Results with confidence intervals (CIs) crossing the null should not be interpreted as indicating no effect. We assessed how often clinicians make this error and explored predictors of correct interpretation.

We conducted a cross-sectional survey of internal medicine residents, fellows, and faculty at McMaster University (Canada) between January and March 2025. Eighteen clinical cases were presented, including 5 cases in which the point estimate indicated benefit or harm and the CI crossed the null, 5 cases in which the point estimate indicated benefit or harm and CIs excluded the null, and eight cases indicating little or no effect. Ten cases used the null and eight cases used a minimally important difference as the decision threshold. Respondents judged whether an effect was present. Co-primary outcomes were the proportions correctly identifying (1) a benefit when the point estimate indicated benefit, but the CI crossed the null, and (2) a harm when the point estimate indicated harm, but the CI crossed the null. We used logistic regression to explore associations between respondent characteristics and correct responses.

Among 109 respondents (65.3% trainees, 34.7% faculty), most lacked graduate-level research training (83.2%) and leadership or formal training in systematic reviews (76.2%). When the point estimate showed benefit or harm but the CI crossed the null, only 13% to 18% correctly recognized an effect, compared with 81% to 98% when the CI excluded the null. Similar patterns were observed when using the minimally important difference threshold. No characteristics predicted correct interpretations.

Clinicians commonly misinterpret a CI crossing the null as “no effect,” disregarding the point estimate and decision threshold. This pervasive error risks misleading clinical decisions and guideline recommendations, underscoring the need for education on proper interpretation of CIs.

When research studies report treatment effects, they commonly include a CI, which reflects the uncertainty associated with the result. Many clinicians assume that if a CI includes “no effect,” the treatment must not work. This is misguided. We surveyed internal medicine residents, fellows, and faculty physicians to see how they interpret treatment effects when CIs are wide or cross commonly used decision thresholds. We found that most clinicians incorrectly concluded there was “no effect” whenever the CI crossed the null, even when the best estimate of the effect suggested benefit or harm. Fewer than one in five clinicians correctly identified an effect in these situations. These findings show that clinicians frequently rely on statistical significance rather than the size and direction of the estimated effect when interpreting study results. This misunderstanding may lead to poor clinical decisions, inaccurate guideline recommendations, and misleading communication with patients. Education and clearer reporting are needed to help clinicians interpret CIs correctly and focus on what the results actually mean for decision-making.