Comparison of tests for categorical data from a stratified cluster randomized trial

Fourth-Grade Cooking and Physical Activity Intervention Reveals Associations With Cooking Experience and Sex

OBJECTIVE

Examine the impact of Fuel for Fun: Cooking with Kids Plus Parents and Play (FFF) on children's culinary self-efficacy, attitude, fruit and vegetable (FV) preferences, physical activity (PA), and body mass index.

DESIGN

Randomized controlled trial.

SETTING

Eight elementary schools in 2 Northern Colorado districts.

PARTICIPANTS

Fourth-grade students; 7-month interventions: school (S.FFF)-theory-based cooking + tasting lessons, active recess, lesson-driven cafeteria promotions; or school + family (S+F.FFF) with added family nights and home activities.

MAIN OUTCOME MEASURE(S)

Cooking self-efficacy and attitudes, FV preferences, PA, and measured height/weight.

ANALYSIS

Individual outcomes nested by classroom, school, and district and assessed > 12 months with repeated measures controlled by sex and baseline cooking experience, with a significance level of P < 0.05.

RESULTS

The sample included 1,428 youth, 38 teachers, 4 cohorts, 50% boys, 75% White, and 15% Hispanic. No intervention effect was observed. Those who cooked retained higher self-efficacy, attitude, and FV preferences (P < 0.001). Girls reported higher self-efficacy and attitude than boys. Moderate-to-vigorous PA and metabolic equivalent minutes increased for all students; boys retained higher levels (P < 0.001). Body mass index percentile remained stable.

CONCLUSIONS AND IMPLICATIONS

Cooking and sex were associated with all outcome measures and should be considered for intervention tailoring. Treatment impacts were not evident nesting by classroom, school, and district. Accurate assessment of school-based interventions requires rejecting student independence from group assignment assumptions.

Testing conditional independence in sets of I × J tables by means of moment and correlation score tests with application to HPV vaccine

A new testing approach is described for improving statistical tests of independence in sets of tables stratified on one or more relevant factors in case of categorical (nominal or ordinal) variables. Common tests of independence that exploit the ordinality of one of the variables use a restricted-alternative approach. A different, relaxed-null method is presented. Specifically, the M-moment score tests and the correlation tests are introduced. Using multinomial-Poisson homogeneous modeling theory, it is shown that these tests are computationally and conceptually simple, and simulation results suggest that they can perform better than other common tests of conditional independence. To illustrate, the proposed tests are used to better understand the human papillomavirus type-specific infection by exploring the intention to vaccinate. Copyright © 2016 John Wiley & Sons, Ltd.

Genetic diversity and investigation of polledness in divergent goat populations using 52 088 SNPs

The recent availability of a genome-wide SNP array for the goat genome dramatically increases the power to investigate aspects of genetic diversity and to conduct genome-wide association studies in this important domestic species. We collected and analysed genotypes from 52 088 SNPs in Boer, Cashmere and Rangeland goats that had both polled and horned individuals. Principal components analysis revealed a clear genetic division between animals for each population, and model-based clustering successfully detected evidence of admixture that matched aspects of their recorded history. For example, shared co-ancestry was detected, suggesting Boer goats have been introgressed into the Rangeland population. Further, allele frequency data successfully tracked the altered genetic profile that has taken place after 40 years of breeding Australian Cashmere goats using the Rangeland animals as the founding population. Genome-wide association mapping of the POLL locus revealed a strong signal on goat chromosome 1. The 769-kb critical interval contained the polled intersex syndrome locus, confirming the genetic basis in non-European animals is the same as identified previously in Saanen goats. Interestingly, analysis of the haplotypes carried by a small set of sex-reversed animals, known to be associated with polledness, revealed some animals carried the wild-type chromosome associated with the presence of horns. This suggests a more complex basis for the relationship between polledness and the intersex condition than initially thought while validating the application of the goat SNP50 BeadChip for fine-mapping traits in goat.

Clustering and meso-level variables in cross-sectional surveys: an example of food aid during the Bosnian crisis

BACKGROUND

Focus groups, rapid assessment procedures, key informant interviews and institutional reviews of local health services provide valuable insights on health service resources and performance. A long-standing challenge of health planning is to combine this sort of qualitative evidence in a unified analysis with quantitative evidence from household surveys. A particular challenge in this regard is to take account of the neighbourhood or clustering effects, recognising that these can be informative or incidental.

METHODS

An example of food aid and food sufficiency from the Bosnian emergency (1995-96) illustrates two Lamothe cluster-adjustments of the Mantel Haenszel (MH) procedure, one assuming a fixed odds ratio and the other allowing for informative clustering by not assuming a fixed odds ratio. We compared these with conventional generalised estimating equations and a generalised linear mixed (GLMM) model, using a Laplace adjustment.

RESULTS

The MH adjustment assuming incidental clustering generated a final model very similar to GEE. The adjustment that does not assume a fixed odds ratio produced a final multivariate model and effect sizes very similar to GLMM.

DISCUSSION

In medium or large data sets with stratified last stage random sampling, the cluster adjusted MH is substantially more conservative than the naïve MH computation. In the example of food aid in the Bosnian crisis, the cluster adjusted MH that does not assume a fixed odds ratio produced similar results to the GLMM, which identified informative clustering.

Impact evaluation of the Northern Fruit and Vegetable Pilot Programme – a cluster-randomised controlled trial

Objective

The purpose of this impact evaluation was to measure the influence of a government of Ontario, Canada health promotion initiative, the Northern Fruit and Vegetable Pilot Programme (NFVPP), on elementary school-aged children’s psychosocial variables regarding fruit and vegetables, and fruit and vegetable consumption patterns.

Design

A cluster-randomised controlled trial design was used. The NFVPP consisted of three intervention arms: (i) Intervention I: Free Fruit and Vegetable Snack (FFVS) + Enhanced Nutrition Education; (ii) Intervention II: FFVS-alone; and (iii) Control group. Using the Pro-Children Questionnaire, the primary outcome measure was children’s fruit and vegetable consumption, and the secondary outcome measures included differences in children’s awareness, knowledge, self-efficacy, preference, intention and willingness to increase fruit and vegetable consumption.

Setting/Subjects

Twenty-six elementary schools in a defined area of Northern Ontario were eligible to participate in the impact evaluation. A final sample size of 1277 students in grades five to eight was achieved.

Results

Intervention I students consumed more fruit and vegetables at school than their Control counterparts by 0·49 serving/d (P < 0·05). Similarly, Intervention II students consumed more fruit and vegetables at school than Control students by 0·42 serving/d, although this difference was not statistically significant. Among students in both intervention groups, preferences for certain fruit and vegetables shifted from ‘never tried it’ towards ‘like it’.

Conclusions

The NFVPP resulted in positive changes in elementary school-aged children’s fruit and vegetable consumption at school, and favourable preference changes for certain fruit and vegetables.

Intracluster correlation coefficient in multicenter childhood trauma studies

OBJECTIVE

To calculate the intracluster correlation coefficient (ICC) for emergency department (ED) shock rate, early trauma death (ie, death during the first 24 h after arrival at hospital), and in-hospital trauma death rate for multicenter childhood injuries.

METHODS

The National Trauma Data Bank (5th revision), the largest multicenter trauma registry in the US, was used. Data from 80 trauma centers were used to calculate the ICC for in-hospital trauma death rate. Thirty three states provided data for calculation of the ICC for ED shock and early trauma death rate.

RESULTS

From 2000 to 2004, 13% of the 952 242 patients in the National Trauma Data Bank were <15 years old. Approximately 17 000 of these children had injuries with an injury severity score >15, of whom 84% (14 095 subjects) were hospitalized at 80 level I or II trauma centers in 33 states. The ICCs for ED shock rate, early trauma death rate, and in-hospital death rate were 0.005 (95% CI 0.000 to 0.010), 0.014 (95% CI 0.004 to 0.024), and 0.023 (95% CI 0.013 to 0.033), respectively. These ICCs were calculated for boys and girls and also for blunt and penetrating injuries.

CONCLUSION

Clustered childhood trauma studies that aim to compare different aspects of pre-hospital and hospital trauma care should incorporate these ICCs for sample calculation. When cluster randomized clinical trials are mounted, if sample sizes are calculated without adjustment for ICC, then the planned trial is likely to be seriously underpowered.

Developments in cluster randomized trials and Statistics in Medicine

The design and analysis of cluster randomized trials has been a recurrent theme in Statistics in Medicine since the early volumes. In celebration of 25 years of Statistics in Medicine, this paper reviews recent developments, particularly those that featured in the journal. Issues in design such as sample size calculations, matched paired designs, cohort versus cross-sectional designs, and practical design problems are covered. Developments in analysis include modification of robust methods to cope with small numbers of clusters, generalized estimation equations, population averaged and cluster specific models. Finally, issues on presenting data, some other clustering issues and the general problem of evaluating complex interventions are briefly mentioned.

Analysis of clustered data in multicentre trauma studies

INTRODUCTION

In order to design multicentre studies an estimate of the correlation of the observations within each centre is necessary. A standard measure of the correlation between observations within each centre is the Intraclass Correlation Coefficient (ICC).

METHOD

We used the National Trauma Data Bank (NTDB). By 2004, 448 trauma centres (including 110 level I and 123 level II trauma centres) from 43 states and US territories contributed over 1.2 million records to the NTDB. Data of patients directly transported from the scene of injury to level I or II trauma centres were used to calculate the ICC of in-hospital trauma fatality and emergency department (ED) shock rate.

RESULTS

The ICCs of ED shock and in-hospital fatality rate were 0.010 (95% confidence interval (CI): 0.003-0.018) and 0.039 (95% CI: 0.028-0.050), respectively. The ICC of shock in the ED was the highest for penetrating injuries (0.017, 95% CI: 0.003-0.032) and the lowest for women (0.008, 95% CI: 0.002-0.013) although the observed difference between men and women was not statistically significant. The ICC of trauma fatality was the highest for penetrating injuries (0.073, 95% CI: 0.047-0.098), and the lowest for blunt injuries (0.029, 95% CI: 0.020-0.037).

DISCUSSION

Although the calculated ICCs might seem so small as to be ignored, the required sample size in studies with exclusively exposed or non-exposed clusters depends on the ICC and the average number of subjects within clusters. Therefore, investigators should be aware of the influence that these ICCs might have on sample size and power of their studies.