Objective: To evaluate the READER model for critical reading by comparing it with a free appraisal, and to explore what factors influence different components of the model.
Design: A randomised controlled trial in which two groups of general practitioners assessed three papers from the general practice section of the BMJ.
Setting: Northern Ireland.
Subjects: 243 general practitioners.
Main outcome measures: Scores given using the READER model (Relevance, Education, Applicability, Discrimination, overall Evaluation) and scores given using a free appraisal for scientific quality and an overall total.
Results: The hierarchical order for the three papers was different for the two groups, according to the total scores. Participants using the READER method (intervention group) gave a significantly lower total score (P < or = 0.01) and a lower score for the scientific quality (P < or = 0.0001) for all three papers. Overall more than one in five (22%), and more men than women, read more than 5 articles a month (P < or = 0.05). Those who were trainers tended to read more articles (P < or = 0.05), and no trainers admitted to reading none. Overall, 58% (135/234) (68% (76/112) of the intervention group) believed that taking part in the exercise would encourage them to be more critical of published articles in the future (P < or = 0.01).
Conclusion: Participants using the READER model gave a consistently lower overall score and applied a more appropriate appraisal to the methodology of the studies. The method was both accurate and repeatable. No intrinsic factors influenced the scores, so the model is appropriate for use by all general practitioners regardless of their seniority, location, teaching or training experience, and the number of articles they read regularly.