Open AccessRandomization in psychiatric intervention research in the general practice setting
Author(s) -
FeltzCornelis CM Van Der,
Adèr HJ
Publication year - 2000
Publication title -
international journal of methods in psychiatric research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.275
H-Index - 73
eISSN - 1557-0657
pISSN - 1049-8931
DOI - 10.1002/mpr.87
Subject(s) - blinding , randomization , intervention (counseling) , psychological intervention , research design , psychology , critical appraisal , clinical study design , randomized controlled trial , placebo , psychiatry , medicine , clinical trial , alternative medicine , mathematics , statistics , surgery , pathology
Abstract Most studies of psychiatric interventions in general practice settings conform only in part to the requirements of randomization, placebo control and blinding as formulated by the Cochrane Collaboration. It is possible, nonetheless, to develop experimental research designs that are sufficiently near to this standard. These must deal with certain methodological issues specific to psychiatric research. This article discusses scientific standards of psychiatric research with special consideration of interventions in general practice settings. These issues are accompanied by concrete examples and suggestions on how to confront the problems. In psychiatric intervention research, equivalence studies with single‐blind outcome assessment, a tested and ethically justified method, are generally used in place of placebo‐controlled studies. The article also examines randomization procedures in greater depth. Randomization can be applied across trial subjects or across doctors' practices. Practical consequences of randomizing across subjects, and specific implementations of it such as crossover and pre‐post designs in general practice settings, are clarified. Overall, a research design using randomization across doctors' practices is judged preferable to one that randomizes across trial subjects. One potential problem is that the control group may become too small, especially when considerable effects are expected from the intervention being studied. One might consider making the control condition smaller in the first place, or, if indicated on ethical grounds, performing an intermediate analysis and then breaking off the study as soon as a statistically significant effect has been demonstrated. Multilevel statistical techniques offer new opportunities for analysis within such designs. Copyright © 2000 Whurr Publishers Ltd.