Background: one of the most attractive alternatives to narrow the evidence-practice gap is the implementation of trustworthy clinical practice guidelines. Although the last decade has seen significant advances in guideline methodology, important limitations still remain. Furthermore, healthcare workers frequently consider that guideline recommendations are alien to their context. This situation reduces recommendation adherence. We hypothesized that including potential guideline users in the recommendation’s development process would increase compliance.

Objectives: to evaluate if a strategy that incorporates clinicians (recommendation users) in the process of recommendations development is feasible and improves recommendation adherence.

Methods (figure 1): the study was carried out in the internal medicine department of the German hospital in Buenos Aires, Argentina, between March and December 2018. Study participants (clinicians working in the department) identified 40 relevant clinical questions that were included in our study and randomized to intervention and control arms. We developed recommendations in response to the 20 questions assigned to the intervention arm following the GRADE approach. In 45-minutes meetings, the study participants, constructed recommendations in response to those questions using Evidence to Decision frameworks and summary of findings tables developed by two GRADE methodologists. To answer the questions assigned to the control group, we adopted published recommendations. All the recommendations (intervention and control) were included in an easily accessible webpage and considered as official guidance. We prospectively identified recommendation adherence opportunities (situations in which an opportunity to adhere to one of the 40 recommendations existed) and recorded whether the clinician’s course of action was consistent with the direction of the proposed recommendation (recommendation adherence). We calculated the relative risk and 95% confidence interval (CI) of recommendation adherence between intervention and control arms. In order to adjust for potential confounding, we constructed a logistic regression model considering several variables. Additionally, to account for the clustered nature of the data we performed a sensitivity analysis considering study questions as the units of analysis.

Results: during the study period, we identified 1004 recommendation adherence opportunities corresponding to the questions assigned to intervention and 1987 to control. Adherence to recommendations in response to questions assigned to the intervention arm was higher than those assigned to the control group, adjusted estimate, odds ratio (OR) 2.14 (95% CI 1.6 0to 2.83). Sensitivity analysis accounting for the clustered nature of the data informed a non-statistical difference between the study arms, P = 0.27.

Conclusions: including guideline users in the recommendation development process was feasible and may increase recommendation adherence.