%) had 3 to 6 years of prior history
in PACE. The 2001 program expenditure for antihypertensives (after adjustment
for rebates) was $48.5 million, or $363 per patient. For prescriptions costing
less than the $6-per-prescription co-payment, the program paid nothing, but
data on medication type and quantity were recorded. Including co-payments,
total spending for antihypertensive medications by the program and patients
was about $60 million.
Table 2 summarizes by class
the number of prescriptions, amount paid, and average cost across all prescriptions.
Calcium channel blockers had the highest average cost ($33.39 per prescription)
and the highest total spending ($17 million); ACE inhibitors were the second
costliest class ($10.5 million). β-Blockers were the most commonly prescribed
antihypertensive drug, but average cost ($15.62) and total spending ($8.0
million) were lower than for ACE inhibitors or calcium channel blockers. Thiazides
were among the least expensive medications ($5.33 per prescription) but accounted
for only 4.3% of prescriptions.
Table 3 presents the numbers
of patients in several major clinical categories who were and were not prescribed
first-line antihypertensive medications. Among patients with CHF, ischemic
heart disease, and diabetes with nephropathy, we found no history of use of
a first-line medication for 33% to 38%. Of patients with no major comorbidities,
82% had no history of use of a thiazide-type diuretic.
For 815 316 prescriptions (40% of total), an evidence-based replacement
medication appeared clinically appropriate after review of all of a patient's
recorded diagnoses and drug use history. The algorithm identified a lower-cost
agent for 631 951 of these prescriptions (31% of total). In 183 365
prescriptions (9% of the total), the replacement (primarily ACE inhibitors)
would have cost more than the existing prescription. If all of the substitutions
were made, spending on antihypertensive medications and associated laboratory
tests would have decreased by $11.6 million in the study year (24% of the
program's antihypertensive drug spending), an annual cost reduction of $8.7
million per 100 000 patients treated for hypertension.
Table 4 summarizes the projected
changes in drug use by drug class. Substitution appeared appropriate for about
half of calcium channel blocker prescriptions (244 645 [48%]); the greatest
savings came from 140 600 prescriptions that could have been replaced
with thiazides, saving $4.32 million. Of existing ACE inhibitor prescriptions,
94 060 (23%) appeared replaceable by thiazides, with associated spending
reduction of $2.05 million (20% of spending on ACE inhibitors). In 71 885 β-blocker
prescriptions (14%), there was an indication for substitution by an ACE inhibitor,
which would have increased spending by $570 000. Most of the $701 622
spending reduction for combination agents was accounted for by calcium channel
blocker–ACE inhibitor combinations that could have been replaced with
combinations containing thiazides.
Over time, the price of generic ACE inhibitors should continue to fall,
so we also considered the potential impact of MAC-level pricing, as used by
Medicaid in most states,18 by recalculating
the potential costs if all replacement ACE inhibitors had been priced at the
average cost paid by the Pennsylvania Medicaid program for a generic ACE inhibitor.
This resulted in a larger reduction in spending of $15.8 million, or 33% of
program spending on hypertension. If the MAC price had been used not just
for substitution but in place of all brand-name ACE inhibitor prescriptions
(analogous to a reference-pricing approach40),
spending would have declined by another $4.7 million, increasing the total
savings to $20.5 million, or a potential reduction of 42% of spending for
If all thiazide users required potassium replacement, the potential
savings would have decreased slightly to $10.20 million; if only 20% required
potassium replacement, the potential savings would have increased to $11.85
We reviewed the medication choices and costs of antihypertensive regimens
in typical elderly patients in a large state drug assistance program over
a 1-year period and measured the cost implications of evidence-based drug
substitution when clinically appropriate in light of their medical histories.
A large proportion of existing prescribing for hypertension was outside of
published guidelines. Adherence to established guidelines for treating hypertension
would have resulted in considerable savings to the health care delivery system
($11.6 million per year, or nearly a quarter of program antihypertensive medication
costs). While some controversy persists regarding the precise details of which
agents are best for first-line therapy, the broad outline of the approach
presented herein would be applicable to most currently proposed evidence-based
guidelines with only minor modifications.
There are many possible reasons for this large divergence between routine
practice on the one hand and clinical trial data and evidence-based recommendations
on the other. Foremost among these is the vigorous marketing of newer, more
costly agents compared with virtually no marketing for older, off-patent drugs.
Such marketing affects both physician prescribing choices and patient preferences.
It is not clear from these analyses to what extent patient vs physician factors
cause these findings; this is an important area for future research.
These findings could have considerable national impact. We found potential
savings of $87.14 per patient treated for hypertension; the level of per-patient
spending on antihypertensive medication in our population is similar to that
found in other research.41 According to the
National Center for Health Statistics, in 2000 almost 22 million elderly Americans
had hypertension.42 Survey data indicate that
63% of older hypertensive patients are being treated,43 translating
to more than 13.5 million elderly patients taking antihypertensive medication.
Applying our findings to that population corresponds to a decrease in spending
for antihypertensive medications of about $1.2 billion per year.
The ALLHAT results15 indicate that these
changes would have been preferred on clinical as well as financial grounds,
although some controversy persists about the role of ACE inhibitors vs thiazides
in some subgroups.16,44 As lower-priced
ACE inhibitors become more widely available, such evidence-based prescribing
would reduce the cost of antihypertensives by at least $2 billion to $3 billion
These data must be viewed in light of several considerations. While
we made many conservative assumptions in our estimates of potential therapeutic
substitutions, we did not interview or examine the patients, review their
primary medical records, or communicate with their physicians. As a result,
we were not able to completely eliminate the possibility of contraindications
for a substitute regimen or hidden indications for the regimens actually prescribed.
Diagnoses based on visits, hospitalizations, and filled prescriptions are
not as accurate as those measured by review of the primary medical record,
but they are a useful way to track whether a given patient ever had a diagnosis
of the conditions we studied. We included data on prior prescriptions back
to 1994 but could not exclude the possibility that some patients may have
taken a first-line agent prior to 1994. We did not define a patient-specific
replacement dose when switching a given patient to an evidence-based regimen,
and we used the average dose of all prescriptions for each major drug class.
Finally, we did not directly address issues of patient adherence45,46 but
assumed that a given patient would comply approximately the same with one
regimen as with another. This assumption has been borne out by the results
of the ALLHAT study,15 which even showed a
slight adherence advantage for thiazides.
The JNC VII guidelines on which we based this analysis were published
in 2003.14 However, these were quite similar
to the recommendations published in JNC VI, which appeared in 1997.47 The main difference was that JNC VII no longer recommended β-blockers
as equivalent to thiazides as first-line treatment for uncomplicated hypertension.14 While our findings focus on the potential savings
from application of the JNC VII recommendations, the results apply equally
well if the JNC VI guidelines, which were current at the time of the prescriptions
studied, are used as the basis for the analysis.
While we cannot conclude that a given substitution would have been appropriate
in any given patient, it is overwhelmingly likely based on extensive clinical
trial data that the proposed substitutions would have been clinically reasonable
in a large proportion of hypertensive patients. Not all patients will tolerate
first-line therapy. In ALLHAT, the 1-year cessation rates were 12.9% for thiazide-type
diuretics and 17.6% for ACE inhibitors,15 and
in recent trials of β-blockers, the cessation rate was about 27%.48,49 Applying these rates to our projections
and assuming that the same proportions of patients who switched to these regimens
would revert to their previous medications would reduce the potential savings
by $1.7 million, to $9.9 million. This amount still represents a 20.5% decrease
in program spending for antihypertensive medications.
These estimates of potential savings are conservative in several ways.
Since many thiaizides cost less than 8¢ per tablet, patients filling
these prescriptions would pay less than the $6 co-payment required for costlier
prescriptions. Thus, for many of the 430 514 prescriptions for which
thiazide replacement was suggested, patients would also have paid less out
of pocket, in addition to the program savings described.
We were also conservative in making exceptions for recommended therapy.
We excluded patients from substitution if they had filled even 1 previous
thiazide prescription. Many of these patients likely did not have intolerable
adverse effects and may well have been able to use thiazides instead of more
expensive antihypertensives, resulting in greater savings, but we did not
consider this possibility. Similarly, we followed the JNC VII guidelines and
considered any diagnosis of angina as an acceptable indication for maintaining
calcium channel blocker treatment,45 even though
other guidelines prefer β-blockers as in such patients.36,45 Likewise,
patients with a diagnosis of benign prostatic hypertrophy may have prostatic
enlargement without urinary symptoms or may have already had surgical correction
of the condition. Nevertheless, we used the JNC VII guidelines, which considered α-blockers
an acceptable therapeutic choice in such patients, although evidence from
ALLHAT would suggest replacing many of these α-blocker prescriptions
Bringing medication use closer to established practice guidelines would
require some initial costs for physician education as well as some additional
office visits at first as patients' prescriptions are converted to new antihypertensive
medications. Recent research addressed this problem directly in examining
the effect of a reference-pricing system and found that the costs of additional
physician visits associated with switching patients' antihypertensive therapy
were very small in relation to the savings achieved.40 Although
additional costs might be incurred when patients switch drug classes, the
savings from these changes would continue over years of subsequent drug use,
offsetting these modest initial implementation and transition costs.
Overall drug spending for Medicare beneficiaries in 2002 was projected
at $87 billion,3 and addition of a Medicare
prescription drug benefit will add substantially to this cost; estimates of
the eventual cost of this program have already increased dramatically from
$400 billion over 10 years to more than $530 billion.8 Initiatives
to improve prescribing could result not only in better clinical outcomes but
also in lower spending on medications for hypertension and, potentially, for
other conditions as well. Such rationalization of prescribing could make the
difference between programs that are fiscally viable and those that are not.50