OUP user menu

Drug treatment of stable angina pectoris and mass dissemination of therapeutic guidelines: a randomized controlled trial

M.-D. Beaulieu, J. Brophy, A. Jacques, R. Blais, R. Battista, R. Lebeau
DOI: http://dx.doi.org/10.1093/qjmed/hch006 21-31 First published online: 31 December 2003

Abstract

Background: Public agencies responsible for implementing health care policies often adapt and disseminate clinical practice guidelines, but the effectiveness of mass dissemination of guidelines is unknown.

Aim: To study the effects of guideline dissemination on physicians’ prescribing practices for the treatment of stable angina pectoris.

Design: Randomized controlled trial.

Methods: A sample of 3293 Quebec physicians were randomly assigned to receive a one-page summary of clinical practice guidelines on the treatment of stable angina (in February 1999), to receive the summary and a reminder (in February and March 1999, respectively), or to receive no intervention (controls). The prescribing profiles of participants, as well as sociodemographic characteristics of the physicians and their patients, were examined for June–December 1999.

Results: The intervention had no effect on prescription rates of β-blockers, antiplatelet agents, or hypolipaemic drugs. Compared to 1997 data for the same physicians, there was an overall 10% increase in appropriate prescription rates, irrespective of the intervention.

Discussion: In-house production and dissemination of clinical practice guidelines may not improve physicians’ practice patterns if there is pre-existing substantial scientific consensus on the issue.

Introduction

When public agencies are responsible for implementing health care policies, their strategy usually includes the in-house production and mass dissemination of clinical practice guidelines, even if such guidelines are available from national or international professional associations. Although such public agencies do not usually ‘reinvent the wheel’, they do commit substantial resources to adapting existing guidelines and disseminating them under their own auspices. This process is perceived as necessary to establish local (or regional) credibility of guidelines, and is often the best way that the agencies can reach all their constituencies.

Literature on the impact of large-scale dissemination of guidelines is scarce and inconsistent, leading to uncertainty about its value. Dissemination of guidelines appears less effective than intense continuing medical education (CME)-based interventions.1–,3 Nonetheless, guideline availability may contribute to meaningful changes in practice patterns when applied to large populations.4,,5 However, the impact of such mass dissemination, and whether this relates to the novelty of the message, the complexity of the guidelines, characteristics of the physicians, or uptake by other stakeholders such as hospitals and local professional associations, is often unclear.3,5,,6

We have previously demonstrated significant under-use of recommended anti-anginal therapies in some regions within the province of Quebec,7 despite the existence of provincial guidelines.8 Therefore, we performed a randomized controlled trial to study the effects of guideline dissemination on physicians’ prescribing practices for the treatment of stable angina pectoris. This study was one component of a government-funded initiative to support the development of a ‘guideline infrastructure’ in the province.

Methods

Study population

The study population and sampling procedure have been described in detail previously.7 Briefly, physicians from three geographically distinct regions were identified through the computerized administrative databases of the Régie de l’Assurance Maladie du Québec (RAMQ, the Quebec Health Insurance Board), which contain data for all physician visits, interventions, and prescriptions for Quebec residents 65 years of age and older. To be eligible for the survey, physicians had to have been part of our 1997 study,7 be the primary prescribing physician (responsible for more than half of all anti-anginal prescriptions) for at least one patient (see below), and still be prescribing cardiovascular medications as of 30 December 1999. Since each prescription delivered by the pharmacist is linked to the physician who wrote the prescription, it was possible, from the database, to include only prescription data from the identified prescribing physician, that is, prescription data is specific to the allocated primary physician, thus avoiding prescriber-prescription misclassification.

The validity of these databases is well established.9 Nevertheless, to confirm the diagnosis of stable angina further, we developed an algorithm using information in the database about medications, interventions, and hospital admission.7 Patients who matched both the ‘claim diagnosis’ and the ‘algorithm diagnosis’ were then deemed to be undergoing treatment for stable angina. For each of these patients, we identified the primary prescribing physician.

Intervention

The existing provincial guidelines for anti-anginal therapy8 had been endorsed by the provincial licensing authority, the College des Médecins du Québec (CMQ), which then developed a user-friendly, one-page summary. This summary incorporated three key messages targeting the most problematic prescribing practices identified in our earlier cross-sectional study,7 namely low prescribing rates for antiplatelet and hypolipaemic drugs and for β-blockers in patients without apparent major contra-indications. The key recommendations in the summary were: (i) to write a prescription for acetylsalicylic acid for patients with stable angina; (ii) to control serum cholesterol, with a target value for low-density lipoprotein cholesterol < 2.6 mmol/l; and (iii) to favour β-blockers as the first choice for anti-angina medication. Data on prescribing rates for the three targeted medication classes by physicians practicing in the same regions as the participating physicians were also included in the one-page summary.

Randomization

The physicians identified in our previous study were randomly assigned, using computer-generated random numbers, to one of three groups. The first group received no intervention (n = 1091), the second group received the one-page summary of the guidelines (n = 1087), and the third group received the one-page summary, followed a month later by a reminder notice, which included stickers to post on patients’ charts (n = 1115). The samples thus assembled were sent to the CMQ, which, by law, is the only body that can access the encrypted physician identifiers used in the administrative databases. The CMQ mailed the one-page summary to the two intervention groups in February 1999, and the reminder notice to the second intervention group in March 1999.

Outcome variables

We studied the prescribing profiles of the participating physicians for the period June–December 1999 by examining data in the RAMQ files, as described previously.7 A 6-month period between the intervention and the outcome measures was chosen to reflect the usual follow-up period for patients with stable angina, as suggested by our visit database. A shorter observation period was deemed to short to show any changes, and a longer one, too far from the intervention to be confident in attributing observed changes to it.

We defined three outcome variables according to the guidelines:8 exposure to β-blockers as the anti-ischaemic regimen of choice (as monotherapy or with calcium-channels blockers and/or nitrates); exposure to antiplatelet drugs (which is supposed to be part of all regimens, unless contra-indications are present); and exposure to hypolipaemic drugs (which were then recommended if LDL-cholesterol was > 2.6 mmol/l). Since each class of drugs has its contra-indications, we considered the exposure to each one as independent, even if it was likely that the majority of patients would receive a drug from each of the three classes. The following classes of medications were considered to represent treatment of stable angina: anti-ischaemics (β-blockers, calcium-channel blockers, long-acting nitrates), antiplatelet agents, and hypolipaemics.

Explanatory variables

In addition to information on physician visits and prescriptions, the RAMQ files contained sociodemographic characteristics of both patients and physicians (including, for the latter, year of graduation and speciality). We created a variable ‘co-treatment’. Co-treatment between a general practitioner and a cardiologist or internist was deemed to exist if there was concomitant billing during the year for a complete major medical examination for a different category of physicians than his principal prescribing physician. The patient could appear only once in the database.

Comorbidities were identified by drug usage: oral hypoglycaemics, including acarbose with or without insulin, as an indicator of diabetes; theophylline with or without β2-agonist, steroid, or ipratropium inhaler, as an indicator of chronic obstructive pulmonary disease (COPD); and any two of the angiotensin-converting enzyme inhibitor, diuretic, or digoxin triad, as an indicator of heart failure.

Statistical analysis

For each outcome variable, multilevel logistic regression was used to study the impact of certain predictors at the patient and physician levels, while taking into account the covariance between observations sharing the same hierarchical structure. We used the formula proposed by Snijders and Bosker,10 which produces intra-class correlation coefficients as measures of the variation between physicians and the logit of the dependent variable. The ML-Win software package (version 1.0) was used.11

Results

Of the 3293 physicians in our initial study,7 967 (29.4%) were not in the database in 1999, hence were considered lost to follow-up. Thus 2326 (70.6%) were available for the current study (Figure 1). Since our database was anonymous, it was impossible to track down what happened to those physicians. The only way not to be in the database, was if a physician did not prescribe any cardiovascular medication as a principal prescriber in any of the three study regions, during the 1999 study period. We hypothesized that the majority had either moved or retired, since the RAMQ prescribing database is up-to-date without delays as far as physicians’ data is concerned. This hypothesis is supported by the observation that physicians who were not found in the 1999 database were elderly, hence with a greater probability of having retired (Table 1). Some may have been on sick leave or stopped prescribing cardiovascular medications, reflecting a career reorientation. However, ‘lost to follow-up’ was equally distributed in the three study groups.

Figure 1

Randomization of physicians in the study.

View this table:
Table 1

Characteristics of the 2326 physicians in the 1999 study database by study group, and comparing those ‘lost to follow-up’ between the 1997 and 1999 studies

CharacteristicControl (n = 767)Guideline (n = 766)Guideline + recall (n = 793)Total 1999 study group (n = 2326)Test of difference between groups χ2 (p value)Lost to follow-up between 1997–1999 (n = 967)Test of difference between 1999 study group vs. lost to follow-up χ2 (p value)
Sex
Male545 (71.1)541 (70.6)553 (69.7)1639 (70.5)0.34 (0.843)687 (71.0)0.11 (0.769)
Female222 (28.9)225 (29.4)240 (30.3)687 (29.5)280 (29.0)
Professional experience
< 10 years47 (6.1)42 (5.5)65 (8.2)154 (6.6)16.14 (0.013)91 (9.4)146.32 (0.000)
10–20 years232 (30.2)256 (33.4)294 (37.1)782 (33.6)269 (27.8)
21–30 years338 (44.1)323 (42.2)292 (36.8)953 (41.0)251 (26.0)
> 30 years150 (19.6)145 (18.9)142 (17.9)437 (18.8)356 (36.8)
Medical training
General practitioner650 (84.7)676 (88.3)665 (83.9)1991 (85.6)8.97 (0.175)856 (88.5)15.92 (0.000)
Cardiologist75 (9.8)65 (8.5)80 (10.1)220 (9.5)52 (5.4)
Internist42 (5.5)25 (3.2)48 (6.1)115 (4.9)59 (6.1)
Mean ± SD number of patients in database according to physician's training
General practitioner6.5 ± 7.06.5 ± 6.46.1 ± 6.76.4 ± 6.7Based on ANOVA test, 449.24 (0.000)Not availableNot calculated
Cardiologist39.8 ± 31.643.4 ± 40.943.6 ± 44.142.3 ± 38.9
Internist8.1 ± 11.610.1 ± 15.29.6 ± 11.19.3 ± 12.6
  • Data are numbers (%), except where stated.

Male doctors (70.5%), general practitioners (GPs) (85.6%), and doctors with > 10 years but < 31 years of experience (74.6%) constituted the majority of the sample (Table 1). On average, GPs had fewer patients represented in the database than did cardiologists and internists. Of the 10 883 patients, about half were men and about half were older than 75 years (Table 2). Approximately 20% of the patients were receiving a medication for each of the three selected comorbidities. Shared treatment by a generalist and a cardiologist or internist was observed for 70.9% of the patients. The three groups (controls and two intervention groups) were comparable in this respect.

View this table:
Table 2

Characteristics of the 10 883 patients in the study sample database, according to physician study group

CharacteristicControl (n = 3569)Guideline (n = 3487)Guideline + recall (n = 3827)χ2 (p value)
Sex
Male1880 (52.7)1731 (49.6)1954 (51.1)6.51 (0.039)
Female1689 (47.3)1756 (50.4)1873 (48.9)
Age
65–69 years834 (23.4)782 (22.4)815 (21.3)6.95 (0.325)
70–74 years989 (27.7)936 (26.8)1072 (28.0)
≥ 75 years1746 (48.9)1769 (50.7)1940 (50.7)
Location of residence
Urban1927 (54.0)2096 (60.1)2356 (61.6)53.61 (0.000)
Rural351 (9.8)322 (9.2)364 (9.5)
Suburban1291 (36.2)1069 (30.7)1107 (28.9)
Medication for
COPD697 (19.5)730 (20.9)774 (20.2)2.16 (0.340)
Heart failure586 (16.4)592 (17.0)687 (18.0)3.15 (0.208)
Diabetes650 (18.2)634 (18.2)728 (19.0)3.86 (0.425)
Co-treatment by a generalist and cardiologist or internist2527 (70.8)2438 (69.9)2749 (71.8)4.45 (0.108)
No. of physicians who prescribed CAD treatment drugs in previous year
11659 (46.5)1618 (46.4)1725 (45.1)4.25 (0.374)
21381 (38.7)1392 (39.9)1534 (40.1)
≥ 3529 (14.8)477 (13.7)568 (14.8)
Mean ± SD no. of visits to principal prescriber4.4 ± 4.44.2 ± 4.04.0 ± 4.6F 7.35 (0.001)a
Principal site of visits to principal prescriber*
Acute care services798 (22.6)832 (24.3)1045 (27.9)35.07 (0.000)
Private office2503 (71.0)2376 (69.4)2475 (66.0)
Other224 (6.4)216 (6.3)231 (6.2)
  • Data are numbers (%), except where stated. COPD, chronic obstructive pulmonary disease; CAD, coronary artery disease.

  • *Figures do not sum to total n of each group, because of lack of a principal site of visits for some patients.

  • aThe test used here is an ANOVA rather than χ2.

The distribution of the prescriptions across the three categories of anti-ischaemic drugs were the following: 2162 (20%) patients were on β-blockers only; 1364 (12.5%) were on β-blockers and nitrates; 1615 (13.9%) were on β-blockers and calcium-channels blockers; and 1760 (16%) were on triple therapy. A total of 4088 (37.6%) of patients had no β-blockers in their anti-angina regimen (either calcium-channels blockers or nitrates or a combination of the two).

Impact of intervention and predictors of prescribing profiles: results of multilevel regression analysis

At the patient level, patients receiving medications for COPD and heart failure were less likely to receive prescriptions for β-blockers and, to a lesser extent, for antiplatelet agents and hypolipaemics (Table 3). However, diabetic patients were more likely to receive the latter two classes of medication. Co-treatment by a generalist and a cardiologist or an internist increased the likelihood of taking a hypolipaemic drug. Patients who received prescriptions for anti-angina medication from more than one physician—but not necessarily the combination of a GP and a cardiologist or internist—were more likely to have prescriptions for a β-blocker and for an antiplatelet drug.

View this table:
Table 3

Summary of multilevel logistic regression models predicting prescription of selected cardiovascular medications in 1999

Class of drug …β-BlockersAntiplateletHypolipaemics
Patient characteristics
Sex
Male1.15 (1.06,1.25)*1.43 (1.32,1.55)*1.12 (1.03,1.22)*
Female (reference category)1.001.001.00
Age
65–69 years (reference category)1.001.001.00
70–74 years0.98 (0.87,1.11)1.01(0.90,1.14)0.92 (0.82,1.03)
≥ 75 years0.79 (0.71,0.88)*0.94 (0.84,1.04)0.49 (0.44,0.54)*
Location of residence
Urban (reference category)1.001.001.00
Rural1.23 (1.01,1.46)*1.13 (0.96,1.32)1.10 (0.92,1.30)
Suburban0.99 (0.89,1.11)1.10 (0.99,1.22)0.93 (0.83,1.05)
Co-treatment by a generalist and cardiologist or internist
No1.001.001.00
Yes1.08 (0.98,1.19)0.94 (0.85,1.03)1.29 (1.17,1.42)*
Medication for COPD
No1.001.001.00
Yes0.36 (0.32,0.40)*0.93 (0.84,1.02)0.82 (0.74,0.92)*
Heart failure
No1.001.001.00
Yes0.61 (0.55,0.68)*0.73 (0.66,0.81)*0.72 (0.64,0.81)*
Diabetes
No1.001.001.00
Yes1.00 (0.90,1.11)1.28 (1.15,1.42)*1.22 (1.10,1.36)*
No. of physicians who prescribed CAD treatment drug in previous year
10.70 (0.64,0.77)*0.75 (0.69,0.82)*1.06 (0.97,1.17)
2 (reference category)1.001.001.00
≥ 31.58 (1.38,1.81)*1.51 (1.33,1.72)*1.13 (0.99,1.28)
Physician characteristics
Group assignment
Control (reference category)1.001.001.00
Guideline1.00 (0.88,1.13)1.05 (0.94,1.18)1.02 (0.90,1.16)
Guideline + recall1.04 (0.92,1.18)1.07 (0.95,1.20)0.95 (0.83,1.08)
Sex
Female (reference category)1.001.001.00
Male0.93 (0.81,1.07)1.00 (0.87,1.14)0.84 (0.72,0.97)*
Medical training
General practitioner0.68 (0.58,0.81)*1.20 (1.03,1.40)*0.86 (0.73,1.02)
Cardiologist (reference category)1.001.001.00
Internist0.94 (0.72,1.23)1.45 (1.12,1.87)*1.10 (0.84,1.45)
Professional experience
0–10 years0.83 (0.62,1.10)1.10 (0.83,1.45)0.92 (0.68,1.24)
11–20 years (reference category)1.001.001.00
> 20 years1.01 (0.89,1.13)0.87 (0.77,0.97)*0.69 (0.61,0.78)
Number of patients by physician
1–50.89 (0.75,1.05)1.01 (0.86,1.19)1.03 (0.86,1.22)
6–10 (reference category)1.001.001.00
11–250.99 (0.85,1.15)1.05 (0.91,1.21)1.18 (1.01,1.38)*
> 250.98 (0.81,1.19)1.25 (1.05,1.49)*1.21 (0.99,1.47)
Intra-class correlation0.0690.0380.074
  • Data are odd ratios (95%CI) for receiving a prescription for the class of drug. COPD, chronic obstructive pulmonary disease; CAD, coronary artery disease. *Statistically significant at p ≤ 0.05.

At the physician level, the intervention (group assignment) had no impact on prescribing rates (Table 3). GPs were less likely to prescribe β-blockers than cardiologists but more likely to prescribe antiplatelet drugs. Doctors with > 20 years of experience tended to prescribe antiplatelet agents significantly less often, whereas high-volume physicians (those with more than 25 patients in the database) had a higher probability of prescribing these drugs. The percentage of variation in prescribing profile attributable to physician characteristics, as estimated by the intra-class correlation coefficients, was 6.9% for β-blockers, 3.8% for antiplatelet agents, and 7.4% for hypolipaemics.

Comparison between 1997 and 1999 profiles

Although the intervention had no impact on prescribing patterns, all prescribing profiles improved from 19977 to 1999 (Figure 2a). We observed an overall increase of 10% in the prescribing rates for antiplatelet agents and β-blockers from 1997 to 1999, and a smaller overall increase in the prescribing rates for hypolipaemic drugs. However, for hypolipaemic drugs, these increases were not distributed equally among patient age groups: greater increases were seen for patients aged ≥ 70 years (Figure 2b). We used logistic regression to explore the predictors of improvement toward the targeted recommendations (Table 4). Being a cardiologist was associated with a greater likelihood of prescribing each of the three classes of medication. Men were more likely to prescribe antiplatelet agents, and physicians with less experience (11–20 years) were less likely to prescribe hypolipaemic drugs.

Figure 2

Prescription of drugs of interest in 1997 study and 1999 (this study), presented as mean percentage of patients in each physician's practice who were receiving the drug.

View this table:
Table 4

Summary of logistic regression models predicting increase in prescriptions of selected cardiovascular medications between 1997 and 1999

Physician characteristicsβ-BlockersAntiplateletsHypolipaemics
Sex
Male0.95 (0.73,1.24)1.3 (1.01,1.73)*0.90 (0.69,1.18)
Female (reference category)1.001.001.00
Professional experience
< 10 years0.56 (0.30,1.07)0.66 (0.34,1.27)1.15 (0.62,2.14)
10–20 years0.86 (0.62,1.18)1.19 (0.87,1.64)0.89 (0.65,1.22)
21–30 years1.08 (0.81,1.45)1.36 (1.02,1.82)*1.12 (0.84,1.49)
> 30 years (reference category)1.001.001.00
Medical training
General practitioner1.46 (0.87,2.46)1.23 (0.73,2.06)1.12 (0.66,1.88)
Cardiologist4.2 (2.28,7.62)*1.9 (1.05,3.39)*1.6 (1.16,2.19)*
Internist (reference category)1.001.001.00
Number of patients in the database
Treated as a continuous variable: contribution of each additional patient1.04 (0.96,3.29)1.08 (0.87,2.56)1.12 (0.79, 2.34)
  • Data are odds ratios (95%CI) for an increase in prescription rate, compared with reduction or no change in prescription rate. *Statistically significant at p ≤ 0.05.

Discussion

Our results augment current knowledge in two respects. First, disseminating a simplified version of current guidelines for treating stable angina had little overall effect on prescribing practices. Secondly, certain physician characteristics were associated with more appropriate prescribing of some drug classes.

In previous experimental studies of continuing medical education (CME) methods, mass dissemination of information in traditional printed format did not significantly alter patterns of practice,3,12,,13 and our observations are consistent with those findings. However, we should be prudent in extrapolating these results to all methods of mass dissemination of information. Diffusion of information is the first step in knowledge transfer,14 and must occur as part of any global strategy for change. Time series and before-and-after studies with controls have suggested that mass dissemination of information alone may contribute to the desired evolution in practices at the health system level.3,5,6,,15 We now need to determine whether we can identify particular situations and specific guideline characteristics that will encourage appropriate change.

The internal characteristics of guidelines, such as their complexity and acceptability, have been shown to influence uptake,4,,6 and our intervention was developed to address perceived shortcomings in practice guidelines for stable angina (complexity, unappealing format, and inaccessibility). Physicians’ agreement with guideline content is also important.15,,16 For example, the Health Service Utilization and Research Commission of Saskatchewan found that three of eight guidelines disseminated to physicians with no additional intervention had an impact on services, four had no significant impact, and one had an impact opposite to that intended.5 The guidelines that had an impact were perceived as conveying an innovation, responded to a perceived need for information, and were not perceived as inconsistent with current trends in treatment. In our negative study, we may have presented physicians with a message that was not perceived as innovative and that did not respond to perceived needs.

With regard to physicians’ uptake of practice guidelines on the pharmacological treatment of stable angina, our results are reassuring in that they show improvements in the prescription rates of β-blockers, antiplatelet agents, and hypolipaemic drugs in patients aged ≥ 65 years, relative to results obtained 2 years earlier for the same physician sample.7 The small and shrinking variation attributable to physician characteristics, as indicated by the intra-class coefficient of the multilevel regressions, confirms a certain standardization of practice (from 0.079 in the 1997 study to 0.069 in 1999 for β-blockers; from 0.044 to 0.038 for antiplatelet agents; and from 0.095 to 0.074 for hypolipaemic drugs). In the current study, patients’ clinical characteristics had more effect on prescribing practices than physicians’ characteristics, which suggests that clinical reasons are the principal determinants of prescribing decisions, as should be the case. Still, certain groups, specifically women and patients aged over 70, were less likely to receive treatment in accordance with the guidelines,8 as was the situation in our 1997 study,7 and in other studies of patients with coronary artery disease.17–,19 We cannot conclude on the appropriateness of treatment on a patient basis, since we do not have the necessary individual clinical information to do so. Nonetheless, the generally favourable evolution of prescribing practices is encouraging, and is in agreement with the overall messages of the current guidelines.

As in 1997,7 patients being treated by GPs were less likely to receive β-blockers. However, we did not identify any consistent association between optimal treatment profiles and the number of angina patients seen by a physician, the physician's specialty, or co-treatment by a GP and a cardiologist or internist. In contrast, studies from the US have identified better practice profiles and better outcomes for patients with acute coronary syndromes who were being treated by a cardiologist, or by a primary care physician and a cardiologist, relative to patients treated by a primary care physician alone.20,,21 Our results do suggest that cardiologists were more likely to have improved their prescribing practices with time. There might be several reasons for this. Cardiologists have more rapid patient turnover, tending to see more new patients in a given period than GPs do, and it is easier to initiate a new regimen in a new patient than to modify the existing regimen of a known patient. Furthermore, it has been suggested that family physicians are slower or more prudent than specialists in modifying their practices, particularly if it means changing a regimen that is already known to work for a given patient.22 We could not assess the contribution of this phenomenon, since we did not analyse differences in practice for new and previously known patients. It is also possible that cardiologists are more comfortable prescribing β-blockers to patients with relative contra-indications, such as older age, diabetes, or cardiac failure, contraindications that are now being challenged on the basis of current research.23 Other reasons might be a greater propensity among cardiologists to change their prescribing practices, or greater exposure of cardiologists to new practice guidelines. Finally, it is likely that cardiologists see patients with more severe degree of coronary diseases, which can influence their prescription profiles and the way they adapt to new knowledge. Although we were able to control for confounding attributable to some co-morbidities, namely diabetes, COPD and heart failure, we were not in a position to assess coronary disease severity.

Our study had several strengths. First, we had access to all information related to physician visits and prescriptions for the entire physician sample and their patients 65 years of age and older. We used a multilevel approach, the best method to take into account the hierarchical structure of the data. We cannot attribute the absence of a difference between intervention and control groups to a lack of power: because of the large sample size, this study had a power of 80% to detect a 7% difference between the study groups at the 5% level.

The study also had some limitations. We are not comfortable generalizing the results to patient populations aged < 65 years for two major reasons. First, younger patients are under a variety of reimbursement schemes, with different co-payment modalities, which could have influenced physicians’ prescription choices in a direction difficult to predict since, in Canada, there are conflicting results of the impact of co-payment on patients’ access to prescription drugs. Second, age can influence physicians’ choices in many ways: the secondary effect profiles of some of the drugs studied may vary with age, particularly for β-blockers (erective dysfunction in younger men and cardiac dysfunction in older subjects, to name only a few); and the aggressiveness of cholesterol control may also vary with patient’ age, as is suggested by our own results.

Also, our sample was not representative of the Quebec physician population as a whole, but represented three specific regions: urban, suburban, and rural. Physicians in the rural region performed consistently better for prescription of β-blockers than physicians in the other two regions (Table 3), which may be a consequence of a practice organization that fosters better integration of GPs and specialists, and a more cohesive milieu for CME and the diffusion of practice guidelines. It would have been interesting to directly link physicians’ practice profiles and their attitudes toward the guidelines, but this could not be done because of confidentiality considerations. However, we have obtained the opinions of a sub-sample of the physicians in this study, which suggested that some still have reservations about using β-blockers because of concerns about side-effects and precipitating cardiac failure (M-D Beaulieu, A Jacques, RN Battista, R Blais, F Goulet, F Pasin, G Bretron, R Lebeau, D Landry, unpublished). Contamination might have occurred between the study groups, either directly (physicians in the intervention groups sharing information with physicians in the control groups) or indirectly (uptake of the guideline messages through the communication channels of various stakeholders and CME activities). Such contamination is indicated by our survey of a sub-sample of the physicians.24 In this study, 90% of respondents, including physicians in the control group, were aware of the guidelines, and 75% had participated in at least one CME activity on the topic during the previous 6 months.

In conclusion, these results suggest that in-house production and dissemination of a clinical practice guideline may not be a sound investment for a health authority if there are pre-existing indications of substantial scientific consensus on the issue. It may be more appropriate to work with professional associations and to support local CME-based interventions. In the specific case of stable angina, messages about the guidelines seem to be disseminated through natural channels of knowledge transfer.

Acknowledgments

We thank Ms Peggy Robinson for her help in the preparation of the manuscript. Dr Brophy receives financial support from le Fonds de Recherche en Santé du Québec (FRSQ). This project was funded by the Health Transition Fund, Health Canada. The results do not necessarily reflect the opinions of Health Canada.

M-DB received financial support from Aventis Pharma in 2000 to attend conferences to present preliminary results of an RCT to evaluate the effectiveness of a workshop to modify physicians’ performances of periodic health examinations in adults. She also received a research grant from this company in 1998 to complete that study, which was also funded by the Medical Research Council of Canada.

References

View Abstract