Introduction
Neglected tropical diseases (NTDs) cause substantial morbidity and mortality, and can result in great cognitive, social, emotional and economic harm.1–6 Therefore, accelerating progress towards their control and elimination could majorly contribute to reaching the Sustainable Development Goal to ‘ensure healthy lives and promote well-being for all.’5 7 For seven of the NTDs—lymphatic filariasis (LF), onchocerciasis (OV), schistosomiasis (SCH), three types of soil-transmitted helminths (STH) and blinding trachoma—delivering preventive chemotherapy (PC) through mass drug administration (MDA) is a central component to preventing morbidity and reducing transmission.6 8 9 Approximately one-fifth of the world’s population requires PC for at least one NTD10 in order to achieve either their control or elimination as public health problems in the next decade.6
MDA involves administering medication to at-risk populations, typically once or twice per year. Populations requiring MDA vary by disease and local epidemiology, ranging from pre-school and/or school-aged children to women of childbearing age to the full population. Treatment coverage thresholds set by WHO vary by disease: ≥65% of the total population for LF, ≥75% of school-age children for SCH and STH, ≥80% of the total population for trachoma, and≥65% of the total or ≥80% of the eligible population for OV.11 12 Regardless, achieving adequate treatment coverage is a prerequisite for meeting control and elimination goals in a timely, efficient and cost-effective manner, and can be part of the criteria to determine when it is appropriate to conduct surveys to show whether MDA can be stopped or its frequency reduced.13 14 Low treatment coverage wastes valuable resources if additional years of MDA are required, and potentially negates the progress towards achieving established NTD road map 2030 goals.6 15 16
Therefore, estimates of programme coverage need to be accurate enough to support programmatic decisions. These decisions include identifying low coverage districts for increased attention and targeted support, and determining when to conduct impact surveys to potentially stop or change MDA frequency. For any health intervention coverage data to be actionable, it needs to be valid and reliable, available in a timely fashion, and collected within real-world financial and logistical constraints—features which often trade off with one another.17
MDA treatment coverage, that is, the proportion of those who need treatment who received and actually swallowed a dose, can be measured using two different sources of data: data from routine programmatic data or from specially implemented, population-representative coverage evaluation surveys.12 Coverage calculated using routinely reported data uses administrative data on the number of persons treated, as recorded by drug distributors during MDAs for the numerator. For the denominator, existing population estimates, normally from national census or programme registers, are used. Use of administrative data enables coverage estimates to be made at granular—district and subdistrict—levels, for each MDA, every year. However, it is potentially susceptible to denominator errors in estimated population size, which can be exacerbated when only subpopulations are targeted for treatment.18–24 Furthermore, if drug distributors do not directly observe patients taking MDA doses, the reported coverage may overestimate the fraction of the population that actually receive and swallow preventive chemotherapy.25–28 There is often a lack of trust in the quality of reported coverage due to limited arithmetic skills among some drug distributors, errors in counting total numbers treated in collated paper-based reports and registers, possible incentives for intentionally inflating the data and difficulty capturing treatments occurring outside the national programme (eg, through the private sector or non-governmental organisations).18–22 24 29 Anecdotally, national NTD programme managers sometimes report challenges due to insufficient/inappropriate data collection tools (eg, there may be treatment registers but no printed summary forms for summarising data).
Alternatively, treatment coverage can be estimated by conducting community-based coverage evaluation surveys. Coverage estimates based on surveyed data are often believed to produce more valid results, and WHO currently recommends that surveys be used to periodically assess the quality of reported coverage.12 14 However, coverage surveys are logistically more complicated and require additional resources and more complex analysis than reported coverage, with estimates usually being available to programme managers less quickly. In addition, coverage evaluation surveys are often designed to be representative at the district level,12 which precludes estimates in small but programmatically important administrative units, such as subdistricts; surveyed coverage may also be susceptible to selection and information bias30 such as recall bias (although one study did find that recall was fairly accurate up to 1 year post-MDA).31 Lastly, there is concern that surveys might be missing the same populations as MDAs, including because people are not at home due to work obligations that can take them away for days or weeks at a time, or because populations to be treated are nomadic and vulnerable populations that may not have fixed or permanent abodes.
The objective of the analyses presented here was to evaluate concordance between reported and surveyed coverage of MDAs conducted in 15 countries between 2008 and 2017. We incorporated coverage estimates as calculated by countries because these are the estimates from which programme managers would make programmatic decisions. Specifically, we aimed to understand (1) how often coverage calculated using routinely reported data and survey data would lead programme managers to make the same programmatic decisions, (2) the magnitude and direction of the difference between these two estimates, and (3) whether there is meaningful variation by region, age group, or country.