An evaluation of harvest plots to display results of meta-analyses in overviews of reviews

Article type
Authors
Hartling L1, Wingert A2, Crick K3, Fernandes R4, Thomson D2, Williams K5
1Cochrane Child Health Field; Alberta Research Centre for Health Evidence, University of Alberta, Canada
2Cochrane Child Health Field, University of Alberta, Canada
3School of Public Health, University of Alberta, Canada
4Cochrane Child Health Field; Clinical Pharmacology Unit, Instituto de Medicina Molecular, Lisbon Academic Medical Centre & Portuguese Branch of the Iberoamerican Cochrane Centre, Portugal
5Cochrane Child Health Field; Department of Developmental Medicine; University of Melbourne; Murdoch Children's Research Institute; The Royal Children's Hospital Melbourne, Australia
Abstract
Background:
Harvest plots are used to display evidence from complex and diverse studies or results graphically. They group studies based on whether they demonstrate a positive, negative or no effect, and can display additional information, such as quality of evidence.

Objectives:
To: 1) determine the feasibility of using harvest plots to depict complex results in overviews of reviews; 2) survey end-users to determine their preferences for graphical display of data, and their understanding of data presented as harvest plots.

Methods:
We developed harvest plots for two overviews. We conducted a survey of 279 members of the Cochrane Child Health Field, which includes pediatric healthcare providers and researchers. We asked whether harvest plots, used in conjunction with tables typically presented in overviews of reviews, are superior to tables alone.

Results:
Fifty-three members responded. On a scale from 0 to 100 (where 100 was most favourable), average responses showed little difference between harvest plots and standard tables with respect to the following features: well suited to display results from meta-analysis (51.1 (SD 26.9) harvest plots, 55.8 (SD 24.6) tables; aesthetically pleasing (57.5 (SD 28.5) harvest plots, 44.5 (SD 25.0) tables); easy to understand (53.2 (SD 26.3) harvest plots, 50.4 (SD 26.1) tables); format intuitive (49.6 (SD 25.2) harvest plots, 44.7 (SD 24.2) tables). Twenty-two (48.9%) respondents indicated that harvest plots could be used in conjunction with tables to display results of meta-analyses; 20 (44.4%) respondents indicated they could be used with some improvements.

Conclusions:
Our goal was to explore a novel form of data presentation to promote uptake of evidence from overviews of reviews. Suitability and ease of understanding the harvest plots were similar to standard tables. Harvest plots were more aesthetically pleasing; however, standard tables were slightly better suited to display results. Neither presentation format was ranked highly on any variable. Further work is needed to identify and evaluate methods for presenting results from overviews to ensure they are readily understood by end-users.