RESUMEN
BACKGROUND: An efficacy evaluation of the AutoAnthro system to measure child (0-59 months) anthropometry in the United States found 3D imaging performed as well as gold-standard manual measurements for biological plausibility and precision. OBJECTIVES: We conducted an effectiveness evaluation of the accuracy of the AutoAnthro system to measure 0- to 59-month-old children's anthropometry in population-based surveys and surveillance systems in households in Guatemala and Kenya and in hospitals in China. METHODS: The evaluation was done using health or nutrition surveillance system platforms among 600 children aged 0-59 months (Guatemala and Kenya) and 300 children aged 0-23 months (China). Field team anthropometrists and their assistants collected manual and scan anthropometric measurements, including length or height, midupper arm circumference (MUAC), and head circumference (HC; China only), from each child. An anthropometry expert and assistant later collected both manual and scan anthropometric measurements on the same child. The expert manual measurements were considered the standard compared to field team scans. RESULTS: Overall, in Guatemala, Kenya, and China, for interrater accuracy, the average biases for length or height were -0.3 cm, -1.9 cm, and -6.2 cm, respectively; for MUAC were 0.9 cm, 1.2 cm, and -0.8 cm, respectively; and for HC was 2.4 cm in China. The inter-technical errors of measurement (inter-TEMs) for length or height were 2.8 cm, 3.4 cm, 5.5 cm, respectively; for MUAC were 1.1 cm, 1.5 cm, and 1.0 cm, respectively; and for HC was 2.8 cm in China. For intrarater precision, the absolute mean difference and intra-TEM (interrater, intramethod TEM) were 0.1 cm for all countries for all manual measurements. For scans, overall, absolute mean differences for length or height were 0.4-0.6 cm; for MUAC were 0.1-0.1 cm; and for HC was 0.4 cm. For the intra-TEM, length or height was 0.5 cm in Guatemala and China and 0.7 cm in Kenya, and other measurements were ≤0.3 cm. CONCLUSIONS: Understanding the factors that cause the many poor scan results and how to correct them will be needed prior to using this instrument in routine, population-based survey and surveillance systems.