Comparison of the costs and data outputs of conventional cluster sampling and lot quality assurance sampling (LQAS) for assessing the coverage of fortified foods in household surveys

Household surveys are essential for assessing the coverage of public health programmes, including large-scale food fortification (LSFF) programmes in developing countries. For decades, survey implementers have predominantly designed and implemented household-based surveys using conventional cluster sampling, but other sampling approaches, such as lot quality assurance sampling (LQAS), should be considered as an alternative. This study compares the costs and data outputs of conventional cluster sampling and LQAS when used to measure the household-level coverage of a hypothetical LSFF programme. Specifically, four survey scenarios were compared using hypothetical results: conventional cluster sampling to calculate the coverage of fortified foods at the national (scenario A) and regional (scenario B) levels, and LQAS to produce pass/fail results at the national (scenario C) and regional (scenario D) levels. For each scenario, sample sizes were calculated using a target coverage of 25%, 50%, and 75%, and used previous surveys to estimate survey budget costs, which consisted of the costs of administration, field workers, other personnel, materials, and laboratory testing. A national level LQAS survey (scenario C) had the lowest estimated costs (69,424 - 73,462 USD), followed by a national level conventional cluster sampling survey (scenario A) (82,620 - 90, 164 USD). There were higher overall costs and larger cost differences between sampling approaches for surveys designed to yield regional estimates. Here, costs for a conventional cluster sampling survey (scenario B; 212,210 - 251, 470 USD) are more than double those for a LQAS survey (scenario D) (113,060 - 129,540 USD). Sample size is the main driver of survey costs in all scenarios, while costs for field teams (salaries and transportation) and laboratory analyses of food samples vary depending on the scenario and coverage threshold; all other survey costs (e.g., ethical approaval, training & field testing) remain relatively stable across different scenarios and thresholds. While LQAS surveys can be implemented at a lower cost due to smaller sample size requirements, the cost savings are less than expected due to the more dispersed distribution of households. Furthermore, because LQAS are initially designed to yield only pass/fail classification rather than estimates of actual coverage, they may not provide the actionable insights required in routine programme monitoring. When selecting a survey sampling ap