All measures and estimates were calculated for de facto children, defined as the members of the household 0–59 mo of age who slept in the household the previous night ( 17). We applied the 2006 WHO standard ( 14) macro to the raw DHS data files to derive estimates for HAZ, WHZ, prevalence of stunting, and prevalence of wasting among children 0–59 mo of age. The absolute difference in HAZ by MOB of reporting should be close to 0 if there are no systematic errors in age reporting. The degree of bias in MOB reporting for each survey was assessed by the absolute difference between the mean HAZ values for January and December ( 12). Therefore, surveys with larger differences in HAZ by MOB may indicate errors in age reporting, and thus biased HAZ estimates. For example, a child who is erroneously recorded as being born earlier in the year than the true birth month is actually younger than reported and therefore likely to be assigned an inappropriately low HAZ for age. This is because children who are born early in the year are likely to be randomly assigned later birth months, and vice versa for children born later in the year. In situations with poor date of birth information, however, discontinuities in estimates of mean HAZ by MOB are more likely to occur near the end/beginning of a calendar year.

The indices enable a relative assessment of the robustness of data underlying the estimation of metrics of HAZ (e.g., mean HAZ or stunting prevalence) or WHZ (e.g., mean WHZ or wasting or overweight prevalence) across multiple surveys.Īlthough there may be potential seasonal patterns in the relationship between mean HAZ and MOB, there should be no sharp differences in mean HAZ by the MOB within a given birth year. In this study, we describe the development of composite indices of anthropometric data quality for use in multisurvey analysis of child health and nutritional status. Whereas examining several individual indicators is informative for assessing various dimensions of quality within a single survey, for multisurvey analyses, a single aggregate measure of relative anthropometric data quality, which combines several data quality indicators, would better enable researchers to account for heterogeneity in the quality of anthropometric data collected across countries and over time. Several indicators have been used to assess anthropometric data quality including the pattern of age heaping ( 10), missingness of data on child height ( 11), proportion of biologically implausible values ( 6), misreporting of month of birth (MOB) for age estimation ( 12), and effect of random error ( 7). In the context of multisurvey analyses, accounting for variability in anthropometric data quality is particularly important because the quality of anthropometric data is unlikely to be uniform across surveys ( 6).Īnthropometric data quality may be affected by survey design (e.g., sampling strategy, questionnaire design, and measurement tools), implementation (e.g., nonresponse rate, management of field operations, staff training in data collection and anthropometry measurement, and method of data entry), and data processing procedures ( 6, 8, 9). However, the validity and reliability of survey-based metrics of child nutritional status depend on the quality of the anthropometric data ( 6– 8). These surveys are used to estimate and compare the nutritional status of young children within and between countries, to monitor secular trends, and to measure responses to public health interventions ( 4, 5). The Demographic and Health Surveys (DHS) Program conducts population-representative surveys in LMICs, including anthropometric data for children 0–59 mo of age in addition to other measures of health and development ( 3). Tracking progress toward global goals and identifying high-priority areas for investments are based on country-level prevalence estimates and trends in child malnutrition in low- and middle-income countries (LMICs) ( 2). Reductions in the global burden of malnutrition are central to the Sustainable Development Goals, which include specific targets related to stunting in children <5 y of age ( 1).

Ultraball x jeff chen