ABSTRACT
Coverage surveys for mass drug administration (MDA) rely on respondent recall and often permit proxy responses, whereby another household member is allowed to respond on behalf of an absent individual. In this secondary analysis of coverage surveys in Malawi, Burkina Faso, and Uganda, we explore the characteristics of individuals who require proxy responses and quantify the association between proxy responses and reported drug coverage. The adjusted logistic regression model found that men 11-39 years and women 11-18 years who were eligible for MDA had greater odds of requiring a proxy response compared with ineligible men and women in the same age groups. A hierarchical multivariable analysis found that proxy responses had 1.70 times the odds of reporting ingestion of MDA drugs compared with first-person responses, controlling for age and sex (95% CI: 1.17, 2.46). This finding is surprising, given that individuals absent during a coverage survey may also have been absent during the MDA, and suggests that proxy responses may be leading to an inflation of survey estimates of drug coverage. This study highlights the possibility for recall bias in proxy responses to MDA coverage; however, excluding absent individuals from coverage surveys would introduce a new bias. Further research is necessary to determine the best method for obtaining information on drug coverage when individuals are absent.
Subject(s)
Anthelmintics/administration & dosage , Anti-Bacterial Agents/administration & dosage , Antiparasitic Agents/administration & dosage , Mass Drug Administration/statistics & numerical data , Proxy , Adolescent , Adult , Albendazole/administration & dosage , Azithromycin/administration & dosage , Burkina Faso , Child , Demography , Female , Humans , Ivermectin/administration & dosage , Logistic Models , Malawi , Male , Mass Drug Administration/trends , Mental Recall , Praziquantel/administration & dosage , Uganda , Young AdultABSTRACT
Coverage evaluation surveys (CESs) are an important complement to routinely reported drug coverage estimates following mass drug administration for neglected tropical diseases (NTDs). Although the WHO recommends the routine use of CESs, they are rarely implemented. Reasons for this low uptake are multifaceted; one is uncertainty on the best sampling method. We conducted a multicountry study to compare the statistical characteristics, cost, time, and complexity of three commonly used CES sampling methods: the Expanded Program on Immunization's (EPI's) 30 × 7 cluster survey, a stratified design with systematic sampling within strata to enable lot quality assurance sampling (S-LQAS) decision rules, and probability sampling with segmentation (PSS). The three CES methods were used in Burkina Faso, Honduras, Malawi, and Uganda, and results were compared across the country sites. All three CES methods were found to be feasible. The S-LQAS approach took the least amount of time to complete and, consequently, was the least expensive; however, all three methods cost less than $5,000 per district. The PSS design resulted in an unbiased, equal-probability sample of the target populations. By contrast, the EPI approach had inherent bias related to the selection of households. Because of modifications needed to maintain feasibility, the S-LQAS method also resulted in a non-probability sample with less precision than the other two methods. Given the comparable cost and time of the three sampling methods and the statistical advantages of the PSS method, the PSS method was deemed to be the best for CESs in NTD programs.
