Article type
Year
Abstract
Introduction: It has been suggested that the presentation format of trial results, that is, the information 'frame', may influence perceptions about the worth of a treatment, and hence, impact upon clinical practices.
Objective: We undertook a systematic review of the published literature to examine the effects of information framing on the practices of physicians.
Methods: Potentially relevant articles were retrieved through electronic and bibliographic searches using specific framing terms and supplemented by hand-searching (known author names, specific journals).
Results: Of 3691 articles retrieved, abstract review by two authors selected 192 for detailed assessment. The four-author reading-group judged 37 papers relevant to this review; 26 contained background information, and eleven, the basis of the review, were reports of studies assessing the effect of framing on physicians' practices. All eleven reported intended, as opposed to actual, practice. Six papers investigated the effect of framing treatment effects in terms of relative or absolute risk reduction (RRR/ARR), or as the number needed to treat (NNT). Data were extracted from all six studies and could be combined for two studies in two instances. In both cases, the results were not pooled due to the significant heterogeneity. In 5 of the 6 ARR/RRR/NNT papers, results framed in RRR terms were significantly more likely than ARR/NNT to be rated effective or elicit a 'treat' response from physicians. Physician gender/experience did not modify the framing impact, but other possible effect modifiers were not explored. Only the sixth paper used a control group. Those authors found a minor effect of framing that was highly sensitive to the level of risk. The remaining five papers framed outcomes in gain (eg. increased survival) or loss (eg. increased mortality) terms. Pooling of results was not possible. Overall, 'gain frames' were viewed more positively by physicians than 'loss frames', but the framing effect was modified by the risk level, type of decision, subjects' pre-existing beliefs about treatments and their clinical experience. All of the papers had methodological flaws: only one used a control group; nine exhibited some degree of within subject comparison; only two selected subjects randomly; four failed to clarify whether any 'between groups' differences existed.
Discussion: Framing appears to influence physicians' opinions of treatments, but the existing literature is limited; it is confined to reported practice only; the studies have numerous methodological flaws; both the magnitude of the framing effect and the extent to which it is modified by other factors affecting decision-making are poorly quantified.
Objective: We undertook a systematic review of the published literature to examine the effects of information framing on the practices of physicians.
Methods: Potentially relevant articles were retrieved through electronic and bibliographic searches using specific framing terms and supplemented by hand-searching (known author names, specific journals).
Results: Of 3691 articles retrieved, abstract review by two authors selected 192 for detailed assessment. The four-author reading-group judged 37 papers relevant to this review; 26 contained background information, and eleven, the basis of the review, were reports of studies assessing the effect of framing on physicians' practices. All eleven reported intended, as opposed to actual, practice. Six papers investigated the effect of framing treatment effects in terms of relative or absolute risk reduction (RRR/ARR), or as the number needed to treat (NNT). Data were extracted from all six studies and could be combined for two studies in two instances. In both cases, the results were not pooled due to the significant heterogeneity. In 5 of the 6 ARR/RRR/NNT papers, results framed in RRR terms were significantly more likely than ARR/NNT to be rated effective or elicit a 'treat' response from physicians. Physician gender/experience did not modify the framing impact, but other possible effect modifiers were not explored. Only the sixth paper used a control group. Those authors found a minor effect of framing that was highly sensitive to the level of risk. The remaining five papers framed outcomes in gain (eg. increased survival) or loss (eg. increased mortality) terms. Pooling of results was not possible. Overall, 'gain frames' were viewed more positively by physicians than 'loss frames', but the framing effect was modified by the risk level, type of decision, subjects' pre-existing beliefs about treatments and their clinical experience. All of the papers had methodological flaws: only one used a control group; nine exhibited some degree of within subject comparison; only two selected subjects randomly; four failed to clarify whether any 'between groups' differences existed.
Discussion: Framing appears to influence physicians' opinions of treatments, but the existing literature is limited; it is confined to reported practice only; the studies have numerous methodological flaws; both the magnitude of the framing effect and the extent to which it is modified by other factors affecting decision-making are poorly quantified.