TY - JOUR
T1 - Principles for designing randomized preventive trials in mental health
T2 - An emerging developmental epidemiology paradigm
AU - Brown, C. Hendricks
AU - Liao, Jason
N1 - Funding Information:
'This paper is an update of two papers presented at the National Conference on Prevention sponsored bytheNationalInstituteofMentalHealthinWashington,D.C.,in1991and1996. The authors wish to acknowledge the many contributions of all members of the Panel on Strategies for Effective Intervention, National Conference on Prevention Research. Discussions with the Chair, Kathryn Barnard, and Carolyn Webster-Stratton, Peter Steinglass, and David Cordray were especially helpful. We would especially like to thank the principal investigators and their colleagues of NIMH prevention centers at Arizona State (IrwinSan-dier), Johns Hopkins (Phil Leaf), Oregon Social Learning Center (John Reid), and the Michigan Prevention Research Center (Richard Price) as well as principal investigators and their colleagues in the conduct disorder research demonstration projects at Duke-Vanderbilt-Penn State-University of Washington Fast-Track Trial (John Coie), and the University of Illinois at Chicago (Nancy Guerra, Pat Tolan, and Rowell Huesmann) as well as Project ADEPT (Gregory Clarke) and the family depression prevention project at Harvard (William Beardslee) for providing detailed information for inclusion in this paper, even at the risk of potential criticism of the designs. David Hawkins also provided valuable comments, as did the AJCP reviewers. Special thanks are due to Sheppard Kellam who led the 1994 conference on prevention science at Hopkins which provided a structure for this cross-site paper and whose continued discussions led to broadening these design concerns. Discussions withSandy Braver, Eve Moscicski, Doreen Spilton-Koretz, and Pat Mrazek were especially helpful. Kathleen Cervenka provided editorial help, and Wei Wang contributed some of the graphics. Finally, we would like to thank our methodological colleagues at the Prevention Science and Methodology Group (PSMG) who have provided many insights. Among this group, Mike Stoolmiller at OSLC graciously provided data on school and classroom variation.A number of comments made by those just mentioned have found their way into this paper, we hope in ways that do justice to their original thoughts. At the same time, only the authors are responsible for the views presented here. Work on this paper was supported by joint funding from the NIMH Prevention Branch and the Epidemiology Branch under Grant No. MH40859, "Designs and Analyses for Mental Health Preventive Trials," the NIMH Prevention Branch under Grants No. MH01259, "Methodologic Advances in Mental Illness Prevention," No. MH38725, "Epidemiologic Center for Early Risk Behaviors," and No. MH42968, "Periodic Outcome of Two Preventive Trials."
PY - 1999/10
Y1 - 1999/10
N2 - An emerging population-based paradigm is now being used to guide the design of preventive trials used to test developmental models. We discuss elements of the designs of several ongoing randomized preventive trials involving reduction of risk for children of divorce, for children who exhibit behavioral or learning problems, and for children whose parents are being treated for depression. To test developmental models using this paradigm, we introduce three classes of design issues: design for prerandomization, design for intervention, and design for postintervention. For each of these areas, we present quantitative results from power calculations. Both scientific and cost implications of these power calculations are discussed in terms of variation among subjects on preintervention measures, unit of intervention, assignment, balancing, number of pretest and posttest measures, and the examination of moderation effects.
AB - An emerging population-based paradigm is now being used to guide the design of preventive trials used to test developmental models. We discuss elements of the designs of several ongoing randomized preventive trials involving reduction of risk for children of divorce, for children who exhibit behavioral or learning problems, and for children whose parents are being treated for depression. To test developmental models using this paradigm, we introduce three classes of design issues: design for prerandomization, design for intervention, and design for postintervention. For each of these areas, we present quantitative results from power calculations. Both scientific and cost implications of these power calculations are discussed in terms of variation among subjects on preintervention measures, unit of intervention, assignment, balancing, number of pretest and posttest measures, and the examination of moderation effects.
KW - Attrition
KW - Developmental epidemiologic preventive trials
KW - Doubly randomized designs
KW - Growth models
KW - Participation bias
KW - Prevention science
KW - Replication
KW - Selection bias
KW - Statistical interactions
KW - Statistical power
KW - Trial within a trial
KW - Unit of intervention
UR - http://www.scopus.com/inward/record.url?scp=0033202593&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0033202593&partnerID=8YFLogxK
U2 - 10.1023/A:1022142021441
DO - 10.1023/A:1022142021441
M3 - Article
C2 - 10676544
AN - SCOPUS:0033202593
SN - 0091-0562
VL - 27
SP - 673
EP - 710
JO - American Journal of Community Psychology
JF - American Journal of Community Psychology
IS - 5
ER -
