Background: The Portuguese National Health Directorate has issued clinical practice guidelines on prescription of anti-inflammatory drugs and COX-2 inhibitors, acid suppressive therapy and proton pump inhibitors, and anti-platelets. However, their effectiveness in changing actual practice is unknown. The objectives will be to compare the effectiveness of educational outreach visits in the implementation of clinical guidelines in primary care in Portugal against usual implementation strategies and to conduct a cost-effectiveness analysis of this method.

Methods: The trial will be a parallel, cluster-randomized, unblinded, trial in primary care, with a 1:1 allocation ratio. This study will assess the effect of educational outreach visits on physician compliance with prescription guidelines. The general study hypothesis is whether educational outreach visits are superior to usual implementation of guidelines regarding the reduction of inappropriate prescribing (compliance with prescription guidelines). All National Health Service primary care units in the Lisbon (Portugal) region will be invited to participate. Units will be eligible if they are using an Electronic Health Record to issue prescriptions and have at least four doctors willing to participate. Doctors in intervention units will receive three educational outreach visits (one for each guideline) during a six months period, while the control group doctors will be offered an unrelated group training session (on using the international classification for primary care). Intervention visits will be one on one 15 minutes discussions conducted by guideline authors or trained family physicians at the physician's workplace. There are two primary outcomes, measured at the physician's level. One is the proportion of COX-2 inhibitors prescribed within the entire NSAID class, in defined daily doses 18 months after the intervention. The other is the proportion of omeprazole within the entire proton pump inhibitors class, in defined daily doses at 18 months post-intervention. Prescription data will be collected from the regional pharmacy claims database.

The trial will be a parallel, cluster-randomized, unblinded, trial in primary care, with a 1:1 allocation ratio. This study will assess the effect of educational outreach visits on physician compliance with prescription guidelines. The general study hypothesis is whether educational outreach visits are superior to usual implementation of guidelines regarding the reduction of inappropriate prescribing (compliance with prescription guidelines). All National Health Service primary care units in the Lisbon (Portugal) region will be invited to participate. Units will be eligible if they are using an Electronic Health Record to issue prescriptions and have at least four doctors willing to participate. Doctors in intervention units will receive three educational outreach visits (one for each guideline) during a six months period, while the control group doctors will be offered an unrelated group training session (on using the international classification for primary care). Intervention visits will be one on one 15 minutes discussions conducted by guideline authors or trained family physicians at the physician's workplace. There are two primary outcomes, measured at the physician's level. One is the proportion of COX-2 inhibitors prescribed within the entire NSAID class, in defined daily doses 18 months after the intervention. The other is the proportion of omeprazole within the entire proton pump inhibitors class, in defined daily doses at 18 months post-intervention.Besides the above primary endpoints, the secondary outcomes will be the short (1 month) and medium-term (6 months) effects of educational outreach visits in the prescription of these two drug classes. Other secondary endpoints will be the short, medium and long-term effects of educational outreach visits in the prescription of clopidogrel. Finally, the trial will determine the cost effectiveness of educational outreach visits.

Allocation to intervention and blinding:

Clusters will be enrolled and allocated sequentially. To achieve a good balance regarding baseline characteristics, the allocation sequence will be determined by minimization (Ivers NI 2012). The units will be minimized by number of doctors (4 or 5, 6 or 7, 8 to12), median baseline prescription of COX-2 inhibitors (above or below the regional median), median baseline prescription of omeprazole (above or below the regional median), proportion of doctors with less than 10 years of practice after completing vocational training (above or below 50%), and type of primary care unit (family health unit or personalized care unit). All doctors sending the consent statement before the cluster allocation will be included in the study.

The sequence of intervention visits for each unit will be determined by simple randomization.

Allocation concealment will be ensured by the following procedures: the trial manager (hired and not part of the authors research team) will assign a sequential number to each unit as participation forms are received; only anonymized data about participating units will be sent to the trial statistician (sequential number and minimization variables); data will be sent in two batches, one for each half of all units; the statistician will allocate units to each trial arm using minimization; the sequence of visits will be determined using the random sequence generator from Random.org (http://www.random.org/sequences/); the statistician will return allocation information to the trial manager.

Due to the nature of the intervention, neither family physicians, nor detailers can be blinded. Outcomes are routinely collected by the regional health administration independently of the researchers or the trial and will only be sent to researchers after the intervention has ended. Data analysts (the trial statistician plus one of the members of the research team) will be blinded to group and visit sequence allocation until all analysis are completed.

Intervention and comparison:

Doctors in units randomized to the intervention will have three educational outreach visits during a six month period. These visits will promote the implementation of governmental guidelines on the prescription of the following agents: non-steroidal anti-inflammatory drugs (NSAIDs) and COX-2 inhibitors, acid secretion modifiers, and antiplatelets. The outcomes for the trial were chosen according to the main key-message from each guideline. For NSAIDs, COX-2 inhibitors should be prescribed only in patients with increased gastrointestinal risk who do not tolerate a classical NSAID associated with a gastroprotective agent; for acid secretion modifiers, all proton pump inhibitors are equivalent in effectiveness so omeprazole should be used in most patients as it is the least expensive; for antiplatelets, there is no benefit of maintaining long term clopidogrel after a myocardial infarction, acute coronary syndrome or percutaneous coronary intervention.

Intervention group: During each 15 to 20 minutes visit an academic detailer will promote one of the guidelines to a family doctor (up to three physicians may be present in each visit if they wish to, but one to one visits will be preferred and encouraged). The detailer will also distribute a point of care summary highlighting the main messages.

Control group: Usual guideline implementation consists of passive dissemination by their publication on the National Health Directorate's website. Doctors in units randomized to the control group will be offered an unrelated training session (coding with the International Classification of Primary Care, second edition) as a token of good will for participating in the trial.

Cost analysis:

Global prescription spending will be defined as the sum of the cost of all drug prescriptions of NSAIDs (ATC M01A), acid suppressive therapy (proton pump inhibitors ATC A02BC and their alternatives: H2-receptor antagonists ATC A02BA, antacids ATC A02A, misoprostol ATC A02BB01, and sucralfate ATC A02BX02), and antiplatelets (ATC B01AC) , up to 18 months after the intervention. These costs will be compared with the amount spent training the detailers, preparing and printing educational materials, travel expenses to intervention units, payment of detailers, program coordination, and physician time spent with a detailer rather than with a patient. Costs will be analyzed from the point of view of an educational outreach program rather than from conducting research. Therefore research related costs (such as researcher time for data collection and analysis) will not be considered. Similarly, the unrelated training session offered to the control units will not be accounted for because it is only intended to improve recruitment and it would not be necessary for the implementation of an educational outreach program.

Data collection:

Researchers will have access to prescription data through a data monitoring system operated by the Regional Health Administration. Data will be collected and provided by employees from this Administration according to researcher defined specifications. Importantly, researchers will not be directly involved in data collection. This information arrives with a two month delay from the date the prescription is dispensed. The prescription information contained there can be either for acute conditions - single prescriptions with up to two packages to be dispensed within 30 days; or for chronic usage - three identical prescriptions (up to two packages each) to be dispensed within 6 months. Within the drug classes of this study, only NSAIDs cannot be prescribed for chronic usage. Adverse events cannot be collected in this study because only prescription data is available.

Detailers will record whether the planned visit was effectively accomplished, whether it had the planned duration, the number of physicians (including residents) present, whether the visit was made on patient visit time or off hours, the number of times the detailer had previously visited that physician, and feedback from the physician about the educational materials.

Sample size:

The research team has obtained pilot data from all physicians of three primary care units. This data was used to estimate within unit variability and the intra-cluster correlation coefficient (ICC). Data was also gathered for all the units in the Regional Health Administration to estimate the mean prescription and standard deviation for the primary outcomes. The mean proportion of omeprazole dispensed was 54.0% of all PPIs (standard deviation, SD, 10.1%) and the ICC was 0.027. The mean proportion of COX-2 inhibitors dispensed was 20.6% of all oral NSAIDs (SD 7.4%) and the ICC was 0.249. The Cochrane review on educational outreach visits found a median adjusted risk difference for improvement in compliance with desired practice of 5.6% (interquartile range 3.0% to 9.0%) in previous trials (O' Brien, 2008). Therefore, we chose to calculate our sample size assuming the intervention would lead to about 5% absolute difference between intervention and control units.

If one assumes a mean cluster size of 6 doctors per unit, a 1:1 allocation ratio of controls per intervention unit, an alpha type error of 0.025, and a dropout rate of about 15%, then a sample of 110 doctors in each group will allow for 80% power to demonstrate a 5% absolute increase in the proportion omeprazole and a 5% absolute decrease in COX-2 inhibitors. To recruit the necessary number of physicians about 38 primary care units will be required. STATA 12.0 (STATA Corp, TX, USA) and its sampsi and sampclus commands were used to calculate sample size.

Statistical methods:

Physicians will be analyzed according to their randomly allocated group regardless of adherence to the intervention (intention to treat analysis). If physicians transfer to another unit within the health region we will still be able to monitor their prescriptions. If the transfer is to a different health region (i.e. not the Lisbon one) we will contact the physician and ask for the missing prescription data. In both cases, prescription of clopidogrel will be adjusted to the new patient list. If a physician retires or we are unable to retrieve data from a physician who moved to a different region then we will use the last working month's prescription.

Both groups will be compared on primary outcomes using a generalized mixed-effects model. The ratio of COX-2 inhibitors to the entire NSAIDs class and the ratio of omeprazole to the entire proton pump inhibitors class and respective 95% confidence intervals will be calculated. Statistical significance will be assumed for a p-value of less than 0.025. STATA 12.0 (STATA Corp, TX, USA) will be used to conduct the analysis.

Inclusion Criteria:

- Physicians of primary care unit with at least four family physicians

Exclusion Criteria:

• Physicians who are planning to retire within two years
• Physicians without an assigned patient list
• Physicians with an incomplete patient list