School-Based versus Community-Based Sampling for Trachoma Surveillance

Abstract

Trachoma surveillance is typically performed via random sampling of endemic districts. This strategy minimizes bias and allows examination of preschool children, but is also expensive. Surveillance for some other neglected tropical diseases is carried out in schools, which is logistically easier. In the present study, the prevalence of trachomatous inflammation-follicular (TF) from a population-based sample of children from each of 70 communities in Ethiopia was compared with the corresponding school-based estimate, which was calculated for each community by performing examinations in all primary schools in the district. The overall prevalence of TF was 39.1% (95% confidence interval [CI]: 35.0-43.1%) among children aged 1-9 years in the community-based sample and 18.8% (95% CI: 15.9-21.7%) among children in grades 1-3 of the school-based sample. School-based estimates of TF explained 35% of the variation in the community-based prevalences (P < 0.001). When TF prevalence was used as a diagnostic test for detecting a community with > 5% prevalence of ocular chlamydia, the area under the receiver operating characteristic curve was 0.73 (95% CI: 0.60-0.85) for the school-based sample and 0.71 (0.58-0.83) for the community-based sample (P = 0.76). Thus, although school-based monitoring was necessarily biased relative to population-based monitoring of 1- to 9-year olds, the two methods provided a similar amount of information about the community burden of ocular chlamydia in this trachoma-hyperendemic setting. The generalizability of these findings to areas with less prevalent trachoma is unclear.

Figure 1.

Distribution of the difference between the community-based TF prevalence estimate and school-based TF prevalence estimate in 70 Ethiopian communities. The white histogram represents the difference of the community-based sample among 1- to 9-year olds and the school-based sample among grades 1–3. The dark gray histogram represents the difference of the community-based sample among 7- to 12-year olds and school-based sample among 7- to 12-year olds. Positive numbers indicate a higher community-based prevalence of trachomatous inflammation–follicular (TF).

Figure 2.

School-based vs. community-based prevalence of trachomatous inflammation–follicular (TF) in 70 Ethiopian communities. Each marker represents a single community. The size of the marker is a monotonic indicator of the number of school-based examinations performed from each community and the shading of the marker signifies the prevalence of ocular chlamydia from the community-based monitoring. The solid black line represents the mean predicted population-based prevalence over the range of school-based prevalences and the dotted lines the 95% confidence interval for the predicted means.

Figure 3.

Receiver operating characteristic (ROC) curves assessing the diagnostic accuracy of school-based and community-based estimates of TF prevalence. Each column assumes a different dichotomous reference standard to gauge accuracy, with the reference standard based on the estimated prevalence of ocular chlamydia infection as assessed in the population-based sample. Each plot shows an ROC curve for school-based (in black; calculated from grades 1–3) and community-based (in gray) trachoma surveillance in which the community prevalence of trachomatous inflammation–follicular (TF) is treated as the diagnostic test. Below each curve are listed the areas under the curve, the optimal cut-point to maximize diagnostic information, and the sensitivity and specificity at the cut-point.