Logistic regression in psychiatric literature: evaluation of articles published between 2002 and 2005 in a prominent journal

INTRODUCTION: Many articles in journals that specialize in psychiatry use logistic regression to identify the effect of independent variables on the probability of occurrence of an event. Assessments of articles published in other medical specialties have shown that the report of logistic regression is incomplete in many cases and this leads to difficulties in interpretation. OBJECTIVE: Assess the quality related to logistic regression analysis and accomplishment of internal validity criteria of articles published in a high impact journal specialized in psychiatry. METHODS: Two independent reviewers selected (manual search) articles that used logistic regression published in the Archives of General Psychiatry (2002-2005) and evaluated them with the internal validity criteria of the Journal of the American Medical Association (JAMA) and a tool designed to value the quality of logistic models (CML). RESULTS: Of 121 articles assessed, 85 (70.2%) met JAMA's criteria of internal validity. In relation to the CML tool, the most frequently reported criterion was coding of independent variables (90.9%), followed by the report of the selection process of independent variables included at the beginning of the study (87.6%) and by the inclusion of the RR or OR of the model and its respective CI (82.6%). The less frequently reported criterion was the goodness of fit (9.1%), followed by the report of the adjustment process of the model (24.8%). CONCLUSIONS: Although most articles reach high editorial standards, the reporting of logistic regression should be improved.

Logistic Regression; Psychiatry; Evaluation Studies; Bibliometrics; Evidence-based medicine

Associação Brasileira de Saúde Coletiva Av. Dr. Arnaldo, 715 - 2º andar - sl. 3 - Cerqueira César, 01246-904 São Paulo SP Brasil , Tel./FAX: +55 11 3085-5411 - Rio de Janeiro - RJ - Brazil
E-mail: revbrepi@usp.br