Anthropometry in infants under 6 months in rural Kenya

Summary of research¹

Reliability study participants

It is currently estimated that worldwide 8.5 million infants under 6 months are wasted. In poor communities, low rates of exclusive breastfeeding and the introduction of mixed feeding before the age of 3 months expose infants <6 months to risks of microbial contamination and malnutrition. In Kenya, 9.7% infants below 6 months are wasted (weight-forlength z score (WFLz) < -2) and 11% are stunted (length-for-age (LFA) z score < -2).

The Government of Kenya has proposed a strategy in which Community Health Workers (CHWs) are trained to deliver community health services, including basic primary health care, growth monitoring (GM) and referral of critically ill patients to hospital. CHWs will be expected to undertake door-to-door anthropometric screening of children and provide basic nutrition education and counselling.

At rural health facilities, weight is commonly measured in infancy. However, weight-for-age (WFA) alone does not differentiate wasting from stunting and is typically accessible only to those attending mother and child health clinics (MCH). WFL is recommended for the diagnosis of acute malnutrition in this age group, but is rarely routinely assessed because length boards are not usually available and length measurement is potentially unreliable. Among children aged 6-59 months, the Mid Upper Arm Circumference (MUAC) may be used to diagnose severe acute malnutrition (SAM). MUAC is a better predictor of mortality than WFA and within this age range, is age-independent. In rural communities, MUAC could be a valuable tool for use by CHWs for early detection of acute malnutrition in infants. However, reliability of MUAC measurement in early infancy is unknown, and cut-off values to determine intervention thresholds have not been defined. To address the first of these questions, a recent study set out to determine the inter-observer variability and accuracy of MUAC and WFLz measurements taken by CHWs among infants under 6 months in rural Kenya.

The study was conducted from February 2008 through August 2009 in Kilifi District, a rural district on the Kenyan Coast. Kilifi is the second poorest district in Kenya with an estimated 67% of the people living in poverty. The study recruited three cadres of participant: (i) an expert in anthropometry with more than 30 years of experience of anthropometry training and conducting nutritional assessments in Kenya, (ii) health professionals (HPs) comprising nurses and public health officers in-charge of Mother and Child Health (MCH) clinics, and (iii) CHWs based at health centres throughout the district.

Study methods

The research employed a cross-sectional repeatability design using the term ‘reliability’ in a statistical sense to mean the inter-observer variability. For applicability within the health system, the research team used measuring equipment normally in use in government facilities in Kenya, with an exception of the infantometer, which is usually not available. Weight was measured to the nearest 100 g using a ‘hanging’ scale costing 103 USD. The machine was quality controlled every morning using standard weighing stones certified by the Kenya Bureau of Standards. Length was taken using a professional infantometer calibrated to the nearest 1 mm and costing 443 USD. MUAC was measured on the left arm of the child using TALC insertion tape marked to the nearest 2 mm costing 0.25 GBP. All measures followed procedures indicated in the United Nations (1986) guidelines.

Sample sizes were calculated separately for infants older and younger than 90 days. ‘Complete unreliability’ was defined as an intra-class correlation (ICC) of <0.4 and estimated the number of infants required for 90% power to distinguish ICC values of 0.6, which were defined as minimum reliability, from 0.4. This gave a required sample size of 71 infants in each age group for three observers. To allow for possible dropout from the study, the aim was to recruit at least 75 infants for each age group. Sample size for accuracy was 15 infants for every 75 infants recruited.

The study was then undertaken in three stages. First, to establish intrinsic reliability, the expert anthropometrist measured weight, length and MUAC among infants visiting the MCH at Kilifi District Hospital. Measurements were repeated after each cohort of 10 children. First and second set of measurements were recorded on separate forms. At the second stage, a training manual was produced for training HPs and CHWs following guidelines from the United Nations 1986 on anthropometry. The expert and the first author trained six HPs on anthropometry, safety procedures when handling infants and quality control of the measuring equipment. After the training, the HPs were divided into two groups of three each and repeatedly measured weight, length and MUAC of 150 infants (75 infants above and 75 below 90 days old). Each child was measured once by each of the three HPs. For every 5th child, the expert took measurements to determine accuracy. Further training was given to address issues arising in the second stage to establish HPs as trainers for the CHWs.

At the third stage, 18 CHWs were recruited, three from each of six sites: a district hospital, a peri-urban health centre, two rural health centres and two rural dispensaries. HPs conducted a 1-day practical training on anthropometry, safety procedures when handling infants and equipment quality control. This aimed to replicate the type of training that could be provided operationally. Then, each group of three CHWs independently measured MUAC, weight and length among 150 infants (75 under and 75 over 90 days old) at their health facilities. A single MUAC, weight and length measurement were taken once by each CHW on each infant. Measures were blinded from each other. At each facility, one HP took measures on every 5th infant to estimate accuracy.

Key findings

Key findings of the study were that among CHWs, ICCs pooled across the six sites (924 infants) were 0.96 (95% CI 0.95 0.96) for MUAC and 0.71 (95% CI 0.68-0.74) for WFLz. MUAC measures by CHWs differed little from their trainers: the mean difference in MUAC was 0.65 mm (95% CI 0.023-1.07), with no significant difference in variance (P = 0.075).

There are few published data on the accuracy of anthropometry in early infancy. None have used a systematic approach and included CHWs as used in this study. In this study, there was better concordance between CHWs and their trainers for MUAC than for WFLz.

In other studies, MUAC achieved high specificity (>95%) and varied sensitivity (48-58%) in identifying infants with severe malnutrition (WFLz <-3) and low-birth- weight (weight < 2500g). In such studies, however, there is a trade-off between specificity and sensitivity, and therefore reported levels of sensitivity and specificity should be interpreted within the study context and the cut-off used.

Absolute measures of MUAC, weight and length were more reliable than calculated z scores. Length was the least reliable when measured by HPs (pooled ICC was 0.82). Overall, WFLz was the least reliable anthropometric index (overall ICC was 0.71, and at one site, met the ‘completely unreliable’ criteria). The most likely explanation is that WFLz is very sensitive to changes in the absolute measurements, e.g. a 1cm change in length of a 6kg/65cm infant results in a 21% change in WFLz. Errors in calculating z scores were not studied.

This study evaluated CHWs across a representative range of sites in a rural district in a realistic setting after a typical practical training. Such an arrangement is likely to be similar to proposed changes by the Government of Kenya’s Ministry of Health when aiming to better identify infants in situ within their community at risk of malnutrition.

However, evaluation in the context of research may have resulted in a better than normal environment for measuring infants in that there was more time with fewer interruptions and greater supervision. This may limit the application of these findings to rapid assessment or a door-to-door visit scenario. Secondly, the measuring equipment was in good condition and regularly calibrated. This may not normally be the case in rural public health facilities in Kenya. Thirdly, owing to a delay in equipment availability, the expert anthropometrist was unable to take weight and length measurements in the same group of infants. It was therefore not possible to calculate and estimate expert’s intra-observer variation of WFLz. Finally, the majority of the infants involved in this study were recruited from the routine GM clinics and not randomly selected within the community, thus they were relatively healthy.

The authors conclude that CHWs can be trained to take absolute MUAC, weight and length measurements accurately and reliably among infants age <6 months. However, the length-based z score indices, LFAz and WFLz, are the least reliable anthropometric measures. Further studies of the generalisability of these findings in other settings and to assess the relationship of MUAC with mortality and illness to establish appropriate cut-off values for MUAC use among infants under 6 months old are needed.

¹Mwangome. M et al (2012) Reliability and accuracy of anthropometry performed by community health workers among infants under 6 months in rural Kenya. Tropical Medicine and International Health. Doi.10.1111/j1365- 3156.2012.02959.