| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
|
|
|
|||||||
|
|||||||||
TRACK to:
|
|
Electronic letters published:
![[Read Comment]](/icons/misc/sectionDOWN.gif){kind=icon}
More on Clusters
GRTs: a Timely and Useful Tool in Family Practice Research
|
|
|||
|
James J. Diamond, Philadelphia, USA research faculty Jefferson Medical College
Send response to journal:
|
May I add two additional comments to this interesting and important discussion on clustering? In an earlier comment to this paper, Richard Kryscio mentioned the David Murray and colleagues paper in the March 2004 AJPH. One aspect of this paper is its extensive set of references. Folks have been writing about nonindependence and clusters for over 20 years and they have the citations for anybody who wants to read from the "start to the current." One paper, among others I'm sure, that was concerned about nonindependence and clusters and dealt with it methodologically is in Pediatrics, February 1997: Horbar JD et al. Hospital and patient characteristics associated with variation in 28-day mortality rates for very low birth weight infants, Pediatrics, 1997, 99, 149-156. As they noted on page 150 "Statistical tests are adjusted for the fact that infants within each NICU represent a cluster of individuals who are not independent." Competing interests: None declared |
|||
|
|
|||
|
Richard J Kryscio, Lexington, Ky Biostatistician, University of Kentucky
Send response to journal:
|
Randomizing doctors, medical practices or even entire communities to interventions but taking observations on individual patients or families is not a new idea in public health research but it has received considerable attention in the past few years. Aside from references provided in the Zyzanski et al article, readers will find the recent review papers by Murray et al and by Donner and Klar interesting and informative. Articles in the current issue of the Annals emphasize common facts that every researcher must consider when conducting a group randomized trial (GRT). These center about accounting for the correlation among responses from the same cluster in both the design of the trial as illustrated by Killip et al as well as in the analysis of the data as illustrated by Reed. The key concept in both articles is the product of the intraclass correlation coefficient (ICC) and the average cluster size. The ICC measures the degree of correlation among responses in the same cluster. The product is a measure of the deviation the GRT has from the familiar simple randomized trial. Since this product is sensitive to the value of the ICC, a catalog of ICC values encountered in family practice studies would be helpful to researchers. Many interventions in the family practice setting such as exercise, weight loss, smoking cessation, and drug prevention programs benefit from and rely on the mutual support and camaraderie that can only be gained when subjects are treated in batches or groups. In vaccine trials GRTs yield not only an estimate of the direct effect of the vaccine (i.e. did it reduce disease incidence when compared to placebo ?) but also by grouping responses an estimate of its indirect effect (i.e. the effect a vaccination program has on a community where not all individuals volunteer to be immunized). These are not achievable with the simple randomized trial. Hence, the GRT has become an integral part of public health research. Family practice researchers are encouraged to become familiar with the rudimentary properties as well as the relative merits of these designs. References: Donner, A., Klar, N. Pitfalls and controversies in cluster randomized trials. American Journal of Public Health, 2004, 94(3), 416-422. Killip, S., Mahfoud, Z., Pearce, K. What is an intraclass correlation coefficient ? Crucial concepts for primary care researchers. Annals of Family Medicine, 2004, 2(3), 204-208. Murray, D.M., Varnell, S.P., and Blitstein, J.L. Design and analysis of group-randomized trials: a review of recent methodological developments. American Journal of Public Health, 2004, 94(3): 423-432. Reed, J.F.III Adjusted chi-square statistics: application to clustered binary data in primary care. Annals of Family Medicine, 2004, 2(3), 201-203 Zyzanski, S.J., Flocke, S.A., Dickinson, L.M. On the nature and analysis of clustered data. Editorial in Journal of Family Medicine, 2004, 2(3), 199-200. Competing interests: None declared |
|||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
TRACK: