aFor all outcomes (except fear avoidance beliefs), a larger score indicates greater agreement or likelihood in the clinicians’ intentions and beliefs in performing the particular behaviour (i.e. not referring for plain X-ray or advising patients to stay active)
bAdjusted effects from models fitted using generalised estimating equations with exchangeable correlation structure and robust variance estimation to allow for clustering within practices. Effect estimates are the difference in means, with the exception of the outcome ‘knowledge’, where the effect estimate is an odds ratio. Models adjusted for pre-specified confounding variables noted in Fig. 2 of the trial protocol .
cRD risk difference. RD calculated from marginal probabilities. Confidence intervals were calculated using a pairwise comparison of margins after fitting a GEE model using Stata, allowing for clustering of observations within practices
dThe Knowledge variable was coded as indicating inadequate (0) or adequate (1) knowledge about key messages of the guideline