Context Rising costs of medications and inequities in access have sparked calls
for drug policy reform in the United States and Canada. Control of drug expenditures
by prescription cost-sharing for elderly persons and poor persons is a contentious
issue because little is known about the health impact in these subgroups.
Objectives To determine (1) the impact of introducing prescription drug cost-sharing
on use of essential and less essential drugs among elderly persons and welfare
recipients and (2) rates of emergency department (ED) visits and serious adverse
events associated with reductions in drug use before and after policy implementation.
Design and Setting Interrupted time-series analysis of data from 32 months before and 17
months after introduction of a prescription coinsurance and deductible cost-sharing
policy in Quebec in 1996. Separate 10-month prepolicy control and postpolicy
cohort studies were conducted to estimate the impact of the drug reform on
adverse events.
Participants A random sample of 93 950 elderly persons and 55 333 adult
welfare medication recipients.
Main Outcome Measures Mean daily number of essential and less essential drugs used per month,
ED visits, and serious adverse events (hospitalization, nursing home admission,
and mortality) before and after policy introduction.
Results After cost-sharing was introduced, use of essential drugs decreased
by 9.12% (95% confidence interval [CI], 8.7%-9.6%) in elderly persons and
by 14.42% (95% CI, 13.3%-15.6%) in welfare recipients; use of less essential
drugs decreased by 15.14% (95% CI, 14.4%-15.9%) and 22.39% (95% CI, 20.9%-23.9%),
respectively. The rate (per 10 000 person-months) of serious adverse
events associated with reductions in use of essential drugs increased from
5.8 in the prepolicy control cohort to 12.6 in the postpolicy cohort in elderly
persons (a net increase of 6.8 [95% CI, 5.6-8.0]) and from 14.7 to 27.6 in
welfare recipients (a net increase of 12.9 [95% CI, 10.2-15.5]). Emergency
department visit rates related to reductions in the use of essential drugs
also increased by 14.2 (95% CI, 8.5-19.9) per 10 000 person-months in
elderly persons (prepolicy control cohort, 32.9; postpolicy cohort, 47.1)
and by 54.2 (95% CI, 33.5-74.8) among welfare recipients (prepolicy control
cohort, 69.6; postpolicy cohort, 123.8). These increases were primarily due
to an increase in the proportion of recipients who reduced their use of essential
drugs. Reductions in the use of less essential drugs were not associated with
an increase in risk of adverse events or ED visits.
Conclusions In our study, increased cost-sharing for prescription drugs in elderly
persons and welfare recipients was followed by reductions in use of essential
drugs and a higher rate of serious adverse events and ED visits associated
with these reductions.
Rising costs of medications and inequities in access to medication have
sparked calls for drug policy reform in the United States and Canada.1,2 One of the most contentious issues
is the introduction of cost-sharing to control drug expenditures. Cost-sharing
is intended to deter the use of drug therapies that do little to improve health.3-6 But cost-effectiveness
rests on the assumption that individuals will have the capacity to pay for
essential drugs and that they will make rational choices about which drugs
to use and abandon. Otherwise, the use of essential drugs will be curtailed
to control drug expenditures and short-term savings in the drug budget may
be offset by downstream costs in the treatment of potentially preventable
illnesses.7-22
Cost-sharing policies that require the consumer to pay a percentage
of the cost of the prescribed medication may be less likely to produce unintended
health effects. But the impact of this type of cost-sharing on the use of
health services has not been evaluated in poor persons, and only one inconclusive
evaluation has been conducted for elderly persons.10
In 1996, the Canadian province of Quebec attempted to enhance equity
of access to prescription drugs by legislating mandatory drug insurance for
all residents.23 To help finance this coverage,
a deductible and a 25% coinsurance charge was instituted for previously insured
beneficiaries who received free medication: elderly persons and welfare recipients.
The existence of comprehensive, linkable databases on drug and health care
use in Quebec provided a unique opportunity to evaluate the impact of cost-sharing
on the use of essential and less essential drugs and determine the rate of
emergency department (ED) visits and serious adverse events (hospitalization,
nursing home admission, and mortality) associated with reductions in drug
use before and after the implementation of the cost-sharing policy.
Quebec Drug Policy Reform
Prior to the drug policy reform, welfare recipients and low-income elderly
persons (5.7% of the total elderly population in Quebec) received free medication,
and all other elderly persons paid $2 per prescription up to a maximum of
$100 per year (all values are in Canadian dollars). On August 1, 1996, both
groups became subject to a 25% coinsurance fee up to an annual maximum of
$200 for welfare recipients and income-indexed ceilings of $200, $500, and
$750 for elderly persons. On January 1, 1997, an annual deductible of $100
was added to the 25% coinsurance fee, with the annual deductible and ceiling
prorated on a quarterly basis. On July 1, 1997, the annual ceiling and deductible
were prorated on a monthly basis to reduce the amount a recipient would have
to pay per month to a maximum of $16.67 ($200 plan) to $62.50 ($750 plan).
Children and selected subgroups were exempted from cost-sharing and continued
to receive free medication. A computerized, real-time link between the provincial
health plan administration (Régie de l'assurance maladie du Québec
[RAMQ]) and pharmacies was put in place to administer the policy.
Design and Study Populations
To study the intended impact of cost-sharing and to evaluate the health
effects of reducing drug use, we identified a subset of essential and less
essential drugs for evaluation using published classifications (Table 1).18,24,25Essential drugs were defined as medications that prevent
deterioration in health or prolong life and would not likely be prescribed
in the absence of a definitive diagnosis. Because there was no information
on indication for treatment, the second criterion was added to limit the analysis
to only those people who were using drugs for which reductions in use would
be expected to have an impact on health status. The most commonly used drugs
excluded using this criterion were antibiotics and gastro-protective medications,
drugs targeted by many intervention programs to reduce overuse.26Less essential drugs were defined as medications that may
provide relief of symptoms but will likely have no effect on the underlying
disease process.25 Drugs included in the essential
and less essential categories were independently identified and then rated
by the 7 clinical coinvestigators. A medication was included if 6 of the 7
clinical coinvestigators agreed with the classification of the respective
drug and it was insured through the RAMQ provincial health care plan. The
subset of drugs included in the evaluation of essential and less essential
drug use represented 21% of all insured drugs and 55% of all prescriptions
dispensed.
An interrupted time-series design was used to estimate changes in prescription
drug use after the introduction of the drug policy. Monthly measures of overall
drug use as well as the use of essential and less essential drugs in the 3
years preceding the policy and in the first 17 postpolicy months were calculated
for individual persons. A random sample of 120 000 welfare recipients
and 120 000 elderly persons was selected. Persons were eligible for inclusion
in the analysis if they were alive on the date of policy implementation, filled
at least 1 prescription for a drug in the year before the policy, and were
not exempted from the cost-sharing policy. The evaluation of essential and
less essential drugs was limited to persons who had filled at least 1 prescription
for drugs within these categories in the prepolicy year. Individuals contributed
to the analysis only during the months in which they were enrolled.
Separate prepolicy control and postpolicy cohort studies were conducted
to determine the impact of the policy on the rates of adverse events and ED
visits associated with reductions in use of essential and less essential drugs.
Each study produced an estimate of the risk associated with reduction in drug
use, the proportion of the population making reductions, the population attributable
fraction, and the rate of adverse events and ED visits associated with reductions
in drug use. The 2 studies were conducted in comparable 10-month periods before
(August 1995 to 1996) and after (August 1996-1997) policy implementation (Figure 1). The prepolicy control study provided
an estimate of the expected rate of adverse events due to reductions in drug
use prior to policy implementation. The estimation of an expected rate was
important because even when drugs are free, individuals will experience adverse
drug events due to injudicious reductions in needed therapy, because of forgetfulness,
adverse effects, or misperceptions about the importance of drug treatment.27 The difference in the rate in the prepolicy control
study vs the postpolicy study was used to estimate the impact of the drug
reform on adverse events. This approach had several advantages. First, it
voided biases related to ecological fallacy28
because changes in drug use were linked at the level of the individual with
the occurrence of adverse events and ED visits. Second, it provided a means
of isolating the effect of the drug policy from other health care policies
that were implemented in the same 4-year period that may have reduced the
rate of ED visits and hospitalizations, unrelated to prescription drug use
(hospital closures and reallocation of service locations). Finally, the prepolicy
and postpolicy cohort study approach verified the assumption that the primary
impact of cost-sharing would be to increase the prevalence of reductions of
drug use rather than changing the "biological risk" associated with rationing
or stopping therapy. Thus, the studies were designed to estimate both the
risk and the population attributable fraction or the share of adverse events
and ED visits due to reductions in drug use, in the prepolicy and postpolicy
periods.
For this analysis, study populations were limited to regular recipients
of essential or less essential drugs, defined as persons who had a supply
of the respective medication in each of the 12 months prior to the follow-up
period or new users with a minimum of 6 months of continuous use.
Four provincial health databases, validated in previous research,29-31 were linked by unique
encrypted health numbers. The beneficiary demographic database provided data
on drug plan eligibility, death, and beneficiary characteristics. The prescription
claims database, which includes the drug, quantity, date, and duration for
each prescription dispensed from community-based pharmacies, was used to measure
medication use. The physician claims database, which includes the date, type,
and location of service delivery (eg, inpatient, emergency, clinic), was used
to measure ED visits and hospitalization-institutionalization. The hospitalization
database was used to validate claims-based measures of hospitalization-institutionalization.
The number of drugs available each day was calculated from prescription
claims records using methods developed to convert the date, drug, and duration
of prescriptions dispensed into a drug-by-day matrix.32
In each of the 53 months of the time series, a matrix of monthly mean daily
drug use was then constructed for each beneficiary (for all drugs and separately
for essential and less essential drugs). The first 3 months of the time series
and of coverage for newly eligible recipients were excluded to avoid artificially
lower values for drug use in the first few months of available prescription
information. The month immediately prior to policy implementation also was
excluded because of possible prescription stockpiling, leaving 49 months for
analysis.
For the prepolicy and postpolicy cohort studies, reductions in drug
use were measured first by estimating an expected daily drug use for each
person. The resulting expected values were then compared with observed use
in the 10-month follow-up period. The expected use value was estimated as
the level predicted for the last baseline month by a linear trend fit to each
person's mean monthly daily drug use in the baseline year. This method conservatively
assumed that expected drug use would remain constant rather than increase
during follow-up. In addition, it was assumed that the impact of reductions
in drug use would cumulate over time. Therefore, time-dependent measures were
used to summarize differences between expected and observed use during the
follow-up period. Time-dependent measures of drug use also provided a means
of adjusting for unusual drug consumption patterns triggered by the features
of the drug policy. For instance, the deductible and maximum ceilings instituted
a pattern whereby reductions in one month may be compensated for by increases
in the next when drugs were free for those persons reaching the spending ceiling.
Cumulative mean monthly increases (observed > expected) and reductions (observed
< expected) in drug use were calculated as the sum (from the first follow-up
month) of the monthly difference in observed and expected drug use divided
by the number of follow-up months. For example, an individual who had an expected
value of 5 drugs per month and who filled prescriptions for 3 drugs in the
first 2 months of follow-up and 8 in the third month would have a mean cumulative
reduction value in month 3 of 1.33 ([2 + 2 + 0]/3) and a mean cumulative increase
of 1 ([0 + 0 + 3]/3). To respect the temporal sequence between reductions
in drug use and adverse events, we linked the occurrence of adverse events
in any 1 month to the patient's mean drug use pattern in the preceding months.
ED Visits and Adverse Events
For each person in each month, the number of distinct days with a claim
for an ED visit was counted. Adverse events, defined as the first occurrence
of acute care hospitalization, long-term care admission, or death, were established
using the date and location of health service in the physician claims database
(for hospitalizations and institutionalizations) and the beneficiary demographic
database (for death). Follow-up was terminated after the occurrence of the
first adverse event because prescription drugs dispensed during hospital stays
are not recorded in claims files, and there are no available methods to establish
long-term predictions of expected drug use at the level of the individual.
The impact of the drug policy on monthly drug use was adjusted for linear
trend across time, season (11 monthly indicator variables), and individual
characteristics (age, sex, type of insurance plan, baseline medication use,
education, and income). Education and income were approximated by the mean
census value33 for residents within the same
postal code area (mean, 366 households). Analysis of adverse events and ED
visits was adjusted for age, sex, number of months hospitalized, Charlson
comorbidity values34,35 in the
baseline year, expected level of medication use, and mean cumulative increases
in drug use. For ED visits, we also adjusted for the individual's frequency
of ED visits in the baseline year.
To estimate the impact of the policy on the level of drug use over time,
random-effects, pooled-time series regression36
was used based on 49 prepolicy and postpolicy months. The individual was the
unit of analysis, and an autoregressive first-order correlation structure
(AR[1]) was assumed to represent the dependence among subsequent observations.
Alternate correlation structures also were assessed (AR [2] and AR [12]),
but as there was no material impact of the choice of correlation structure
on the estimated regression coefficients or SEs, AR (1) results were presented.
To assess postpolicy reductions in drug use, we first used the 32 prepolicy
months to estimate the effect of time, seasonal variation, and beneficiary
characteristics and then to predict, for each person, expected use for the
17 months postpolicy as well as the absolute and relative differences between
observed and expected use.
The occurrence of adverse events associated with reduction in use of
essential and less essential drugs was estimated by hazard ratios derived
from a Cox model using time-dependent covariates, modeled as continuous variables,
to represent monthly changes in drug use. Proportional hazards and linearity
assumptions were verified. The frequency of ED visits was modeled by Poisson
regression, assuming a first-order autoregressive correlation structure in
a generalized estimating equation approach37,38
and empirical SEs to account for overdispersion. Outcomes in a particular
month were examined in relation to cumulative mean monthly measures of drug
use in the previous month. For the prepolicy control and postpolicy cohort
studies, the overall fraction of adverse events and ED visits associated with
reductions in drug use was estimated separately using the approach of Greenland
and Drescher39 and multiplied by the overall
adverse event and ED visit rates to estimate the monthly rate of adverse events
and ED visits associated with reductions in drug use. As no estimate of the
intercept is available in the Cox model, logistic regression was used to estimate
the proportion of adverse events associated with reductions (population-attributable
fraction) for adverse events. The regression coefficients produced by the
Cox and logistic models were almost identical. Standard errors and confidence
intervals (CIs) for the difference in these adverse event rates in the prepolicy
and postpolicy studies were estimated by bootstrapping,40,41
using 200 bootstrap samples.
Among the 120 000 elderly persons sampled, 117 408 were eligible
for cost-sharing. Of the 117 408, 93 950 (80.0%) used prescription
medication in the prepolicy year, and 70 801 of the 93 950 (75.3%)
used essential drugs (Table 2).
Among the 120 000 welfare recipients sampled, 84 272 were adults
of whom 55 333 (65.7%) used medication in the prepolicy year and 25 820
(46.7%) used medications classified as essential drugs. Elderly persons used
more drugs per day and had higher costs for therapy than welfare recipients.
However, the proportions hospitalized in the prepolicy year were similar.
Elderly persons and welfare recipients used at least twice as many essential
as less essential drugs. The most commonly used essential drugs in the elderly
group were aspirin (13.2%), levothyroxine sodium (7.5%), and furosemide (6.0%).
Among welfare recipients, the most commonly used essential drugs were salbutamol
(5.9%) and levothyroxine sodium (5.0%). Benzodiazepines were the most commonly
used less essential medication in the elderly and welfare recipient groups.
The introduction of the drug policy was followed by substantial and
statistically significant reductions in the overall number of drugs used per
day by both elderly persons (9.14%; 95% CI, 8.8%-9.5%) and adult welfare recipients
(15.94%; 95% CI, 15.0%-16.9%). Similar reductions also were seen for the use
of essential drugs as well as a slower rate of increase in drug use over time
(policy/time interaction: elderly persons: = −0.02;
P<.001;
welfare recipients: = −0.009;
P<.001)
(Figure 2). In the postpolicy period, elderly persons showed a 9.12%
(95% CI, 8.7%-9.6%) reduction in the number of essential drugs used per day
(0.17 drugs; 95% CI, 0.16-0.18). Absolute and relative reductions were higher
among welfare recipients (14.4%; 95% CI, 13.3%-15.6% and absolute reduction:
0.21; 95% CI, 0.19-0.23 essential drugs per day).
Relative reductions were greater in the use of less essential drugs
by elderly persons and welfare recipients (15.14%; 95% CI, 14.4%-15.9% and
22.39%; 95% CI, 20.9%-23.9%, respectively) than for essential drugs (Figure 3). However, because fewer less essential
drugs were used per day, the absolute size of the reduction was smaller for
less essential drugs (elderly persons, 0.10 and welfare recipients, 0.15)
than for essential drugs (elderly persons, 0.17 and welfare recipients, 0.21).
Also, there was a significant decrease in the slope of less essential drug
use over time in the
postpolicy period (policy/time interaction) for the elderly
persons ( = −0.009;
P<.001) and for the welfare
recipients ( = −0.008;
P<.001).
As expected, in both the prepolicy and postpolicy studies, there was
a significantly higher rate of adverse events and ED visits in those individuals
who reduced their use of essential drugs vs those who did not (Table 3). Dose-response relationships were evident between the magnitude
of the reduction and the rates of both outcomes. For example, in the prepolicy
control study, the rates of adverse events in those with no reduction (≤0.1
drugs/d), minor reduction (>0.1 to 0.5 drugs/d), and major reductions (≥1
drugs/d) were 256, 272, and 385 per 10 000 person-months, respectively.
Reduction of 1 medication would be equivalent to stopping 1 drug or rationing
2 drugs to half the expected use. Risks associated with reductions in drug
use leveled off at higher reductions, reaching a plateau at approximately
1.5 medications per day. The estimated risk ratio for a 1-drug reduction,
adjusted for cumulative increases, was similar for the prepolicy (1.57) and
postpolicy (1.56) studies. Adjustment for beneficiary characteristics lowered
the estimated risk ratios, primarily because individuals who had a greater
number of comorbidities and used more medication were more likely to reduce
their use of drugs and to experience adverse events and ED visits.
A greater proportion of elderly persons (7.4% more) in the postpolicy
study compared with the prepolicy control study reduced their use of essential
drugs (Table 3). As a result,
the estimated proportion of adverse events associated with drug reductions
increased from 2.0% in the prepolicy to 3.9% in the postpolicy period. This
change was accompanied by an estimated increase in the monthly rate of adverse
events related to reductions in essential drug use from 5.8 to 12.6 per 10 000
in the prepolicy and postpolicy periods; a net increase of 6.8 adverse events
per 10 000 per month. Reduction in the use of essential drugs also was
associated with a significant increase in the rate of ED visits (risk ratio,
1.17-1.19). The higher frequency of reductions in essential drug use in the
postpolicy period resulted in a net increased rate in monthly ED visits related
to reductions in drug use of 14.2 per 10 000.
Welfare recipients, as compared with elderly persons, experienced similar
risks of adverse events and ED visits in relationship to reductions in essential
drug use (Table 4); however, the
frequency of postpolicy reductions in essential drug use was considerably
greater (13.6% and 7.4%, respectively). As a result, the drug policy was associated
with a substantially greater net increase in the rates of adverse events and
ED visits in relation to reductions in the use of essential drugs. In the
postpolicy study, relative to the prepolicy study, the monthly rate of adverse
events increased by 12.9 per 10 000 and ED visit rates by 54.2 per 10 000
person-months.
Reductions in less essential drug use had no significant impact on adverse
events or ED visits among elderly persons and welfare recipients in either
the prepolicy control or postpolicy studies (Table 5).
Increased cost-sharing for prescription drugs had the desired effect
of reducing the use of less essential drugs but also the unintended effect
of reducing the use of drugs that are essential for disease management and
prevention. As a result, in the postpolicy period, there was an increase in
the rate of adverse events and ED visits related to reductions in essential
drug use. These effects are consistent with those reported by Soumerai et
al18,20 in the evaluation of prescription
reimbursement caps in US Medicaid programs.
This study had several limitations. In the absence of an out-of-province
control group, we cannot confirm that reductions in drug use were solely related
to cost-sharing policies. However, no other plausible explanation has emerged
and the suddenness of reductions in drug use that occurred immediately after
the introduction of the cost-sharing policy strengthens the conclusion that
changes in drug use were likely to be related to the cost-sharing policy.
Prescription claims files do not indicate what drugs were taken, only medication
purchased. Although prescription refill rates provide a reasonably accurate
measure of medication compliance,42-44
reductions in drug use could have been overestimated if individuals received
free samples or purchased equivalent over-the-counter preparations (eg, aspirin)
after policy implementation. However, these individuals would be falsely classified
as having reduced medication use, and as a result, the risk associated with
reductions in drug use in the postpolicy studies would be underestimated.
Indications for therapy were unknown. Drugs classified as less essential
may have been required therapy for some individuals (eg, benzodiazepines for
panic disorder), whereas some essential therapeutic drugs may have been prescribed
without adequate clinical indication (eg, diuretics for transitory elevation
in blood pressure). This misclassification would likely lead to an underestimation
of both the potential benefits of reducing the use of less essential drugs
and the risks of reducing essential drug therapy.
Our study suggests that the primary mechanism by which cost-sharing
affected the rate of adverse events was by increasing the proportion of people
who made reductions in the use of essential drugs. We cannot confirm that
reductions in essential drug use led to a deterioration in health status,
but we believe that this is a plausible explanation for several reasons. First,
there was a dose-response relationship between the magnitude of the reduction
in the use of essential drugs and the risk of adverse events and ED visits.
Second, reductions were associated with an increase in the risk of adverse
events in the prepolicy and postpolicy period, a phenomenon that would be
expected if reductions represented medication noncompliance. Finally, the
risk associated with reduction was specific to essential drugs, for which
there is clinical trial evidence of efficacy.
The challenge for insurers has been to craft health care policies that
provide adequate access to drug therapy while simultaneously exercising fiscally
responsible control over the drug budget. Consumer cost-sharing has been the
principal method of fiscal control because it assumes that people will value
what they pay for and as a result, they will reduce their use of unnecessary
medication when they are required to contribute a portion of the payment.4,6 While this reasoning may apply to many
consumer goods, cost-sharing has been shown to have unintended effects in
health care, such as increasing hospital admissions.16,17,19,20,45-48
Consumers may not have the information needed to make wise decisions about
necessary treatment. We estimate that for elderly persons alone, the drug
policy reform in Quebec may result in 7000 additional adverse events per million
annually. In light of the substantial impact that drug policy can have on
the population's health, there is a need to redress the relative scarcity
of scientific data on the outcomes of policy interventions. Our results suggest
that more stringent cost-sharing pharmaceutical cost containment policies
in other parts of Canada49 and the United States1,4 may contribute to avoidable illnesses.
1.Soumerai SB, Ross-Degnan D. Inadequate prescription-drug coverage for medicare enrollees: a call
to action.
N Engl J Med.1999;340:722-728.Google Scholar 2.Minister of Public Works and Government Services. Canadian Health Action: Building on the Legacy: Synthesis
Reports and Issues Papers. Ottawa, Ontario: National Forum on Health; 1997. Cat No. H21-126/5-2-1997E.
3.Soumerai SB, Ross-Degnan D, Fortess EE, Abelson J. A critical analysis of studies of state drug reimbursement policies:
research in need of discipline.
Milbank Q.1993;71:217-252.Google Scholar 4.Levy R. Prescription cost sharing economic and health impacts, and implications
for health policy.
Pharmacoeconomics.1992;2:219-237.Google Scholar 5.Huttin C. The use of prescription charges.
Health Policy.1994;27:53-73.Google Scholar 6.Reeder CE, Lingle EW, Schulz RM.
et al. Economic impact of cost-containment strategies in third party programmes
in the US (part 1).
Pharmacoeconomics.1993;4:92-103.Google Scholar 7.Hurley J, Arbuthnot Johnson N. The effects of co-payments within drug reimbursement programs.
Can Public Policy.1991;17:473-489.Google Scholar 8.Ryan M, Birch S. Charging for health care: evidence on the utilisation of NHS prescribed
drugs.
Soc Sci Med.1991;33:681-687.Google Scholar 9.Martin BC, McMillan JA. The impact of implementing a more restrictive prescription limit on
Medicaid recipients.
Med Care.1996;34:686-701.Google Scholar 10.Johnson RE, Goodman RM, Hornbrook MC, Eldredge MB. The effect of increased prescription drug cost-sharing on medical care
utilization and expenses of elderly health maintenance organization members.
Med Care.1997;35:1119-1131.Google Scholar 11.Leibowitz A, Manning WG, Newhouse JP. The demand for prescription drugs as a function of cost-sharing.
Soc Sci Med.1985;21:1063-1069.Google Scholar 12.Harris BL, Stergachis A, Ried LD. The effect of drug co-payments on utilization and cost of pharmaceuticals
in a health maintenance organization.
Med Care.1990;28:907-917.Google Scholar 13.O'Brien B. The effect of patient charges on the utilization of prescription medicines.
J Health Econ.1989;8:109-132.Google Scholar 14.Smith DG. The effects of copayments and generic substitution on the use and costs
of prescription drugs.
Inquiry.1993;30:189-198.Google Scholar 15.Nelson AA, Reeder CE, Dickson M. The effect of a Medicaid drug copayment program on the utilization
and cost of prescription services.
Med Care.1984;22:724-736.Google Scholar 16.Lohr KN, Brook RH, Kamberg C.
et al. Use of medical care in the RAND health insurance experiment: diagnosis
and service-specific analyses in a randomized controlled trial. Santa Monica, Calif: RAND; 1986.
17.Reeder CE, Nelson AA. The differential impact of copayment on drug use in a Medicaid population.
Inquiry.1985;22:396-403.Google Scholar 18.Soumerai SB, Avorn J, Ross-Degnan D, Gortmaker S. Payment restrictions for prescription drugs under Medicaid: effects
on therapy, cost, and equity.
N Engl J Med.1987;317:550-556.Google Scholar 19.Soumerai SB, McLaughlin TJ, Ross-Degnan D, Casteris CS, Bollini P. Effects of limiting Medicaid drug-reimbursement benefits on the use
of psychotropic agents and acute mental health services by patients with schizophrenia.
N Engl J Med.1994;331:650-655.Google Scholar 20.Soumerai SB, Ross-Degnan D, Avorn J, McLaughlin TJ, Choodnovskkiy I. Effects of Medicaid drug-payment limits on admission to hospitals and
nursing homes.
N Engl J Med.1991;325:1072-1077.Google Scholar 21.Brook RH, Ware JE, Rogers WH.
et al. Does free care improve adults' health? results from a randomized controlled
trial.
N Engl J Med.1983;309:1426-1434.Google Scholar 22.Newhouse JP.The Insurance Experiment Group. Free for All? Lessons From the RAND Health Insurance Experiment . London, England: Harvard University Press; 1993.
23.Martin M. Quebec considering universal drug insurance.
CMAJ.1996;154:1264.Google Scholar 24.World Health Organization. Essential drugs.
WHO Drug Information.1999;13:249-262.Google Scholar 25.McGavock H, Wilson-Davis K, Niblock RWF. Unsuspected patterns of drug utilization revealed by interrogation
of a regional general practitioner prescribing database.
Pharmacoepidemiol Drug Saf.1992;1:73-80.Google Scholar 26.Soumerai SB, McLaughlin TJ, Avorn J. Improving drug prescribing in primary care: a critical analysis of
the experimental literature.
Milbank Q.1989;67:268-317.Google Scholar 27.Goetghebeur E, Molenberghs G, Katz J. Estimating the causal effect of compliance on binary outcomes in randomized
controlled trials.
Stat Med.1998;17:341-355.Google Scholar 28.Morgenstern H. Ecological studies. In: Rothman K, Greenland S, eds. Modern Epidemiology. Philadelphia, Pa: Lippincott-Raven; 1998:459-480.
29.Tamblyn RM, Lavoie G, Petrella L, Monette J. The use of prescription claims databases in pharmacoepidemiological
research: the accuracy and comprehensiveness of the prescription claims database
in Quebec.
J Clin Epidemiol.1995;48:999-1009.Google Scholar 30.Tamblyn RM, Reid T, Mayo N, McLeod PJ, Churchill-Smith M. Using medical services claims to assess injuries in the elderly: the
sensitivity of diagnostic and procedure codes for injury ascertainment.
J Clin Epidemiol.2000;53:183-194.Google Scholar 31.Demers M. Cout des services médicaux durant la dernière
année de vie au Québec en 1991. Québec, Direction générale de l'évaluation
et du contrôle, Régie de l'assurance-maladie du Québec;
1994:ix-39.
32.Tamblyn R, Abrahamowicz M. Drug utilization patterns. In: Armitage P, Coulton T, eds. Encyclopedia of
Biostatistics. West Sussex, England: John Wiley & Sons Ltd; 1998:1235-1247.
33.Wilkins K. Use of postal codes and addresses in the analysis of health data.
Health Rep.1993;5:157-177.Google Scholar 34.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal
studies: development and validation.
J Chronic Dis.1987;40:373-383.Google Scholar 35.Deyo RA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative
data: a response.
J Clin Epidemiol.1993;46:1075-1079.Google Scholar 36.Ward MM, Leigh JP. Pooled time series regression analysis in longitudinal studies.
J Clin Epidemiol.1993;46:645-659.Google Scholar 37.Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models.
Biometrics.1986;73:13-22.Google Scholar 38.Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes.
Biometrics.1986;42:121-130.Google Scholar 39.Greenland S, Drescher K. Maximum likelihood estimation of the attributable fraction from logistic
models.
Biometrics.1993;49:865-872.Google Scholar 40.Rockhill B, Weinberg CR, Newman B. Population attributable fraction estimation for established breast
cancer risk factors: considering the issues of high prevalence and unmodifiability.
Am J Epidemiol.1998;147:826-833.Google Scholar 41.Efron B, Tibshirani R. An Introduction to the Bootstrap. New York, NY: Chapman & Hall; 1993.
42.Grymonpre RE, Didur CD, Montgomery PR, Sitar DS. Pill count, self-report, and pharmacy claims data to measure medication
adherence in the elderly.
Ann Pharmacother.1998;32:749-754.Google Scholar 43.Inui TS, Carter WB, Pecoraro RE, Pearlman RA, Dohan JJ. Variations in patient compliance with common long-term drugs.
Med Care.1980;18:986-993.Google Scholar 44.Lau HS, de Boer A, Beuning KS, Porsius A. Validation of pharmacy records in drug exposure assessment.
J Clin Epidemiol.1997;50:619-625.Google Scholar 45.Lurie N, Ward NB, Shapiro MF, Gallego C, Vaghaiwalla R, Brook RH. Termination of medical benefits: a follow-up study one year later.
N Engl J Med.1986;314:1266-1268.Google Scholar 46.Simon GE, Grothaus L, Durham ML, VonKorff M, Pabiniak C. Inpact of visit copayments on outpatient mental health utilization
by members of a health maintenance organization.
Am J Psychiatry.1996;153:331-338.Google Scholar 47.Bloom BS, Jacobs J. Cost effects of restricting cost-effective therapy.
Med Care.1985;23:872-879.Google Scholar 48.Simon GE, VonKorff M, Durham ML. Predictors of outpatient mental health utilization by primary care
patients in a health maintenance organization.
Am J Psychiatry.1994;151:908-913.Google Scholar 49.Bacovsky RA. Federal, Provincial and Territorial Government-Sponsored Drug Plans
and Drug Databases. Health Canada 1997; Background information Prepared for the Conference
on National Approaches to Pharmacare on January 18-20, 1998 (Saskatoon, Saskatchewan).