A.P. Coelho Figueira Freire1, D. Fernández Hernando2, M. Elkins3
1Central Washington University, Health Sciences, Ellensburg, United States, 2Hospital Fundación Jiménez Díaz, Madrid, Spain, 3University of Sydney, Faculty of Medicine and Health, Sydney, Australia
Background: Many statistical authorities have recommended reporting the effect of an intervention with a 95% confidence interval (CI) instead of with a p-value.CIs show how precisely a study has estimated the magnitude of the intervention’s effect and can contribute to interpretation of whether an intervention is clinically worthwhile. The physiotherapy profession is making efforts to increase the use of CIs. In 2022, an initiative by the International Society of Physiotherapy Journal Editors encouraged routine use of CIs instead of p-values. Establishing a baseline prevalence prior to that initiative will help to later discern its effect and reinforces the need for transition by researchers to reporting CIs.
Purpose: To determine the prevalence of use of CIs in current reporting of between-group differences in randomised trials of physiotherapy interventions. To identify if CIs, when presented, are interpreted by authors.
Methods: We randomly selected 100 trials published in 2021 from the Physiotherapy Evidence Database (PEDro). We included only the primary report of each research trial, therefore pilot studies and secondary analyses were excluded. Eligibility was not restricted by language of publication or area of physiotherapy practice. Two assessors independently extracted information regarding sample size, country, funding and quality criteria according to PEDro scale. We determined whether the analysis of all, some or none of the between-group comparisons was reported using CIs. We also extracted whether the types of outcomes reported with CIs were continuous, dichotomous or both; and recorded whether the CI were presented numerically or graphically. We also recorded whether at least one CI was interpreted. Interpretation included any mention of the clinical implications of the CI limit(s), any reference to the null value being inside or outside the CI, and/or any reference to location of the CI relative to the smallest worthwhile effect. Data were summarised with descriptive statistics.
Results: Most (26%) trials were in the musculoskeletal subdiscipline. The mean PEDro score (0-10) was 5 (SD 1) and the mean sample size was 138 (SD 221). Most trials were conducted in Asia (44%) and received funding (45%). Regarding CI prevalence, 53% of trials did not present any between-group CIs, 20% presented CIs for some outcomes and 27% presented CIs for all outcomes. This represents a marginal improvement over 2016, when 42% of trials reported some or all outcomes with CIs. Outcomes that presented CIs were mainly continuous (77%) and reported numerically (96%). Among the 47 trials that presented some or all outcomes with a between-group CI, only 6 (13%) interpreted the CI.
Conclusions: Most randomised trials of physiotherapy interventions still fail to report CIs. Additionally, among authors that are reporting CIs, a small minority provided an interpretation or mention of the clinical implications of the CIs.
Implications: There is still a high prevalence of randomised trials that are failing to provide researchers and clinicians with a more informative and precise view of effects on physiotherapy interventions. Ultimately this can impact clinicians’ ability to make more-informed decisions in their clinical practice. Therefore, awareness of the value of CIs is still needed.
Funding acknowledgements: This work was unfunded.
Keywords:
Randomised trials
Confidence intervals
Physiotherapy interventions
Randomised trials
Confidence intervals
Physiotherapy interventions
Topics:
Research methodology, knowledge translation & implementation science
Professional practice: other
Research methodology, knowledge translation & implementation science
Professional practice: other
Did this work require ethics approval? No
Reason: Ethics approval was not required because the study examined previously published randomised trials only.
All authors, affiliations and abstracts have been published as submitted.
