Article Text

Download PDFPDF
Criteria for undernutrition screening in hospitalised infants under 6 months: a diagnostic accuracy study in a resource-poor setting
  1. Ifeyinwa Obiageli Ezeofor1,
  2. Ada Lizbeth Garcia2,
  3. Charlotte Margaret Wright1
  1. 1 Child Health, School of Medicine, Dentistry and Nursing, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, UK
  2. 2 Human Nutrition, School of Medicine, Dentistry and Nursing, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, UK
  1. Correspondence to Dr Ada Lizbeth Garcia, Human Nutrition, School of Medicine, Dentistry and Nursing, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow G31 2ER, UK; Ada.Garcia{at}


Purpose We aimed to describe the prevalence of undernutrition in hospitalised infants aged under 6 months and test the utility of simple index measures to detect undernutrition.

Design Diagnostic accuracy study: weight, length, mid-upper arm circumference (MUAC), triceps and subscapular skinfolds were measured in infants aged 2 weeks to 6 months admitted to a Teaching Hospital in Enugu, Nigeria. Index criteria : low (<−2SD) weight-for-age Z-scores (WAZ), weight-for-length Z-scores (WLZ); MUAC <11 cm. Reference definition: weight faltering (conditional weight gain below fifth percentile for healthy Nigerian infants) or sum of skinfolds (SSF) <10 mm.

Results Of 125 hospitalised infants, only 5% (6) were admitted specifically for undernutrition, but low SSF were found in 33% (41) and, 24% (25) with known birth weight had weight faltering, giving an undernutrition prevalence of 36%. Low WAZ was the most discriminating predictor of undernutrition (sensitivity 69%, positive predictive value 86%, likelihood ratio 5.5; area under receiver operator curves 0.90) followed by MUAC (73%, 73%, 4.9; 0.86), while WLZ performed least well (49%, 67%, 2.9; 0.84). Where both MUAC and WAZ were low, there was sensitivity 90%, positive predictive value 82% and likelihood ratio 8.7.

Conclusions Infants aged under 6 months admitted to hospital in Nigeria had a high prevalence of undernutrition. In young, high-risk population, a low WAZ alone was a valuable screening criterion, while combining weight with MUAC gave even higher discrimination. Measurement of length to calculate WLZ was a less useful predictor in this population.

  • nutrition
  • tropical paediatrics
  • screening

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • Undernutrition is more common in hospitalised infants due to disrupted feeding and underlying disease effects.

  • In low- and middle-income countries, undernutrition is usually identified using measurements of weight, height and mid-upper arm circumference.

  • Screening criteria for children aged under 6 months are lacking.

What this study adds?

  • Undernutrition in hospitalised infants under 6 months of age in Nigeria is common.

  • Mid-upper arm circumference and weight-for-age Z-scores are the best-performing measures to screen for undernutrition in very young infants.

  • Using weight-for-length Z-scores in a hospital setting missed many cases of low body fat and faltering growth.


The prevalence of undernutrition at population level in the first 6 months of life has previously been assumed to be low1 because of the relatively short time after birth and the assumption that infants will be protected by exclusive breast feeding. However, recent studies in early infancy in countries with higher prevalence of undernutrition have challenged these assumptions and highlighted the scale of the problem, with reported rates for moderate wasting in healthy infants under 6 months of between 15% and 34%.1–3 The main factors associated with high undernutrition prevalence in this age group are lack of exclusive breast feeding and underlying medical or neurodevelopmental problems.4

Childhood undernutrition at hospital admission has been reported as a problem even in affluent countries5 6 and when there is a higher background prevalence of malnutrition, admission may exacerbate and perpetuate the cycle of undernutrition, while malnutrition in hospitalised children increases the risk for complications and may prolong hospital stay.7 Despite this, studies from hospitalised children living in resource-poor settings are scarce. A prospective survey of infants aged 2–59 months in Gambia reported a high prevalence of wasting (41%), stunting (16.9%) and underweight (35.7%) in infants admitted to a paediatric ward of a rural health centre with a diagnosis of severe pneumonia,8 but this study included few children aged under 6 months. Undernutrition prevalence is usually reported using the common indicators for wasting (weight-for-length Z-scores, WLZ), stunting (lengh-for-age Z-scores) and underweight (weight-for-age Z-scores, WAZ) based on WHO references. The pros/cons of different indicators to define undernutrition is a complex issue, because each of them measures different aspects, which will result in different prevalence rates.9 ,10 ,11 For example, it has been argued that using WAZ to define undernutrition in hospitalised infants rather than WLZ is likely to result in overestimation of undernutrition, because of the acute nature of disease-associated malnutrition.12 On the other hand, undernutrition can be the result of chronic deficit of energy and nutrient intake and prolonged periods of inflammation. Other common screening methods for undernutrition, such as mid-upper arm circumference (MUAC) or skinfolds, are less often used in hospital settings. In more affluent countries, change in weight over time rather than a single measurement is used to avoid identifying naturally short rather than undernourished children. The limitation of this approach is the need for at least two measurements and some allowance for baseline size, as healthy large infants will tend to drop toward average, while initially small infants will tend to rise.13

A further problem is a lack of suitable standards for very young infants. The WHO growth chart project published standards for subscapular and triceps skinfolds in 2007, which were only from age 3 months to 5 years.14 15 MUAC has been recommended as the best case-detection method for severe malnutrition in field surveys, because of its high performance in terms of age dependence, precision, accuracy, sensitivity and specificity.16 17 Its use in clinical settings in affluent countries with low malnutrition prevalence has been debatable, because of the sensitivity and cut-offs, but it is advocated for settings with high undernutrition prevalence.18 MUAC measurement is simple, cheap and acceptable; therefore, it is recommended that programmes’ screening and treating severe malnutrition move toward a MUAC-based case-detection, referral and admission criteria.16 However, as there are currently no threshold MUAC values for infants younger than 6 months, the recommended values for children 6–60 months of age are often applied and the performance of MUAC in hospitalised very young infants has not been explored. MUAC has been much promoted in recent years as a simple and useful measurement in clinical settings (Berkley et al, 2005), but the threshold to use in young infants is not clear. A study in rural Gambian infants aged 6–14 weeks19 recommended using MUAC <11 cm, after finding this to be more effective than WLZ of <−2 in predicting malnutrition-related mortality.

There is, thus, a need to establish which measures easily used in a hospital setting will identify undernutrition in very young infants in populations with high prevalence. Therefore, we aimed to:

  1. Determine the prevalence of confirmed undernutrition in hospitalised infants aged under 6 months in a resource-poor setting, using weight faltering (WF) defined by conditional weight gain (CWG) or low fat stores, as a reference definition.

  2. Evaluate the performance and utility of WLZ, WAZ and MUAC as index screening methods to identify confirmed undernutrition in very young infants.


Study design and setting

This was a diagnostic accuracy study, conducted at the University of Nigeria Teaching Hospital (UNTH), Enugu, Nigeria using cross-sectional data collected as part of a PhD study programme.20 This is a large referral hospital with over 500 beds that serves Enugu city with a population of over 700 00021 and community services that extend to various states in the country, particularly those in the southeast geopolitical zone. Social and economic activities in the zone are farming and small-scale and medium-scale trading. There is a high prevalence of childhood morbidity due to infectious diseases and inequality in maternal and child healthcare access are major public health issues.20 22


Infants admitted to the paediatric wards aged up to 6 months were recruited using purposive sampling between February and July 2012. Power calculation conducted in advance suggested that 200 hospitalised children in total would give 95% power to estimate the prevalence of undernutrition estimated at 14% with a precision of ±5%. Infants under the age of 2 weeks were later excluded to avoid including children with newborn issues who in other centres would be managed in neonatal units.

Weight data were also collected on healthy infants attending the Infant Welfare Clinic at the same hospital to provide healthy norms for conditional weight gain.

Data collection

After obtaining informed consent from mothers of the participating infants, anthropometric measurements were taken using WHO standard operating techniques23 by one trained nutritionist researcher. Weights were measured using a SECA 385 electronic baby scale, regularly calibrated with a known weight. Recumbent length was measured using a SECA 416 infantometer. MUAC was measured using a narrow, reusable non-stretch tape. Skinfolds (triceps and subscapular) were measured using Holtain skinfold callipers. Each measure was collected at least three times until three were within acceptable range of each and the average of these three readings was used for each measurement. Gestational age, date of birth, weight at birth and information on diagnosis were obtained from each individual infant’s hospital case note.

Data for the healthy comparison group were collected by approaching mothers attending the well-baby clinic and, with their consent, recording that days weight as well as transcribing all previous weights from the child’s clinic card at birth, 6 weeks, 3 and 6 months. Children who were known to be malnourished or unwell were excluded. The infants were weighed in the nude using paediatric weighing scales with pan, which were routinely maintained and calibrated.


Data were analysed using SPSS (IBM SPSS Statistics for Windows, V.25.0; IBM, Armonk, New York, USA). Anthropometric Z-scores were calculated using the WHO anthropometric software, adjusted for gestational age at birth.24 Due to the lack of skinfold norms for infants under 3 months, a threshold of 10 mm for sum of skinfolds (SSF) was used to define low skinfolds, based on the empirical lower limit observed across the first year in a UK survey.7 20 CWG is a measure of WAZ change which adjusts for regression to the mean, the tendency for small infants to catch up toward the median and for larger infants to drift down toward the median.25 In this study, this was calculated as the change in WAZ (SD scores) from birth to the later age when data were collected, as follows:

Embedded Image

where B is the regression coefficient of birth weight Z regressed onto later WAZ in a healthy population. Because of a lack of norms for CWG under age 4 weeks and uncertainty about local growth rates, regression constants and norms for weight gain were established using the retrospective weights of the healthy infants attending the hospital Infant Welfare clinic. Based on this data (see online supplementary table), the threshold for WF in the hospitalised infants was defined as a CWG of −2SD.20 The reference definition of undernutrition was then defined as SSF < 10 mm or CWG <−2SD. Where there was no CWG due to the lack of birth weight, an infant was defined as test negative unless they had low SSF.

For each of the three index measures (WLZ, WAZ and MUAC), we calculated their sensitivity and positive predictive value (PPV) to detect undernutrition, as well as the likelihood ratio (LR): the ratio of the percentage test positive for each measure to the prevalence in those who were test negative. The three indicators above were also compared using receiver operator curves (ROCs). Thresholds for low WLZ and WAZ were set at −2 SD, the WHO thresholds for moderate malnutrition.23 For MUAC, a threshold of < 11 cm was used as described above.


Data from the hospital Infant Welfare Clinic were obtained for 411 healthy infants, yielding 1480 measurements in total (online supplementary table). The regression constants of the later weight compared with birth weight decreased with age, but the lower limit for CWG remained around −2 SD. A total of 125 hospitalised infants aged 2 weeks to 6 months were recruited. Overall 22% were born preterm, but prematurity was given as the reason for admission in only 9%. The the most common reason for admission was sepsis, followed by respiratory problems and surgical conditions. Only six cases were admitted specifically because of severe undernutrition, but the mean weight and CWG scores were low on average (table 1).

Table 1

Variation in characteristics of subjects by age

The proportion with low WLZ, WAZ or MUAC varied from 25% for WLZ to 36% for MUAC, while 24% had low CWG and 33% had low SSF (table 2). This meant that overall 45 infants (36%) met our a priori reference definition of undernutrition (low SSF or CWG). The great majority of these had low skinfolds, with only four (8%) having low CWG in isolation (figure 1). A much higher proportion of infants aged under 3 months were below the thresholds for most measures, with the exception of WLZ.

Figure 1

Degree of overlap between slow weight gain and low skinfolds. Values are percentage of all children with each characteristic. CWG, conditional weight gain.

Table 2

Prevalence of undernutrition using different criteria, broken down by age

MUAC had the highest sensitivity under 3 months but the lowest at 3–6 months, while WLZ had the lowest sensitivity under age 3 months and overall detected only half of all cases. WAZ had the highest PPV at all ages and the highest sensitivity overall, detecting more than two-thirds of the cases. The LR was highest overall for low WAZ in infants aged 3–6 months, but in infants under 3 months MUAC had the higher LR (table 3). When comparing the three values using ROC, WAZ consistently had the highest area under the curve (table 3, figure 2). Overall 42% of children had either low WAZ or MUAC or both and this criterion detected 90% of cases with a PPV of 82% and a LR of 8.7 (table 3).

Figure 2

Weight-for-length Z-scores (WLZ), weight for age Z-scores and mid-upper arm circumference (MUAC) compared using receiver operator curves with the outcome of undernutrition.

Table 3

Extent to which field measures (MUAC, WLZ and WAZ) predict undernutrition (low CWG and/or low skinfolds)


All anthropometric measures are only a screen for undernutrition rather than a gold standard measure of it and all have advantages as well as disadvantages. While the widely used WHO public health ‘definitions’ of acute malnutrition are important to allow consistency between countries and centres, it is important to remember that these are only screening definitions, which by their nature would be expected to misclassify some healthy children as undernourished, while missing other true cases. Some children with WAZ may be naturally short rather than undernourished, while other children with normal weight could be tall but emaciated. WLZ and MUAC are intended to identify wasting, but low values for these might also screen in healthy children with a narrow build and in infancy it might not detect children who are growing slowly and becoming stunted. Thus, it is important to validate these criteria against measures that are more specific to undernutrition. Here we have used skinfolds, which are a more direct assessment of the size of fat stores that are essential to sustain health, rather than lean mass. CWG detects children whose weight has dropped below its expected trajectory, which should avoid the misclassification of genetically small or lean children as undernourished.

The performance of any screening measure, in terms of its positive or negative predictive values depends crucially on the underlying prevalence of the condition being targeted. In this study, though few of the hospitalised infants were admitted specifically for treatment of undernutrition, over a third had low fat stores, with or without slow weight gain, which likely reflects the high rates of pathology found in this hospitalised population. In practice, most children with low WAZ were also thin or growing slowly so that, despite it theoretical limitations, it was actually a low WAZ that most reliable detected children with undernutrition, followed by MUAC, while weight for length performed least well. This presumably reflects a greater tendency for very young infants to respond to nutritional compromise by growing slowly, rather than wasting.

A limitation of this study was that it was undertaken in only one centre and there is a need to replicate this study in other units with different case mixes, where results might be different.

However so far there have been very few studies in this young age group. A similar recent study in infants aged 1–12 months found low rates of malnutrition in hospitalised infants in the UK, but much higher rates in a tertiary referral unit in Iran.7 Similar rates of undernutrition have been reported previously in another Nigerian hospital-based study,26 but it was not clear there how much this simply reflected the high background prevalence of undernutrition in Nigeria.27 However in our study, it was clear that undernutrition was much more common in hospitalised infants than in the well-baby clinic, since rates of WF were four to five times more than the lower fifth percentile found in healthy Nigerian children. This suggests that in a Nigerian setting, levels of undernutrition are high for very young sick infants during hospitalisation, which could have implications for recovery and prolonged hospital stay.

The limitation of any study of malnutrition is the absence of any gold standard measure and the results of different studies will vary depending on the measure used. In more affluent, low-prevalence settings, measures of weight gain, rather than a single weight are favoured as they are assumed to be more specific. Their limitation is the need for at least two measures collected over time and any velocity measure of this kind is thus more prone to error, as it relies on weights at two time points. In this sample we cannot assume that the birth weight was collected accurately, but it is reassuring that most infants with low CWG also had low skinfolds, suggesting that most were accurate. Around one in six children did not have a recorded birth weight and this is more likely in settings without universal maternity provision, making velocity measures impractical for general use. A very similar CWG measure, also using a fifth percentile threshold has been successfully used in another study to identify infants at risk for growth faltering in Nigerian infants at the 6–8 week postpartum check.28 The 5% threshold for CWG, although apparently less stringent than the −2 Z-scores threshold (equivalent to the second centile), was in fact comparably stringent to the other measures used. The lower limits were established for this study using healthy Nigerian infants, but the average weight gain seen there was slower than rates seen in UK infants.9 It must also be recognised that for some infants the exact date of birth may not be known accurately, which would also introduce some imprecision into the WAZ.

Skinfolds have not yet been much used in a developing world context to assess undernutrition, although norms for skinfolds were included in the WHO growth chart project14 and it is the level of fat reserves that is most likely to determine the level of severity of undernutrition. The problem for our study was that the WHO norms only started at 3 months, which would exclude over half of the children screened. However, we had already observed that skinfold levels showed little variation across the first year, making a single lower threshold probably valid.20 Skinfolds will not allow any assessment of abdominal fat, but it seems reasonable to assume that these will be quite closely correlated.

Whatever measure is used, what matters is the extent to which that measure reflects underlying functional impairment or long-term risk, so future studies of this kind should consider functional outcomes such as later morbidity or mortality and longer term growth impairment.


This study highlights the importance of screening for malnutrition at hospital level in these very young infants and that a low WAZ alone is the most valuable screening criterion, while MUAC is also helpful. Combining weight with MUAC gives even higher discrimination, but the addition of a length measurement does not.


Thanks to all the health staff who participated in the study and to the Ford Foundation International Fellowships Program (IFP), New York, USA, for sponsoring IOE’s PhD studentship and the data collection. We also thank SN Ibeziako who facilitated hospital access and help with supervision during data collection in Nigeria Teaching Hospital (UNTH).



  • Twitter @DrAdaGarcia

  • Correction notice This paper has been corrected since it was published online. The first author's affilition has been changed.

  • Contributors CMW conceived the study design. IOE created the questionnaire, collected the data, undertook the initial analyses and produced the first draft of the paper. CMW and ALG helped plan the study and supervised the analyses. CMW undertook further analyses and ALG drafted the paper. All authors contributed to successive drafts and have approved the final draft.

  • Funding Ford Foundation International Fellowships Program (IFP), New York, USA.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval Ethical approval was obtained from the College of Medicine Ethics Committee at the University of Glasgow (Reference No 2011018) and the Medical Research Ethics Committee of the UNTH, Enugu, southeast Nigeria (Reference No NHRE/05/01/2008B–FWA00002458–IRB00002323).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available on reasonable request.

Linked Articles