The Influence of Display and Statistical Factors on the Interpretation of Metaanalysis Results by Physicians
Dados Bibliográficos
| ID | 9104415 |
|---|---|
| Autores | Parminder Raina (0000-0002-8107-3193, University of British Columbia, autor correspondente), Parminder S Raina, Jamie C Brehaut (University of British Columbia), Robert W Platt (0000-0002-5981-8443, GlaxoSmithKline (Canada)), Terry P Klassen (0000-0002-5309-7091, University of Alberta), David Moher (0000-0003-2434-4206, University of Ottawa, autor correspondente), Phil St John (University of British Columbia), Dianne Bryant (0000-0002-7535-8707, University of Ottawa, autor correspondente), Ray Viola (University of Ottawa), Ba’ Pham (0000-0003-1738-3485, GlaxoSmithKline (Canada)) |
| Ano | 2005 |
| Volume | 43 |
| Fascículo | 12 |
| Páginas | 1242-1249 |
| Data de publicação | 2005-12-01 |
| Peer Reviewed | Sim |
| Open Access | Não |
| Tipo | ARTICLE |
| Periódico | Medical Care (JOURNAL) |
| Identificadores do periódico | ISSN: 0025-7079 • E-ISSN: 1537-1948 |
| Editora | Ovid Technologies (Wolters Kluwer Health) (PUBLISHER) |
| DOI | 10.1097/01.mlr.0000185710.62071.7f |
| PMID | 16299436 |
| OpenAlex | W2078049749 |
| Idioma | EN |
| Referências citadas | 15 |
OBJECTIVE: The objective of this study was to determine the extent to which various factors affect the interpretation of metaanalytic results by physicians. STUDY DESIGN: A sample of 120 physicians, selected from The Royal College of Physicians and Surgeons of Canada (RCPSC), was randomly assigned to 1 of 6 groups (n = 20) created from a combination of 3 summary measures and 2 levels of disease severity. The intervention consisted of a written scenario and 4 individual displays of metaanalyses (M-A), each followed by questions related to the interpretation of results of M-A. Two final questions examined statistical familiarity/proficiency with the summary measures used. DATA ANALYSIS: Analyses of variance examined main effects and interactions among 4 factors: summary measure, disease severity, effect size, and statistical consistency of the studies comprising the metaanalysis. Two outcomes were examined: interpretation of the treatment effect and confidence in the interpretation of the treatment effect. PRINCIPAL FINDINGS: Physicians were more likely to favor treatment when the results of the primary randomized, controlled trials (RCTs) were statistically homogeneous (P = 0.001) and when the overall effect size was large (P = 0.001). Also, physicians were more likely to be confident when the results were homogeneous (P = 0.001) and when effect size was large (P = 0.000). Interactions also revealed that the effect of statistical consistency of contributing to RCTs was greatest when data were presented as risk difference for treatment outcome (P = 0.026) and when effect size was small (P = 0.000). CONCLUSIONS: The interpretation of metaanalytic displays is influenced by the overall effect size of M-A, the statistical consistency of the contributing RCTs, and interactions of these factors with display factors
Analysis of variance · Clinical trial · Confidence interval · Consistency (knowledge bases) · Homogeneous · Interpretation (philosophy) · Meta-analysis · Publication bias · Randomized controlled trial · Sample size determination · Statistical significance · Statistics · Traditional medicine · Treatment effect · Internal Medicine · Mathematics · Medicine · Meta-analysis and systematic reviews · Reliability and Agreement in Measurement · Statistical Methods in Clinical Trials
| Velocidade de citação | historical |
|---|---|
| Altamente citado | Não |