Causal and Associational Language in Observational Health Research: A Systematic Evaluation

Abstract

We estimated the degree to which language used in the high-profile medical/public health/epidemiology literature implied causality using language linking exposures to outcomes and action recommendations; examined disconnects between language and recommendations; identified the most common linking phrases; and estimated how strongly linking phrases imply causality. We searched for and screened 1,170 articles from 18 high-profile journals (65 per journal) published from 2010-2019. Based on written framing and systematic guidance, 3 reviewers rated the degree of causality implied in abstracts and full text for exposure/outcome linking language and action recommendations. Reviewers rated the causal implication of exposure/outcome linking language as none (no causal implication) in 13.8%, weak in 34.2%, moderate in 33.2%, and strong in 18.7% of abstracts. The implied causality of action recommendations was higher than the implied causality of linking sentences for 44.5% or commensurate for 40.3% of articles. The most common linking word in abstracts was "associate" (45.7%). Reviewers' ratings of linking word roots were highly heterogeneous; over half of reviewers rated "association" as having at least some causal implication. This research undercuts the assumption that avoiding "causal" words leads to clarity of interpretation in medical research.

Keywords: association; causal inference; causal language; observational study.

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram detailing the search and screening process to arrive at our final sample for a study of causal and associational linking language in observational research and health evaluation literature. Journals included were American Journal of Epidemiology, American Journal of Medicine, American Journal of Preventive Medicine, American Journal of Public Health, Annals of Internal Medicine, BioMed Central Medicine, British Medical Journal, Canadian Medical Association Journal, European Journal of Epidemiology, International Journal of Epidemiology, Journal of Internal Medicine, Journal of the American Medical Association, Journal of the American Medical Association Internal Medicine, The Lancet, Mayo Clinic Proceedings, New England Journal of Medicine, PLOS Medicine, and Social Science and Medicine. JCR, Journal Citation Reports.

Figure 2

Summary scores for the degree of causal implication in linking sentences and action recommendations in a study of causal and associational linking language in observational research and health evaluation literature, showing the frequency of key strength of causal implication metrics for the 1,170 non–randomized-control trial studies in our sample, as indicated by the arbitrating reviewer. A–C) The strength of causal implication ratings for the language ratings in the abstract, discussion, and popout sections; D–F) the strength of causal implication ratings for the action recommendations in the abstract, discussion, and popout sections.

Figure 3

Comparison of the strength of causal implications in the abstracts for the linking phrase and action recommendations in a study of causal and associational linking language in observational research and health evaluation literature, showing the distribution of linking sentence and action recommendation language among the 400/1,170 non–randomized-control trial studies in which there was an action recommendation present in the abstract. A) An unconditional heatmap, with colors representing the number of articles in the strata, and histograms on the top and right showing the overall distribution of ratings for each axis; B) the distributions within each level of linking sentence causal strength.

Figure 4

Strength of causal implication ratings for the most common root linking words in a study of causal and associational linking language in observational research and health evaluation literature, showing the distribution of ratings given by reviewers during the root word rating exercise. On the left side, they are sorted by median rating + the percentage of reviewers who would have to change their ratings in order for the rating to change (A). On the right, the chart is sorted alphabetically (B).

Figure 5

Frequency of indicators of potential causal interest in a study of causal and associational linking language in observational research and health evaluation literature. A) Formal causal model; B) discussion of causal disclaimer statement; C) confounders mentioned; D) causal theory strength; E) control/adjustment variables. These results are from the 390 articles reviewed in full.