Improving Lifejacket Wear Among Boaters on Lake Albert, Uganda

A cluster randomized trial will be conducted for six months. To avoid covariate imbalance at baseline, stratified permuted block randomization with a 1:1 allocation ratio will be done. At least 387 boaters across 7 clusters per arm will be interviewed post baseline to give a 90% power to detect the effect of the intervention if such effect exists. A mixed effects multi-level modeling at 5% alpha level will be done using logical model building procedures

Given that the intervention will be at the community level, a cluster randomized controlled trial is considered most appropriate for testing the hypothesis of this study. The intact groups/clusters of individuals (boaters) rather than individuals themselves, will be randomized to either intervention or control arm. The landing sites herein referred to as clusters, will be the randomization unit. This is because it is difficult to randomize individuals to behavioral intervention since they mix within the cluster. This study will be conducted using an etic epistemological approach guided by a realist ontology in order to objectively measure the effect of the intervention without the influence of the researchers. To reduce covariate imbalance and increase comparability at baseline, clusters will be randomized to either intervention or control arm using stratified permuted block randomization, with landing sites (clusters) as the randomization unit. The clusters will be stratified by estimated population size and baseline prevalence of lifejacket use as compared with the overall average prevalence. All individuals in the intervention cluster will receive the intervention. The main aim of the stratified randomization is to ensure that the clusters are as comparable as possible at baseline. The other factors that might be considered for stratification will be identified at baseline and, in the opinion of the investigator, might interfere with the objectives of the study. Because there are few clusters per arm (7), two strata will be formed. The stratification variable will be the baseline prevalence of lifejacket wear in the landing sites. The risk of contamination (one of the major threats to this study) will be reduced by ensuring that at least one cluster which is not part of the study acts as a buffer. From each stratum, one cluster will be randomly allocated to the intervention arm and the other to the control arm using stratified permuted block randomization with a 1:1 allocation ratio. The allocation will be generated using the 'sealed envelope TM' web-based application by an independent statistician who has no interest in the study. The block sizes will be masked/blinded to increase the randomness.

The study participants will not know if they are receiving the intended intervention or not, since the control arm will also be receiving some messages

The peers will be trained to train their colleagues (fellow boaters) on proper lifejacket use. A training manual has been developed for this.

The Marine Police will be encouraged to continue conducting their community policing on water safety and emphasize lifejacket wear.

Training on proper lifejacket wear will be given to peers who will be the change agents amongst their fellow boaters

The intention to treat analysis will be used to assess the change in prevalence of lifejacket wear among the boaters. Multilevel modeling (cluster level model and boater level model) with a focus on the individual as the unit of analysis while accounting for clustering will be conducted. The generalized estimating equation (GEE) will be used to account for within-cluster variations. The goodness of fit of the model will be tested using the Hosmer Lemeshow test. Comparison between the baseline and end-line will be done and the difference in differences (Diff in Diff) will be reported to show the change of effect in the intervention and control arm.

Inclusion Criteria: The landing sites (clusters) should not be flooded, should not be receiving any known educational/behavior related intervention except the routine marine police messages Exclusion Criteria: Landing sites that are inaccessible because of the bad terrain and poor security

Six months from the start of the study, and will be accessible to the team until completion of PhD graduation

Bennett S, Woods T, Liyanage WM, Smith DL. A simplified general method for cluster-sample surveys of health in developing countries. World Health Stat Q. 1991;44(3):98-106.

Michie S, van Stralen MM, West R. The behaviour change wheel: a new method for characterising and designing behaviour change interventions. Implement Sci. 2011 Apr 23;6:42. doi: 10.1186/1748-5908-6-42.