Impact of Different Sampling Schemes for Decision Making in Soil-Transmitted Helminthiasis Control Programs

Abstract

Starting and stopping preventive chemotherapy (PC) for soil-transmitted helminthiasis is typically based on the prevalence of infection as measured by Kato-Katz (KK) fecal smears. Kato-Katz-based egg counts can vary highly over repeated stool samples and smears. Consequentially, the sensitivity of KK-based surveys depends on the number of stool samples per person and the number of smears per sample. Given finite resources, collecting multiple samples and/or smears means screening fewer individuals, thereby lowering the statistical precision of prevalence estimates. Using population-level data from various epidemiological settings, we assessed the performance of different sampling schemes executed within the confines of the same budget. We recommend the use of single-slide KK for determining prevalence of moderate-to-heavy intensity infection and policy decisions for starting and continuing PC; more sensitive sampling schemes may be required for policy decisions involving stopping PC. Our findings highlight that guidelines should include specific guidance on sampling schemes.

Keywords: Kato-Katz; fecal egg counts; moderate-to-heavy infection; prevalence; soil-transmitted helminths.

Figure 1.

Sensitivity and precision of different sampling schemes for detection of hookworm infection, when based on the same overall budget. Budget is expressed as the number of individuals that can be tested in the context of a 1×1 scheme (1 stool sample collected per person and 1 slide tested per sample). Numbers between brackets under the sampling scheme indicators (“1×1”, “1×2”, etc) represent the number of people who were tested in that scheme, given the budget. Boxes represent the median and interquartile range (25th and 75th percentiles) of the bootstrapped prevalences; whiskers cover the range of bootstrapped values up to a distance of 1.5 times the interquartile range from the outer hinges of the box. Estimated are based on 10 000 bootstraps. (A) Ugandan dataset [5]; (B) Indian dataset [6]; (C) Kenyan data (TUMIKIA) [8, 9]; (D) Tanzanian dataset (Starworms) [10–12].

Figure 2.

Sensitivity and uncertainty of different sampling schemes for detection of moderate-to-heavy intensity hookworm infection in the general population, based on the same overall budget. Budget is expressed as the number of individuals that can be tested in the context of a 1×1 scheme (1 stool sample collected per person and 1 slide tested per sample). Numbers between brackets under the sampling scheme indicators (“1×1”, “1×2”, etc) represent the number of people who were tested in that scheme, given the budget. Boxes represent the median and interquartile range (25th and 75th percentiles) of the bootstrapped prevalences; whiskers cover the range of bootstrapped values up to a distance of 1.5 times the interquartile range from the outer hinges of the box. Estimated are based on 10 000 bootstraps. (A) Ugandan dataset [5]; (B) Indian dataset [6]; (C) Kenyan data (TUMIKIA) [8, 9]; (D) Tanzanian dataset (Starworms) [10–12].