The choice of an appropriate unit of analysis in evaluations in health policy is problematic and errors are frequently made which may undermine study conclusions. This paper describes the structure of randomized trials, issues concerning randomization and replication, and the factors that should be considered when deciding upon the appropriateness of a unit of analysis.
Get full access to this article
View all access options for this article.
References
1.
SoumeraiSBAvornJ.Principles of education outreach (‘academic detailing’) to improve clinical decision making. Journal of the American Medical Association1990; 263: 549–556
2.
DivineGWBrownTFrazierLM.The unit of analysis error in studies about physicians' patient care behaviour. Journal of General Internal Medicine1992; 7: 623–629
3.
Whitting-O'KeefeQEHenkeCSimborgDW.Choosing the correct unit of analysis in medical care experiments. Medical Care1984; 22: 1101–1114
4.
The GUSTO Investigators. An international randomized trial comparing four thrombolytic strategies for acute myocardial infarction. New England Journal of Medicine1993; 329: 673–682
5.
CornfieldJ.Randomization by group: a formal analysis. American Journal of Epidemiology1978; 108: 100–102
6.
FranksPDicksonJC.Comparison of family physicians and internists: process and outcome in adult patients at a community hospital. Medical Care1986; 24: 941–948
7.
DonnerA.A regression approach to the analysis of data arising from cluster randomization. International Journal of Epidemiology1985; 14: 322–326
8.
DonnerABirkettNBuckC.Randomization by cluster: sample size requirements and analysis. American Journal of Epidemiology1981; 114: 906–914
9.
DonnerADonaldA.Analysis of data arising from a stratified design with the cluster as unit of randomization. Statistics in Medicine1987; 6: 43–52
10.
DonnerA.Statistical methodology for paired cluster designs. American Journal of Epidemiology1987; 126: 972–979
11.
ShipleyMJSmithPGDramaixM.Calculation of power for matched pair studies when randomization is by group. International Journal of Epidemiology1989; 18: 457–461
12.
HsiehFY.Sample size formulae for intervention studies with the cluster as unit of randomization. Statistics in Medicine1988; 8: 1195–1201
13.
GoldsteinH.Multilevel models in educational and social research. London: Oxford University Press, 1987
14.
RiceNLeylandA.Multilevel models: applications to health data. Journal of Health Services Research and Policy1996; 1: 154–164
15.
RiceNJonesA.Multilevel models and health economics. Health Economics1997; 6: 561–575
16.
DuffySWSouthMCDayNE.Cluster randomization in large public health trials: the importance of antecedence data. Statistics in Medicine1992; 11: 307–316
17.
LangfordIHBenthamG.Regional variations in mortality rates in England and Wales: an analysis using multi-level modeling. Social Science and Medicine1996; 42: 897–908
18.
FengZMcLerranDGrizzleJ.A comparison of statistical methods for clustered data analysis with Gaussian error. Statistics in Medicine1996; 15: 1973–1806
19.
DiwanVKErikssonBSterkyGTomsonG.Randomization by group in studying the effect of drug information in primary care. International Journal of Epidemiology1992; 21: 124–130
20.
SmithTCSpiegelhalterDJThomasA.Bayesian approaches to random-effects meta analysis: a comparative study. Statistics in Medicine1995; 14: 2685–2699
21.
DavisPGribbenB.Rational prescribing and interpractitioner variation: a multilevel approach. International Journal of Technology Assessment in Health Care1995; 11: 428–442
22.
Van de WerfFTopolEJLeeKLWoodliefLHGrangerCBArmstrongPWVariations in patient management and outcomes for acute myocardial infarction in the United States and other countries: results from the GUSTO trial. Journal of the American Medical Association1995; 273: 1586–1591
23.
FreemantleNHainesAMasonJMEcclesM.CONSORT — an important step towards evidence based health care. Annals of Internal Medicine1997; 126: 81–83
