CONTEXT: Contamination occurs when participants in the control arm of a trial receive an intervention that was intended for the intervention arm. Contamination tends to bias estimates of effectiveness towards no effect, and to reduce a trial's power to detect significant differences in outcomes. This is often assumed to be a problem in trials of educational interventions because of the transferability of verbal and experiential approaches. Cluster randomisation and other trial designs and analyses commonly used to reduce contamination can themselves cause bias or reduce statistical power. METHODS: We conducted a Delphi exercise to elicit experts' opinions about factors more or less likely to lead to contamination, and to rank methods of avoiding contamination when planning trials of educational interventions. RESULTS: Trials in settings with social, geographical or professional overlaps of respondents were considered to be at highest risk of contamination. Interventions that were complex and aimed to change behaviour were thought less likely to contaminate trials than simple interventions or those aimed at increasing knowledge. DISCUSSION: Although the issue of contamination requires detailed consideration of the nature of the intervention and academic judgement, this study adds to the literature on factors affecting contamination in educational studies and its findings will assist researchers in deciding whether they need to choose a cluster randomised trial design. The classification of studies and their likelihood of contamination suggest that complex interventions, also known to have greatest impact on behavioural change, have advantages in minimising contamination. Further work is required to see whether the additional costs of clustered studies are offset by the greater costs of complex interventions, while the aim of the study remains central to any choice of trial design.
RCT Entities:
CONTEXT: Contamination occurs when participants in the control arm of a trial receive an intervention that was intended for the intervention arm. Contamination tends to bias estimates of effectiveness towards no effect, and to reduce a trial's power to detect significant differences in outcomes. This is often assumed to be a problem in trials of educational interventions because of the transferability of verbal and experiential approaches. Cluster randomisation and other trial designs and analyses commonly used to reduce contamination can themselves cause bias or reduce statistical power. METHODS: We conducted a Delphi exercise to elicit experts' opinions about factors more or less likely to lead to contamination, and to rank methods of avoiding contamination when planning trials of educational interventions. RESULTS: Trials in settings with social, geographical or professional overlaps of respondents were considered to be at highest risk of contamination. Interventions that were complex and aimed to change behaviour were thought less likely to contaminate trials than simple interventions or those aimed at increasing knowledge. DISCUSSION: Although the issue of contamination requires detailed consideration of the nature of the intervention and academic judgement, this study adds to the literature on factors affecting contamination in educational studies and its findings will assist researchers in deciding whether they need to choose a cluster randomised trial design. The classification of studies and their likelihood of contamination suggest that complex interventions, also known to have greatest impact on behavioural change, have advantages in minimising contamination. Further work is required to see whether the additional costs of clustered studies are offset by the greater costs of complex interventions, while the aim of the study remains central to any choice of trial design.
Authors: Sunil S Solomon; Gregory M Lucas; David D Celentano; Frangiscos Sifakis; Shruti H Mehta Journal: Am J Epidemiol Date: 2013-06-25 Impact factor: 4.897
Authors: Sarah Allen; Suzanne Held; Shauna Milne-Price; Alma McCormick; Du Feng; Jillian Inouye; Mark Schure; Dottie Castille; Rae B Howe; Mikayla Pitts; Shannen Keene; Lorenda Belone; Nina Wallerstein Journal: Am J Community Psychol Date: 2021-09-17
Authors: Matthias Hoben; Peter G Norton; Liane R Ginsburg; Ruth A Anderson; Greta G Cummings; Holly J Lanham; Janet E Squires; Deanne Taylor; Adrian S Wagg; Carole A Estabrooks Journal: Trials Date: 2017-01-10 Impact factor: 2.279
Authors: Melchor Sánchez-Mendiola; Luis F Kieffer-Escobar; Salvador Marín-Beltrán; Steven M Downing; Alan Schwartz Journal: BMC Med Educ Date: 2012-11-06 Impact factor: 2.463