Purpose:
WHO guidelines for trachoma control include mass treatment distribution of Azithromycin in districts where the prevalence of active trachoma is more than 10% in children ages less than 10 years of age. Cluster random sample (CRS) methodology is recommended to identify districts in need of treatment. We examined the efficiency of CRS in estimating the district level prevalence when the number of clusters selected is limited by the overall cost of the survey.
Methods:
The Kongwa District of Tanzania was used because data are available on population size, and prevalence of active trachoma for all 64 villages (clusters) in the district. One hundred thousand samples of 8, 9, and 10 clusters, were drawn, selecting the clusters with probability proportional to the size. The estimated district prevalence for each sample was compared with the actual prevalence. Two sorting orders were used for the population listing: by geographical proximity, and at random. We report the proportion of samples that produced an estimate for the district outside ± 4%, and ± 5% of the actual prevalence
Results:
The district total population was 266,850 with an overall trachoma prevalence of 15.5%; village prevalences ranged from 0% to 54%. Simulation results are shown below.
Samples selected from geographically sorted listings performed better than samples selected from listings sorted at random.Conclusions Sample selection for diseases that cluster by geographical area should use population listings sorted by geographical proximity to increase the heterogeneity of the selected sample. Trachoma CRS surveys with low proportion of the District clusters selected may lead to erroneous estimates and in consequence erroneous decisions regarding mass treatment
Keywords: trachoma • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology • clinical (human) or epidemiologic studies: systems/equipment/techniques