Qualitative research, observational research, and The BMJ

Many of the reasons that qualitative research is given low priority at The BMJ also apply to observational studies.1 Their results are often not definitive or “likely to change clinical practice and help doctors make better decisions.” Their design means causality cannot be inferred, and confounding is invariably present and difficult to control for. Analyses are often exploratory, and large numbers of associations are evaluated, generating false positive results. The large number of participants and events produce very precise risk estimates that usually fall within the range considered weak in observational research (relative risks and odds ratios between 0.25 and 4.0).2 3 Despite generating fragile, low quality evidence, authors regularly make clinical practice recommendations based on their observational research.4

A recent cohort study exploring pre-pregnancy potato consumption and risk of gestational diabetes risk highlights these problems.5 The analysis was hypothesis generating; causality could not be inferred; the validity of self reported diet was uncertain; hundreds of secondary hypotheses had already been examined and thousands of statistical tests undertaken in this cohort, but false positive findings were not considered; and all risk estimates were between 1.0 and 2.0. Fundamental limitations were not discussed prominently in the paper or press release, as commonly occurs with observational research.6 The paper has no implications for clinical practice and it concluded that intervention trials are needed.5

The BMJ’s editors suggest that most qualitative research without definitive findings or clinical implications would be better published in specialist journals. Why then are observational studies with similar limitations to qualitative research published in The BMJ and not in specialist journals?