Concurrently wasted and stunted children 6‐59 months in Karamoja, Uganda: prevalence and case detection

Abstract We assessed prevalence of concurrently wasted and stunted (WaSt) and explored the overlaps between wasted, stunted, underweight and low mid‐upper arm circumference (MUAC) among children aged 6–59 months in Karamoja, Uganda. We also determined optimal weight‐for‐age (WAZ) and MUAC thresholds for detecting WaSt. We conducted secondary data analysis with 2015–2018 Food Security and Nutrition Assessment (FSNA) cross‐sectional survey datasets from Karamoja. Wasting, stunting and underweight were defined as <−2.0 z‐scores using WHO growth standards. Low MUAC was defined as <12.5 cm. We defined WaSt as concurrent wasting and stunting. Prevalence of WaSt was 4.96% (95% CI [4.64, 5.29]). WaSt was more prevalent in lean than harvest season (5.21% vs. 4.53%; p = .018). About half (53.92%) of WaSt children had low MUAC, and all were underweight. Younger children aged <36 months had more WaSt, particularly males. Males with WaSt had higher median MUAC than females (12.50 vs. 12.10 cm; p < .001). A WAZ <−2.60 threshold detected WaSt with excellent sensitivity (99.02%) and high specificity (90.71%). MUAC threshold <13.20 cm had good sensitivity (81.58%) and moderate specificity (76.15%) to detect WaSt. WaSt prevalence of 5% is a public health concern, given its high mortality risk. All children with WaSt were underweight and half had low MUAC. WAZ and MUAC could be useful tools for detecting WaSt. Prevalence monitoring and prospective studies on WAZ and MUAC cut‐offs for WaSt detection are recommended. Future consideration to integrate WAZ into therapeutic feeding programmes is recommended to detect and treat WaSt children.

MUAC threshold <13.20 cm had good sensitivity (81.58%) and moderate specificity (76.15%) to detect WaSt. WaSt prevalence of 5% is a public health concern, given its high mortality risk. All children with WaSt were underweight and half had low MUAC.
WAZ and MUAC could be useful tools for detecting WaSt. Prevalence monitoring and prospective studies on WAZ and MUAC cut-offs for WaSt detection are recommended. Future consideration to integrate WAZ into therapeutic feeding programmes is recommended to detect and treat WaSt children. Wasting is also diagnosed by mid-upper arm circumference, that is, MUAC <12.5 cm and WHZ <−2.0 are indicators for acute malnutrition (UNHCR & WFP, 2011). Hereafter, acute malnutrition defined as MUAC <12.5 cm is referred to as low MUAC for ease of communication.
Recent evidence indicates that these different forms of undernutrition can coexist in some children Khara, Mwangome, Ngari, & Dolan, 2018;McDonald et al., 2013;Myatt et al., 2018;Schoenbuchner et al., 2019;Wells et al., 2019). Being wasted and stunted (WaSt) has a heightened risk of death equivalent to being severely wasted as defined by WHZ <−3 McDonald et al., 2013 ;Myatt et al., 2018). WaSt children have a 12-fold higher risk of death than children without anthropometric deficit (McDonald et al., 2013;Myatt et al., 2018). On the contrary, stunted children have 1.5-fold risk and wasted children have 2.3-fold risk compared with children without deficits (McDonald et al., 2013;Myatt et al., 2018;Odei Obeng-Amoako et al., 2020). Evidence suggests that children with WaSt, similar to children with severe wasting, could benefit from existing therapeutic feeding programmes to avert mortality (Myatt et al., 2018).
A recent analysis showed a pooled WaSt prevalence of 3% with a range of 0% to 8.0% among children aged 6-59 months in 84 countries . Drivers of the occurrence of WaSt are yet to be fully uncovered. However, age, sex and food insecurity season have been linked to WaSt Myatt et al., 2018;Schoenbuchner et al., 2019). How to screen and detect WaSt cases at the community and health facility levels is a critical prerequisite question for future public health programmes and policy decision-making.
WAZ and MUAC are considered the appropriate anthropometric indicators for detecting WaSt cases at risk of death (Myatt et al., 2018;. WAZ and MUAC for WaSt detection would potentially avoid the logistical complexities and errors associated with height measurements in the field (Myatt, Khara, & Collins, 2006).
Global and contextual evidence on prevalence and case detection of WaSt to inform decision-making is limited. Earlier analysis on WaSt focused on concurrency of WHZ <−2.0, HAZ <−2.0 and WAZ <−2.0 Khara et al., 2018;Myatt et al., 2018;Myatt et al., 2019;Schoenbuchner et al., 2019). Our study seeks to add to the evidence on WaSt by analysing the overlaps between four conventional anthropometric deficits: WHZ <−2.0, HAZ <−2.0, WAZ < −2.0 and MUAC <12.5 cm. Additionally, this analysis provides contextual evidence on the burden and detection of WaSt in a protracted food insecure setting. We used population-based FSNA survey datasets (June 2015-July 2018) to assess the prevalence of WaSt and  (Golden et al., 2006). The SMART methodology is a simplified and standardized household-level survey methodology used to determine the public health situation in humanitarian and nonhumanitarian settings (Golden et al., 2006). The FSNA used a two-stage cluster sampling approach. In the first stage, clusters were selected from an updated list of parishes in each district by using probability proportional to size sampling procedure. The second stage of sampling uti-
Karamoja is a semi-arid agro-pastoral region with a history of decades of conflicts and cattle raiding (Powell, 2010). Food insecurity is persistently high; over half of its 1.4 million population are food insecure (Uganda IPC Technical Working Group, 2019). The region is highly dependent on food rations, provided mostly by United Nations agencies as well as other development partners. Children younger than 5 years in the Karamoja Region compared with children in other parts of Uganda have a higher risk of death. It is estimated that child mortality is 102 deaths per 1,000 live births in Karamoja, compared with 64 per 1,000 live births nationally (UBOS & ICF, 2018).

| Study participants
Children aged 6-59 months with complete information on sex, MUAC, WAZ, WHZ and HAZ were included in this analysis. We identified and removed duplicate records from the database. Additionally, we removed children with bilateral pitting oedema and implausible anthropometric measurements such as height <45 and >120 cm and MUAC >20 cm (MoH Uganda & UNICEF, 2016). Implausible z-scores were removed based on WHO's biological plausibility anthropometric criteria for z-score outliers: HAZ <−6 and HAZ >6, WAZ <−6 and WAZ >5 and WHZ <−5 and WHZ >5 (WHO, 2009).

| Case definitions
We defined wasted, stunted and underweight based on the z-scores of the 2006 WHO growth standards (WHO, 2006) as follows: Wasted: weight for height z-score (WHZ) <−2.0.

| Data management and analysis
The names, types, length, coding schemes, units of measures of the variables and file format in each of the seven survey datasets were standardized and combined into one database (Myatt et al., 2018).
Statistical analyses were conducted using STATA 13.0, SPSS Statistics 23 and Microsoft Excel. We analysed anthropometric z-scores in STATA 13.0 (Leroy, 2011;WHO, 2009). A Venn diagram analysis was used to assess the intersection of children with WHZ <−2, HAZ <−2, WAZ <−2 and MUAC <12.5 cm (Venn, 1880). We adjusted for clustering effect of the multistage sampling during data analysis. We described the prevalence of WaSt and the characteristics of children with WaSt using summary statistics. We summarized continuous variables with medians and interquartile ranges (IQRs). We used percentages to describe categorical variables. We used chi-square to test the significance of differences between the proportions of children with and without nutritional deficits. After testing for normality of the continuous variables, we used Wilcoxon rank-sum test to compare the median values because they were not normally distributed (Bonita, Beaglehole, & Kjellström, 2006). Similar to an earlier analysis, we used Mann-Whitney U test to calculate the common language effect size (CLES) statistic to assess the probability of a higher median of WHZ and HAZ in wasted-only, stunted-only compared with WaSt cases (Conroy, 2012;Myatt et al., 2018). Receiver operating characteristic (ROC) curves and the Youden Index (sensitivity + specificity-100%) were used to determine optimal cut-off values of MUAC and WAZ for detecting WaSt in SPSS Statistics 23 (Myatt et al., 2018;Youden, 1950).

| Ethical considerations
We received approval to access the Karamoja FSNA datasets from the Office of the Prime Minister, Uganda. We sought a waiver of consent to use the FSNA datasets for this study from the Makerere Uni-      Table 6).

| DISCUSSION
We found a pooled WaSt prevalence of 4.96% in our analysis. Previous studies reported WaSt prevalence between 1% and 3% Saaka & Galaa, 2016 Routine national and subnational level nutrition surveys such as FSNA and SMART surveys need to be modified to include WaSt indicators to inform programme and policy decision-making.
WaSt prevalence varied by food security seasons in our study, similar to a Gambian study where being born at the start of the hunger season was a risk factor for early linear growth faltering and stunting (Schoenbuchner et al., 2019). These findings imply that WaSt prevalence could be seasonal and routine monitoring of WaSt prevalence would be required to inform effective detection and treatment.
About half (53.92%) of the children with WaSt had low MUAC.
An overlap between WaSt and low MUAC implies a child had simultaneously experienced the four common anthropometric deficits: wasting, stunting, underweight and low MUAC. Also, all WaSt children were underweight, as previously reported (Myatt et al., 2018). Thus, WaSt could be a useful indicator of concurrent anthropometric deficits (i.e., wasting, stunting and underweight) for communicating programmatic and policy outcomes (Myatt et al., 2018). Further research is needed to validate this finding.
In the present analysis, the proportion of children with different forms of anthropometric deficits were associated with age. Both underweight and low MUAC were common among children aged 6-11 months. However, the proportion of children with low MUAC declined thereafter, underweight rate was higher at age 12-23 months but decreased steadily after 24 months. Wasting was common among children <24 months but less among >24 months old.
Stunting was most prevalent beyond age 24 months. The higher underweight rate in the age group 36-48 months may indicate higher stunting rather than wasting rate (Waterlow, 1974). A child's body prioritizes weight gain over linear growth during nutritional recovery (Dewey et al., 2005). Wasting and stunting tend to be risk factors for each other (Myatt et al., 2018). A longitudinal study showed that wasting predicted the odds of stunting by three times. Conversely, stunting predicted the odds of wasting by two times (Schoenbuchner et al., 2019). Because of coexistence of growth faltering, integrated programming to address undernutrition in young children is needed (Bergeron & Castleman, 2012;Wells et al., 2019). Also, therapeutic feeding programmes should be optimized to include WaSt children and to promote linear growth in wasted children during treatment (Bergeron & Castleman, 2012;Khara & Dolan, 2014).
Younger children aged <36 months particularly males were more likely to have WaSt. The estimated WaSt male:female prevalence ratio is similar to findings reported in an earlier study (Myatt et al., 2018). Garenne et al. also showed that before age 30 months, males were 1.6 times more likely to be WaSt than females; however, the sex difference disappeared after age 30 months . These findings indicate that male children aged <36 months have a higher risk of WaSt Khara et al., 2018;Myatt et al., 2018). Given that children under 5 years are at risk of undernutrition, all children especially those <36 months old irrespective of sex should be targeted through a community-based WaSt screening programme.
Currently, the reasons for the observed sex difference in WaSt prevalence remain unclear and require further investigations. In a subpopulation analysis, we found that fewer males were WaSt and had low MUAC compared with those with WaSt-only, whereas more females had WaSt and low MUAC than those with WaSt-only. The sex disparities in childhood undernutrition could be due to the sex differences in body composition mainly muscle mass and body fat distribution measured by the different anthropometric indicators (Admassu et al., 2018;Park, Park, Kim, Kim, & Chung, 2011). Future studies should analyse the overlaps between wasting, stunting, underweight and MUAC to clarify understanding of sex and age patterns of WaSt.
In the present analysis, children with WaSt were more severely stunted than stunted-only children. The level of wasting did not differ between wasted-only and WaSt cases. This is contrary to previous studies where WaSt cases were both severely wasted and stunted compared with wasted-only or stunted-only cases Myatt et al., 2018

| Strengths and limitations
The dataset used in this analysis was collected based on a standardized protocol during each round of the survey. However, our study is not devoid of limitations usually associated with retrospective analysis. Random error is a common challenge associated with a fixed sample size of a historic dataset. Nonetheless, random error associated with secondary dataset was likely minimized because of the large sample size of the dataset and the probability proportional to population size sampling procedure used in the FSNA surveys (Hulley et al., 2001). Selection bias is an inherent limitation of existing datasets. We cleaned and censored incomplete data, missing data and implausible anthropometric data to minimize selection bias. Children aged 6-59 months who had complete anthropometric data were eligible for the present analysis. There was no sex difference between children included in the present analysis and those who were excluded due to incomplete anthropometric data. However, children in this study were older than those excluded for having incomplete data on the four anthropometric indicators. Our findings should be interpreted in light of this potential selection bias. Information bias is another limitation associated with a secondary dataset. The sources of information bias could be due to missing data and erroneous anthropometric measurements by field enumerators. Regular and standardized training of field enumerators, the use of standard operating procedures for sampling and collecting data in FSNA surveys probably minimized information bias in datasets. Causal inference about WaSt could not be made in our analysis because the design was cross-sectional. However, the findings could be useful baseline information for future studies on WaSt prevalence and case detection. The evidence on WaSt prevalence estimates and case detection would be useful for decisionmaking on WaSt programming and planning.

| CONCLUSION
WaSt prevalence of 5% found in Karamoja is a public health concern given the increased risk of death associated with WaSt.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.