A comparison of imputation strategies in cluster randomized trials with missing binary outcomes

被引:12
作者
Caille, Agnes [1 ,2 ,3 ,4 ]
Leyrat, Clemence [1 ,2 ,3 ]
Giraudeau, Bruno [1 ,2 ,3 ,4 ]
机构
[1] INSERM, U1153, Paris, France
[2] INSERM, CIC 1415, 2 Bd Tonnelle, F-37044 Tours 9, France
[3] CHRU Tours, Tours, France
[4] Univ Francois Rabelais Tours, PRES Ctr, Val de Loire Univ, Tours, France
关键词
cluster randomized trial; missing data; outcome; multiple imputation; MULTIPLE-IMPUTATION; CLINICAL-TRIALS; CORRELATION-COEFFICIENT; INTRACLASS CORRELATION; TREAT; INTENTION; DESIGN; MODELS;
D O I
10.1177/0962280214530030
中图分类号
R19 [保健组织与事业(卫生事业管理)];
学科分类号
摘要
In cluster randomized trials, clusters of subjects are randomized rather than subjects themselves, and missing outcomes are a concern as in individual randomized trials. We assessed strategies for handling missing data when analysing cluster randomized trials with a binary outcome; strategies included complete case, adjusted complete case, and simple and multiple imputation approaches. We performed a simulation study to assess bias and coverage rate of the population-averaged intervention-effect estimate. Both multiple imputation with a random-effects logistic regression model or classical logistic regression provided unbiased estimates of the intervention effect. Both strategies also showed good coverage properties, even slightly better for multiple imputation with a random-effects logistic regression approach. Finally, this latter approach led to a slightly negatively biased intracluster correlation coefficient estimate but less than that with a classical logistic regression model strategy. We applied these strategies to a real trial randomizing households and comparing ivermectin and malathion to treat head lice.
引用
收藏
页码:2650 / 2669
页数:20
相关论文
共 47 条
[1]  
Altman Douglas G, 2009, Open Med, V3, pe51
[2]   A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome [J].
Ambler, Gareth ;
Omar, Rumana Z. ;
Royston, Patrick .
STATISTICAL METHODS IN MEDICAL RESEARCH, 2007, 16 (03) :277-298
[3]   A comparison of the statistical power of different methods for the analysis of cluster randomization trials with binary outcomes [J].
Austin, Peter C. .
STATISTICS IN MEDICINE, 2007, 26 (19) :3550-3565
[4]   Analysis of dichotomous outcome data for community intervention studies [J].
Bellamy, SL ;
Gibberd, R ;
Hancock, L ;
Howley, P ;
Kennedy, B ;
Klar, N ;
Lipsitz, S ;
Ryan, L .
STATISTICAL METHODS IN MEDICAL RESEARCH, 2000, 9 (02) :135-159
[5]   Robustness of a multivariate normal approximation for imputation of incomplete binary data [J].
Bernaards, Coen A. ;
Belin, Thomas R. ;
Schafer, Joseph L. .
STATISTICS IN MEDICINE, 2007, 26 (06) :1368-1382
[6]   The design of simulation studies in medical statistics [J].
Burton, Andrea ;
Altman, Douglas G. ;
Royston, Patrick ;
Holder, Roger L. .
STATISTICS IN MEDICINE, 2006, 25 (24) :4279-4292
[7]   Dichotomizing a continuous outcome in cluster randomized trials: impact on power [J].
Caille, Agnes ;
Leyrat, Clemence ;
Giraudeau, Bruno .
STATISTICS IN MEDICINE, 2012, 31 (24) :2822-2832
[8]   Determinants of the intracluster correlation coefficient in cluster randomized trials: the case of implementation research [J].
Campbell, MK ;
Fayers, PM ;
Grimshaw, JM .
CLINICAL TRIALS, 2005, 2 (02) :99-107
[9]   CONSORT statement: extension to cluster randomised trials [J].
Campbell, MK ;
Elbourne, DR ;
Altman, DG .
BMJ-BRITISH MEDICAL JOURNAL, 2004, 328 (7441) :702-708
[10]  
Carpenter JamesR., 2013, MULTIPLE IMPUTATION