Klaus Munkholm1, Maria Faurholt-Jepsen2, John P A Ioannidis3, Lars G Hemkens4. 1. Nordic Cochrane Centre, Rigshospitalet, Copenhagen, Denmark. Electronic address: km@cochrane.dk. 2. Psychiatric Center Copenhagen, Copenhagen, Denmark. 3. Departments of Medicine, Health Research and Policy, Biomedical Data Sciences, and Statistics and the Meta-Research Innovation Center at Stanford (METRICS), Stanford University, Stanford, CA, USA. 4. Department of Clinical Research, Basel Institute for Clinical Epidemiology and Biostatistics, University of Basel, Basel, Switzerland.
Abstract
OBJECTIVES: When reporting observational studies, authors should explicitly discuss the potential for confounding and other biases, but it is unclear to what extent this is carried out within the psychiatric field. STUDY DESIGN AND SETTING: We reviewed a random sample of 120 articles in the five psychiatric specialty journals with the highest 5-year impact factor in 2015-2018. We evaluated how confounding and bias was considered in the reporting of the discussion and abstract and assessed the relationship with yearly citations. RESULTS: The term "confounding" was explicitly mentioned in the abstract or discussion in 66 articles (55.0%; 95% confidence interval (CI): 46.1-63.6) and the term "bias" in 68 articles (56.7%; 95% CI: 47.7-65.2). The authors of 25 articles (20.8%; 95% CI: 14.5-28.9) acknowledged unadjusted confounders. With one exception (0.8%, 95% CI: 0.0-4.6), authors never expressed any caution, limitation, or uncertainty in relation to confounding or other bias in their conclusions or in the abstract. Articles acknowledging nonadjusted confounders were not less frequently cited than articles that did not (median 7.9 vs. 5.6 citations per year, P = 0.03). CONCLUSION: Confounding is overall inadequately addressed in the reporting and bias is often ignored in the interpretation of high-impact observational research in psychiatry.
OBJECTIVES: When reporting observational studies, authors should explicitly discuss the potential for confounding and other biases, but it is unclear to what extent this is carried out within the psychiatric field. STUDY DESIGN AND SETTING: We reviewed a random sample of 120 articles in the five psychiatric specialty journals with the highest 5-year impact factor in 2015-2018. We evaluated how confounding and bias was considered in the reporting of the discussion and abstract and assessed the relationship with yearly citations. RESULTS: The term "confounding" was explicitly mentioned in the abstract or discussion in 66 articles (55.0%; 95% confidence interval (CI): 46.1-63.6) and the term "bias" in 68 articles (56.7%; 95% CI: 47.7-65.2). The authors of 25 articles (20.8%; 95% CI: 14.5-28.9) acknowledged unadjusted confounders. With one exception (0.8%, 95% CI: 0.0-4.6), authors never expressed any caution, limitation, or uncertainty in relation to confounding or other bias in their conclusions or in the abstract. Articles acknowledging nonadjusted confounders were not less frequently cited than articles that did not (median 7.9 vs. 5.6 citations per year, P = 0.03). CONCLUSION: Confounding is overall inadequately addressed in the reporting and bias is often ignored in the interpretation of high-impact observational research in psychiatry.