The World Health Organization (WHO) recommends Transmission Assessment Surveys (TAS) to determine when an evaluation unit (EU) (a designated population survey area) has achieved elimination of transmission of the vector-borne macroparasitic disease Lymphatic Filariasis (LF). These determinations are based on combining data from multiple survey units within an EU; it is unclear how underlying cluster-level variation influences the outcome of the TAS at EU level. We simulate LF infection distribution in an EU and compare three methods for assessing whether LF elimination has occurred based on currently recommended decision thresholds and sampling methods. We simulate an EU divided into clusters of varying size and disease prevalence. We produce 1000 samples according to LF TAS examples and WHO guidelines and compare three decision-making approaches: lot quality assurance sampling (LQAS) (recommended by WHO), one-sided interval estimate (CI), and nth order statistic (MAX). Summary statistics demonstrating the "pass" rate for the EU under different disease transmission conditions are generated using a versatile SAS® macro. As the prevalence of LF decreases, the LQAS and CI approaches produce increased likelihood of a pass outcome for an EU while some cluster units may still have a high likelihood of transmission. The MAX provides an alternative that increases the likelihood of determining a pass only once the whole area has a low likelihood of transmission. LQAS and CI approaches designed to estimate the LF prevalence in the EU miss hotspots that will continue to transmit infection while the MAX approach focuses on identifying clusters with high risk of transmission. The current TAS methodology has a flaw that may result in false predictions of LF transmission interruption throughout an EU. Modifying the TAS methodology to address results from extreme clusters rather than being based on mean prevalence over an EU will result in greater success for global elimination of LF.
4 October 2011 in The American Journal on Addictions
The World Health Organization recommends mass drug administration (MDA) with azithromycin to eliminate trachoma as a public health problem. MDA decisions are based on prevalence estimates from two-stage cluster surveys. There is a need to mathematically evaluate current trachoma survey designs. Our study aimed to characterize the effects of the number of units sampled on the precision and cost of trachomatous inflammation-follicular (TF) estimates. A population of 30 districts was simulated to represent the breadth of possible TF distributions in Amhara, Ethiopia. Samples of varying numbers of clusters (14-34) and households (10-60) were selected. Sampling schemes were evaluated based on precision, proportion of incorrect and low MDA decisions made, and estimated cost. The number of clusters sampled had a greater impact on precision than the number of households. The most efficient scheme depended on the underlying TF prevalence in a district. For lower prevalence areas (< 10%) the most cost-efficient design (providing adequate precision while minimizing cost) was 20 clusters of 20-30 households. For higher prevalence areas (> 10%), the most efficient design was 15-20 clusters of 20-30 households. For longer-running programs, using context-specific survey designs would allow for practical precision while reducing survey costs. Sampling 15 clusters of 20-30 households in suspected moderate-to-high prevalence districts and 20 clusters of 20-30 households in districts suspected to be near the 5% threshold appears to be a balanced approach.
9 September 2022 in JMIR Public Health and Surveillance
Background: Surveillance data are essential public health resources for guiding policy and allocation of human and capital resources. These data often consist of large collections of information based on nonrandom sample designs. Population estimates based on such data may be impacted by the underlying sample distribution compared to the true population of interest. In this study, we simulate a population of interest and allow response rates to vary in nonrandom ways to illustrate and measure the effect this has on population-based estimates of an important public health policy outcome. Objective: The aim of this study was to illustrate the effect of nonrandom missingness on population-based survey sample estimation. Methods: We simulated a population of respondents answering a survey question about their satisfaction with their community’s policy regarding vaccination mandates for government personnel. We allowed response rates to differ between the generally satisfied and dissatisfied and considered the effect of common efforts to control for potential bias such as sampling weights, sample size inflation, and hypothesis tests for determining missingness at random. We compared these conditions via mean squared errors and sampling variability to characterize the bias in estimation arising under these different approaches. Results: Sample estimates present clear and quantifiable bias, even in the most favorable response profile. On a 5-point Likert scale, nonrandom missingness resulted in errors averaging to almost a full point away from the truth. Efforts to mitigate bias through sample size inflation and sampling weights have negligible effects on the overall results. Additionally, hypothesis testing for departures from random missingness rarely detect the nonrandom missingness across the widest range of response profiles considered. Conclusions: Our results suggest that assuming surveillance data are missing at random during analysis could provide estimates that are widely different from what we might see in the whole population. Policy decisions based on such potentially biased estimates could be devastating in terms of community disengagement and health disparities. Alternative approaches to analysis that move away from broad generalization of a mismeasured population at risk are necessary to identify the marginalized groups, where overall response may be very different from those observed in measured respondents.