Abstract
Background:
The sub-Saharan Africa region faces a significant burden of child undernutrition. Indicators of child nutrition are often interrelated; however, previous research has typically analyzed them separately.
Methods:
This article investigated the risk of joint child stunting, underweight, and wasting in children under 5 years old in Malawi, utilizing trivariate logistic regression. The study used data from the 2024 Malawi Demographic and Health Survey. All computations were carried out using R software version 4.5.3.
Results:
The findings revealed that stunting and underweight frequently occurred together in children, sharing similar risk factors. However, the cases of simultaneous stunting and wasting, as well as underweight and wasting, could not be adequately explained by the available factors. The likelihood of experiencing both stunting and underweight was lower in children aged 12 months and above, those from middle- or high-income households, and those living in rural areas. Conversely, the risk of these two conditions was higher in children with higher birth orders.
Conclusion:
Analyzing the risk factors associated with concurrent child undernutrition outcomes provides valuable data for developing child health policies aimed at eliminating multiple forms of undernutrition.
1 Introduction
Child undernutrition remains a public health concern in the least developed nations, particularly in sub-Saharan Africa and South-East Asia (, ). Undernutrition refers to having insufficient nutrients and energy in a person's body, which can cause adverse effects on body composition, function, and health outcomes (). This is measured using a quantitative standard normal probability distribution score (z-score) of one of the three anthropometric indicators: height-for-age (HAZ) for stunting, weight-for-height (WHZ) for wasting, and weight-for-age (WAZ) for underweight (–). A z-score of less than or equal to -2.0 in any of these measures is considered an indicator of undernourishment in a child (). Poor nutrition in children can result in stunted growth, developmental issues, and even higher rates of child mortality (, ).
Risk factors for child undernutrition include the child's sex, age, birth order, birth interval, breastfeeding duration, promptness of breastfeeding initiation, and whether the child has fever symptoms. Other risk factors include the mother's age at childbirth, occupation, education level, the household's access to clean drinking water, maternal height, household wealth index, and the type of toilet facility used by the household (, –). For instance, male babies and first- or second-born children are more likely to experience stunting, while younger children are more likely to experience wasting. Additionally, drinking water from unimproved sources is associated with stunting in children (, ). Research shows that timely and adequate interventions for child undernutrition can prevent negative consequences and promote catch-up growth in later stages of life (). To devise appropriate interventions, it is necessary to conduct more research on the dynamics of child undernutrition in high-burden settings.
In many cases, a malnourished child exhibits multiple indicators of undernutrition, often due to shared social disadvantages that contribute to several health deficiencies (–). This provides an opportunity to study multiple nutrition indicators simultaneously (). Previous research has primarily focused on identifying the risk factors associated with child undernutrition using a single anthropometric indicator (–). This study employs a trivariate binary logistic regression analysis to examine the common risk factors for joint stunting, wasting, and underweight among children in Malawi.
The child undernutrition indicators are often paired; however, most public health research analyses them separately due to limitations in available statistical methods. Many studies focus on individual risk factors for stunting, underweight, and wasting, rather than looking at these indicators together. This research addresses that gap by analyzing the nutrition indicators jointly and identifying common risk factors for any two indicators simultaneously. This comprehensive analysis will enable policymakers to take decisive action based on the identified drivers of child undernutrition.
This study is organized as follows: Section 2 presents the data and statistical methods used, and Section 3 gives the results. This is followed by the discussion of findings in Section 4 and the conclusion in Section 5.
2 Methods
2.1 Data
The research utilized data from the household member recode file of the 2024 Malawi Demographic and Health Survey (MDHS), which aimed at collecting an up-to-date estimates regarding the basic demographic and health indicators in Malawi. The survey was implemented between 12 May and 31 August 2024 by the National Statistical Office (NSO) of Malawi, with the technical support from the ICF and DHS programme of the United States of America. The survey involved a two-stage cluster-stratified sampling technique in which 792 clusters were randomly sampled from across the country in the first stage and 23,070 households at second stage (). A subsample of these households produced children's data reported by biological mothers, which had a total sample of 4,231 children aged 0–59 months. The interest of this study was to analyze joint undernutrition outcomes in the study area including clusters with extreme values. Some children had missing values in the response variables. This study has dropped the missing values and used complete case analysis, which involved a sample of 3,323 children.
The study involved three dependent variables. The first was whether or not the under-five child was stunted, which was characterized by height-for-age standardized score (HAZ) of less than or equal to –2.0. The second variable was whether the under-five child was underweight using weight-for-age standardized score (WAZ) of less than or equal to –2.0. The third variable was whether the under-five child was wasted, based on weight-for-height standardized score (WHZ) of less than or equal to –2.0 (). Each child had measurements on these variables that were taken around the same time; hence, a joint or trivariate regression analysis was carried out. The independent variables included the child's sex, age in months, birth order number, preceding birth interval in months, and whether the child had malaria diagnosed by rapid test at the time of the survey (, , ). In addition, the mother's education level, place of residence, and household's wealth index were also included in the modeling (, ).
The 2024 MDHS data analyzed during this research are publicly available for users through the following link: https://www.dhsprogram.com/data/dataset/Malawi_Standard-DHS_2024.cfm?flag=1. Data cleaning was carried out using the STATA package version 12.0.
2.2 Trivariate logistic regression model and its estimation
Let (Yi1, Yi2, Yi3) represents number of observed cases of joint stunting, underweight, and wasting in i-th child surveyed. Let the actual measurements be represented by (yi1, yi2, yi3), where i = 1, 2, ..., n is the unit of analysis. At each observation, let 1 represents the outcome that a child was stunted, or underweight, or wasted, and 0 otherwise. This gave rise to a trivariate Bernoulli experiment with the following ordered triple outcome:
each event in Equation (1) occurring with respective probability:
where p000+p001+p010+p100+p011+p101+p110+p111 = 1. If the probabilities of events of 1's in Equation (2) are fixed and identical across the observations, then the set (Yi1, Yi2, Yi3) has a trivariate Bernoulli probability distribution, with the following mass function (, ):
As it can be appreciated from the exponential form of the trivariate Bernoulli distribution in Equation 3, the marginal and joint event probabilities of the three variables given in Equation 2 can be estimated though seven log odds natural parameters in a trivariate logistic model (, ). The trivariate logistic regression is given by
The probabilities in Equation (4) were estimated from the log odds link functions given below:
where is vector of independent variables' values observed on i-th child; were the fixed effects, while ϵi represented the model's error term at each level of the response variable, which was assumed to have mean zero ().
Solving the system of equations in Equation 5 provided the following estimates of the probabilities given in model Equation 4
where the probability p000 in Equation (6) is given by Dai () and Islam ():
The model parameters were estimated using maximum likelihood given by
where A is as defined in Equation 7. The partial derivatives of the log-likelihood function derived from Equation (8) with respect to the model fixed parameters pertinent to particular marginal model provided the score functions (), which helped to find the model solutions. The random effects were estimated in turn, using the updated log-likelihood function. The fixed effect ML estimates have usual interpretation as marginal log odds of success in the respective outcome, when comparing one level of a covariate to the other called reference level. In this research, the interest was on estimating risk factors of joint occurrence of stunting and underweight, stunting and wasting, as well and underweight and wasting. The joint model estimates were calculated using the R package VGAM () in R software version 4.5.3. The R codes used are given as an Appendix.
3 Results
3.1 Data summary
The summary of the data given in Table 1 showed that cases of low height-for-age (i.e., stunting), low weight-for-age (i.e., underweight), and weight-for-height (i.e., wasting) were present in less than 5% of the studied children. The conditions were generally more prevalent in male children, those aged 12 months and above, those staying in poor households, and those with preceding birth interval of not more than 36 months. In addition, three conditions were also in large proportions in children who breastfed 13 months and above, those whose mothers had no education or primary education, and children residing in rural areas. The chi-squared test results showed that stunting was significantly associated with a child's sex and preceding birth interval. While underweight was associated with the child's age group, his or her preceding birth interval, duration of breastfeeding, mother's education, household's wealth, and place of residence, wasting was associated with the child's age group, and his or her duration of breastfeeding.
Table 1
| Stunting | Underweight | Wasting | |||||
|---|---|---|---|---|---|---|---|
| Characteristic | n (%) | Stunted (%) | χ2 p-value | Underweight (%) | χ2p-value | Wasted (%) | χ2p-value |
| Overall sample | 5,370 (100) | 105 (01.96) | 125 (02.33) | 128 (02.38) | |||
| Child's sex | 0.029 | 0.220 | 0.166 | ||||
| Male | 2,654 (49.42) | 63 (02.37) | 55 (02.07) | 71 (02.68) | |||
| Female | 2,716 (50.58) | 42 (01.55) | 70 (02.58) | 57 (02.10) | |||
| Child's age group | 0.308 | 0.011 | 0.001 | ||||
| 0–11 months | 1,160 (21.60) | 17 (01.47) | 13 (01.12) | 19 (01.64) | |||
| 12–23 months | 1,170 (21.79) | 21 (01.79) | 28 (02.39) | 46 (03.93) | |||
| 24–35 months | 1,028 (19.4) | 22 (02.14) | 35 (03.40) | 27 (02.63) | |||
| 36–47 months | 1,002 (18.66) | 27 (02.69) | 23 (02.30) | 21 (02.10) | |||
| 48–59 months | 1,010 (18.81) | 18 (01.78) | 26 (02.58) | 15 (01.49) | |||
| Preceding birth interval | < 0.001 | 0.069 | 0.934 | ||||
| < 24 months | 291 (05.42) | 9 (03.09) | 10 (03.44) | 6 (02.06) | |||
| 24–36 months | 628 (11.69) | 26 (04.14) | 21 (03.34) | 15 (02.39) | |||
| >36 months | 4,451 (82.89) | 70 (01.57) | 94 (02.11) | 107 (02.4) | |||
| Malaria result | 0.292 | 0.877 | 0.427 | ||||
| Negative | 3,472 (64.66) | 73 (02.10) | 80 (02.30) | 87 (02.51) | |||
| Positive | 1,898 (35.34) | 32 (01.69) | 45 (02.37) | 41 (02.16) | |||
| Mother's education | 0.415 | 0.056 | 0.800 | ||||
| No education | 376 (07.00) | 9 (02.39) | 12 (03.19) | 9 (02.39) | |||
| Primary | 3,489 (64.97) | 74 (02.12) | 91 (02.61) | 88 (02.52) | |||
| Secondary | 1,393 (25.94) | 20 (01.44) | 21 (01.51) | 29 (02.08) | |||
| Higher | 112 (02.09) | 2 (01.79) | 1 (00.89) | 2 (01.79) | |||
| Household's wealth | 0.104 | < 0.001 | 0.820 | ||||
| Poor | 2,371 (44.15) | 57 (02.40) | 79 (03.33) | 59 (02.49) | |||
| Middle | 883 (16.44) | 15 (01.70) | 11 (01.25) | 22 (02.49) | |||
| Rich | 2,116 (39.4) | 33 (01.56) | 35 (01.65) | 47 (02.22) | |||
| Place of residence | 0.447 | 0.004 | 0.838 | ||||
| Urban | 970 (18.06) | 16 (01.65) | 16 (01.65) | 24 (02.47) | |||
| Rural | 4,400 (81.94) | 89 (02.02) | 109 (02.48) | 104 (02.36) | |||
| Mean birthorder (SD) | 2.85 (01.93) | ||||||
Distribution of unweighted cases of stunting, underweight, and wasting by the socio-demographic characteristics of the children, 2024 MDHS complete cases data.
SD = standard deviation, χ2 = chi-squared test, n = sample size.
3.2 Effects of socio-demographic factors on joint child stunting and underweight, and wasting based on the trivariate logistic regression model
The model estimates presented in Table 2 indicated that the log odds of being both stunted and underweight were significantly lower in children aged 12 months and older, those with preceding birth interval of 24 months and above, children from middle income or rich families, and those residing in rural areas. The log odds of both stunting and underweight increased with increasing birth order. While the effects of being malarial, being female, and mother's education were not significant, the estimate of model intercept showed that cases of joint stunting and underweight were on the rise in the study population.
Table 2
| Variable | Stunting Log odds (SE, p-value) | Underweight Log odds (SE, p-value) | Stunting and underweight Log odds (SE, p-value) |
|---|---|---|---|
| Intercept | –1.51 (1.29, 0.2439) | –3.27 (1.29, 0.0114) | 28.5 (12.4, 0.0218) |
| Child's sex | |||
| Male* | |||
| Female | –0.32 (0.29, 0.2689) | 0.13 (0.27, 0.6236) | 1.43 (1.83, 0.4344) |
| Child's age group | |||
| 0–11 months* | |||
| 12–23 months | –0.26 (0.49, 0.5921) | –0.03 (0.48, 0.9565) | –13.0 (5.87, 0.0265) |
| 24–35 months | –0.13 (0.44, 0.7651) | 0.49 (0.42, 0.2396) | –7.50 (3.88, 0.0531) |
| 36–47 months | –0.31 (0.49, 0.5305) | –0.02 (0.48, 0.9626) | –12.2 (5.75, 0.0334) |
| 48–59 months | –0.12 (0.45, 0.7967) | 0.22 (0.44, 0.6245) | –3.77 (3.28, 0.2505) |
| Preceding birth interval | |||
| < 24 months* | |||
| 24–36 months | 0.39 (0.44, 0.3761) | 0.54 (0.46, 0.2411) | –4.40 (3.27, 0.1790) |
| >36 months | –1.02 (0.44, 0.0204) | –0.12 (0.44, 0.7802) | –10.6 (5.09, 0.0368) |
| Malaria result | |||
| Negative* | |||
| Positive | –0.01 (0.36, 0.9866) | 0.03 (0.33, 0.9366) | –0.68 (2.18, 0.7535) |
| Mother's education | |||
| No education* | |||
| Primary | –0.46 (0.46, 0.3174) | –0.04 (0.44, 0.9342) | –4.11 (3.23, 0.2038) |
| Secondary | –0.29 (0.60, 0.6277) | –0.78 (0.63, 0.2167) | 6.09 (4.41, 0.1674) |
| Higher | 0.89 (0.98, 0.3622) | 0.57 (1.21, 0.6382) | –11.1 (8.28, 0.1817) |
| Household's wealth | |||
| Poor* | |||
| Middle | –0.56 (0.49, 0.2526) | –1.51 (0.56, 0.0067) | –6.97 (4.75, 0.1421) |
| Rich | –0.61 (0.41, 0.1378) | –0.99 (0.40, 0.0132) | –7.66 (4.14, 0.0642) |
| Place of residence | |||
| Urban* | |||
| Rural | –0.31 (0.49, 0.5310) | –0.32 (0.49, 0.5191) | –10.3 (4.89, 0.0354) |
| Birthorder | –0.03 (0.10, 0.7629) | 0.09 (0.08, 0.3092) | 3.35 (1.46, 0.0223) |
Effects of child characteristics on joint stunting and underweight upon fitting a bivariate logistic regression model to 2024MDHS data.
SE, standard error; (*), reference category.
In the marginal univariate outcomes, the log odds of stunting were found to be significantly lower in children with birth interval of above 36 months, while the log odds of underweight were lower in children from middle and rich wealth quintile households. The p-values indicated that most variables had significant effects in the paired outcome compared with the marginal univariate outcomes. However, the directions of the estimates for most of the child characteristics were consistent between the bivariate and univariate outcomes.
The estimates given in Table 3 are for effects of child characteristics on joint stunting and wasting, as well as on their marginal outcomes. The results showed that the log odds of being both stunted and wasted were lower in children aged 12 months and older, and those residing in rural areas, although the effects were not statistically significant. None of the variables has neither significant effects in univariate outcome case.
Table 3
| Stunting | Wasting | Stunting and wasting | |
|---|---|---|---|
| Variable | Log odds (SE, p-value) | Log odds (SE, p-value) | Log odds (SE, p-value) |
| Intercept | -1.83 (2.11, 0.3859) | -4.08 (2.13, 0.0555) | -48.1 (38.9, 0.2163) |
| Child's sex | |||
| Male* | |||
| Female | –0.30 (0.47, 0.5284) | -0.40 (0.40, 0.3169) | 11.5 (8.07, 0.1526) |
| Child's age group | |||
| 0–11 months* | |||
| 12–23 months | –0.07 (0.54, 0.8983) | 0.63 (0.44, 0.1512) | –0.34 (3.90, 0.9312) |
| 24–35 months | –0.09 (0.61, 0.8894) | 0.18 (0.54, 0.7365) | –5.84 (6.09, 0.3377) |
| 36–47 months | 0.01 (2.18, 0.9952) | 0.05 (2.17, 0.9799) | –26.7 (28.4, 0.3484) |
| 48–59 months | –0.05 (0.69, 0.9390) | –0.24 (0.68, 0.7178) | –9.40 (7.98, 0.2390) |
| Preceding birth interval | |||
| < 24 months* | |||
| 24–36 months | 0.33 (0.94, 0.7280) | 0.74 (1.36, 0.5868) | 5.62 (10.0, 0.5757) |
| >36 months | –0.87 (0.94, 0.3566) | 0.80 (1.31, 0.5404) | 14.6 (13.2, 0.2689) |
| Malaria result | |||
| Negative* | |||
| Positive | –0.20 (0.70, 0.7724) | –0.94 (0.67, 0.1576) | 17.7 (12.9, 0.1723) |
| Mother's education | |||
| No education* | |||
| Primary | –0.48 (0.72, 0.5007) | 0.02 (0.63, 0.9711) | 2.19 (6.17, 0.7225) |
| Secondary | –0.39 (0.82, 0.6365) | –0.40 (0.74, 0.5857) | 11.1 (10.3, 0.2835) |
| Higher | 0.70 (1.36, 0.6071) | 0.27 (1.55, 0.8637) | –6.05 (16.7, 0.7168) |
| Household's wealth | |||
| Poor* | |||
| Middle | –0.68 (0.78, 0.3836) | –0.08 (0.55, 0.8902) | 8.58 (8.16, 0.2926) |
| Rich | –0.41 (0.62, 0.5070) | –0.33 (0.52, 0.5209) | 12.8 (9.83, 0.1936) |
| Place of residence | |||
| Urban* | |||
| Rural | –0.19 (0.53, 0.7171) | 0.12 (0.48, 0.7983) | –5.99 (5.56, 0.2812) |
| Birthorder | –0.03 (0.13, 0.7957) | 0.04 (0.11, 0.7042) | 3.41 (2.61, 0.1910) |
Effects of child characteristics on joint stunting and wasting upon fitting a bivariate logistic regression model to 2024MDHS data.
SE, standard error; (*), reference category.
The results in Table 4 showed that the log odds of being both underweight and wasted were lower in children living in middle income households. The log odds of both underweight and wasting were higher in children aged 12–47 months. In the marginal univariate outcomes, the log odds of underweight were significantly lower in children from middle wealth family, while chances of wasting were low in malarial children and high in children aged 12–23 months. Again, most variables had no significant effects on either underweight or wasting or both.
Table 4
| Variable | Underweight Log odds (SE, p-value) | Wasting Log odds (SE, p-value) | Underweight and wasting Log odds (SE, p-value) |
|---|---|---|---|
| Intercept | –3.34 (1.92, 0.0818) | –4.50 (1.98, 0.0230) | –41.0 (219.2, 0.8517) |
| Child's sex | |||
| Male* | |||
| Female | 0.18 (0.30, 0.5503) | –0.31 (0.29, 0.2834) | –1.22 (1.40, 0.3832) |
| Child's age group | |||
| 0–11 months* | |||
| 12–23 months | 0.35 (0.51, 0.4894) | 0.82 (0.45, 0.0691) | 4.26 (2.38, 0.0734) |
| 24–35 months | 0.64 (0.48, 0.1852) | 0.39 (0.48, 0.4188) | 2.88 (2.03, 0.1550) |
| 36–47 months | 0.31 (0.51, 0.5429) | 0.39 (0.49, 0.4198) | 2.72 (2.13, 0.2024) |
| 48–59 months | 0.24 (0.53, 0.6485) | –0.02 (0.56, 0.9698) | –10.6 (109.2, 0.9226) |
| Preceding birth interval | |||
| < 24 months* | |||
| 24–36 months | 0.31 (0.56, 0.5762) | 0.50 (0.73, 0.4940) | 0.20 (2.59, 0.9374) |
| >36 months | –0.22 (0.51, 0.6602) | 0.69 (0.65, 0.2944) | 3.76 (2.55, 0.1400) |
| Malaria result | |||
| Negative* | |||
| Positive | 0.13 (0.37, 0.7145) | –1.13 (0.54, 0.0371) | –0.47 (2.07, 0.8203) |
| Mother's education | |||
| No education* | |||
| Primary | –0.16 (0.45, 0.7218) | 0.09 (0.51, 0.8626) | –0.07 (0.39, 0.858) |
| Secondary | –1.11 (1.38, 0.4221) | –0.43 (1.05, 0.6815) | –18.0 (109.6, 0.8692) |
| Higher | 0.11 (4.01, 0.9775) | 0.04 (4.71, 0.9926) | 24.6 (275.0, 0.9286) |
| Household's wealth | |||
| Poor* | |||
| Middle | –1.28 (0.64, 0.0466) | –0.18 (0.41, 0.6540) | –5.74 (3.32, 0.0835) |
| Rich | –0.91 (0.71, 0.2009) | –0.41 (0.73, 0.5749) | 19.4(109.5, 0.8597) |
| Place of residence | |||
| Urban* | |||
| Rural | –0.30 (0.82, 0.7105) | 0.33 (0.83, 0.6906) | 18.1 (109.5, 0.8684) |
| Birthorder | 0.08 (0.07, 0.2329) | 0.01 (0.07, 0.9122) | 0.22 (0.33, 0.5109) |
Effects of child characteristics on joint underweight and wasting upon fitting a bivariate logistic regression model to 2024MDHS data.
SE, standard error; (*), reference category.
4 Discussion
This article investigated the risks of multiple undernutrition conditions in a under-five child in Malawi. This was analyzed using a trivariate logistic regression model applied to the 2024 Malawi demographic and health survey data. The findings revealed strong evidence of risk of concurrent stunting and underweight in a child, but weak evidence of joint stunting and wasting, as well as underweight and wasting risks. Cases of paired stunting and underweight were explained by more independent variables involved compared to those of stunting and wasting, or underweight and wasting. Multiple child undernutrition outcomes in a child were expected, as the conditions relate to food intake in the child, making it unlikely for a patient to have one condition only and not the other (, , ).
The likelihood of joint stunting and underweight in children was low among children aged 12 months and older, those with preceding birth interval of 24 months and above, children from middle income or rich families, and those residing in rural areas. After breastfeeding cessation, older children in developing countries often experience periods of nutritional deprivation (), which adversely affects their growth and development. Thus, the observed low risk of stunting and underweight in children over 12 months old is unexpected result. A long and adequate preceding birth interval allows the mother to provide better healthcare, attention, and nutrition for the youngest child at home, which may likely reduce the chances of stunting and underweight in that child (). This also enables proper adherence to exclusive and adequate breastfeeding and hence reducing chances of stunting and underweight (, ). Children from wealthier households are food secure, which is why they have low chances of being stunted and underweight ().
Furthermore, children living in rural areas usually have access to affordable food as Malawi's economy is agrarian and heavily supported by the rural population (). So, it was not surprising in this study to observe low chances of joint stunting and underweight in children residing in rural areas. The findings also indicated that the chances of stunting and underweight increased with increasing birth order. Higher birth order is associated with thinness, probably due to dampened maternal excitement and attention which might reduce dietary care that benefits first-born and second-born children (, , ).
The study had approximately one-fifth of the sample with missing data on both outcome and explanatory variables, and the corresponding cases were dropped to have a complete case analysis. This reduced the sample size and might have biased the estimates of standard errors and p-values in this study (). The VGAM package of R software that was used to process the data could only produce estimates up to two paired outcomes, and not three as planned (). This limited the scope of the estimations in this study, despite availability of statistical methods.
5 Conclusion
This research presents evidence of common risk factors and the simultaneous occurrence of child undernutrition indicators, specifically low height-for-age (stunting) and low weight-for-age (underweight). It emphasizes the necessity of using statistical methods that can analyse these indicators simultaneously. By applying a trivariate logistic regression to the 2024 Malawi Demographic and Health Survey data, the study examined common risk factors associated with concurrent stunting and underweight, stunting and wasting, and underweight and wasting. It was found that the risk of stunting and underweight was low in children older than 12 months, those from wealthier families, and those who resided in rural areas. The risk was higher in children with higher birth order. The risk of joint stunting and wasting or underweight and wasting could not be much explained by the variables involved in the modeling. Given that health outcomes can be paired for an individual child, this study recommends utilizing statistical methods that analyze these outcomes simultaneously rather than sequentially. This approach has the potential to reveal common risk factors related to multiple health issues, facilitating the development of health policies that address all related concerns collectively. Future research could combine these multivariate regression methods with techniques designed to handle missing data.
Statements
Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: https://www.dhsprogram.com/data/dataset/Malawi_Standard-DHS_2024.cfm?flag=1.
Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements, since this study used secondary data collected by another entity. Written informed consent from the [patients/ participants OR patients/participants legal guardian/next of kin] was not required to participate in this study in accordance with the national legislation and the institutional requirements, as this was secondary data analysis.
Author contributions
TK: Conceptualization, Methodology, Investigation, Validation, Supervision, Writing – review & editing, Resources, Visualization, Software, Formal analysis, Data curation, Writing – original draft, Project administration, Funding acquisition.
Funding
The author(s) declared that financial support was not received for this work and/or its publication.
Acknowledgments
The author is grateful to the National Statistical Office of Malawi and the Measure DHS program for the data that were used in this study.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2026.1848833/full#supplementary-material
References
1.
PasrichaSRBiggsBA. Undernutrition among children in south and south-east Asia. J Paediatr Child Health. (2010) 46:497–503. doi: 10.1111/j.1440-1754.2010.01839.x
2.
Dapi NzefaLMonebenimpF. Äng C. Undernutrition among children under five in the Bandja village of Cameroon, Africa South. Afr J Clin Nutr. (2019) 32:46–50. doi: 10.1080/16070658.2018.1448503
3.
SaundersJSmithTStroudM. Malnutrition and undernutrition. Medicine. (2011) 39:45–50. doi: 10.1016/j.mpmed.2010.10.007
4.
KassieGWWorkieDL. Determinants of under-nutrition among children under five years of age in Ethiopia. BMC Public Health. (2020) 20:1–11. doi: 10.1186/s12889-020-08539-2
5.
De LucaAPatelMManthaOPerettiNHankardRePINUT studygroupet al. Promoting the awareness of hospital malnutrition in children: ePINUT 10th anniversary in 2020. Nutr Clin Metab. (2021) 35:85–92. doi: 10.1016/j.nupar.2021.01.001
6.
PermatasariTAEChadirinY. Assessment of undernutrition using the composite index of anthropometric failure (CIAF) and its determinants: a cross-sectional study in the rural area of the Bogor District in Indonesia. BMC Nutr. (2022) 8:133. doi: 10.1186/s40795-022-00627-3
7.
DuranIMartakisKRehbergMSemlerOSchoenauE. Anthropometric measurements to identify undernutrition in children with cerebral palsy. Dev Med Child Neurol. (2019) 61:1168–74. doi: 10.1111/dmcn.14225
8.
AlebelAWagnewFTesemaCKibretGDPetruckaPEshiteS. Effects of undernutrition on survival of human immunodeficiency virus positive children on antiretroviral therapy. Ital J Pediatr. (2018) 44:1–10. doi: 10.1186/s13052-018-0472-2
9.
GeorgiadisAPennyME. Child undernutrition: opportunities beyond the first 1000 days. Lancet Public Health. (2017) 2:e399. doi: 10.1016/S2468-2667(17)30154-8
10.
YisakHGobenaTMesfinF. Prevalence and risk factors for under nutrition among children under five at Haramaya district, Eastern Ethiopia. BMC Pediatr. (2015) 15:1–7. doi: 10.1186/s12887-015-0535-0
11.
AlemayehuMTinsaeFHaileslassieKSeidOGebregziabherGYebyoH. Undernutrition status and associated factors in under-5 children, in Tigray, Northern Ethiopia. Nutrition. (2015) 31:964–70. doi: 10.1016/j.nut.2015.01.013
12.
HallJWaltonMVan OgtropFGuestDBlackKBeardsleyJ. Factors influencing undernutrition among children under 5 years from cocoa-growing communities in Bougainville. BMJ Glob Health. (2020) 5:e002478. doi: 10.1136/bmjgh-2020-002478
13.
NgassaABMerikiHDMbangaCMNzefaLDMbhenyaneXTambeAB. Key predictors of undernutrition among children 6-59 months in the Buea Health District of the Southwest region of Cameroon: a cross sectional community-based survey. BMC Nutr. (2022) 8:148. doi: 10.1186/s40795-022-00646-0
14.
ChowdhuryMRKKhanHTRashidMKabirRIslamSShariful IslamMet al. Differences in risk factors associated with single and multiple concurrent forms of undernutrition (stunting, wasting or underweight) among children under 5 in Bangladesh: a nationally representative cross-sectional study. BMJ Open. (2021) 11:e052814. doi: 10.1136/bmjopen-2021-052814
15.
AnikAIChowdhuryMRKKhanHTMondalMNIPereraNKKaderM. Urban-rural differences in the associated factors of severe under-5 child undernutrition based on the composite index of severe anthropometric failure (CISAF) in Bangladesh. BMC Public Health. (2021) 21:2147. doi: 10.1186/s12889-021-12038-3
16.
FagbamigbeAKandalaNBUthmanA. Demystifying the factors associated with rural-urban gaps in severe acute malnutrition among under-five children in low-and middle-income countries: a decomposition analysis. Sci Rep. (2020) 10:11172. doi: 10.1038/s41598-020-67570-w
17.
KaombeTMBandaJCHamuzaGAMuulaAS. Bivariate logistic regression model diagnostics applied to analysis of outlier cancer patients with comorbid diabetes and hypertension in Malawi. Sci Rep. (2023) 13:8340. doi: 10.1038/s41598-023-35475-z
18.
KangYKimJ. Risk factors for undernutrition among children 0-59 months of age in Myanmar. Matern Child Nutr. (2019) 15:e12821. doi: 10.1111/mcn.12821
19.
CorsiDJ. Mejía-Guevara I, Subramanian S. Risk factors for chronic undernutrition among children in India: estimating relative importance, population attributable risk and fractions. Soc Sci Med. (2016) 157:165–85. doi: 10.1016/j.socscimed.2015.11.014
20.
JeyakumarANikamSNayakS. Prevalence and risk factors of undernutrition among children less than 2 years in urban slums of Pune, Maharashtra, India. Ecol Food Nutr. (2019) 58:456–69. doi: 10.1080/03670244.2019.1613985
21.
National Statistical Office (NSO) Utica International and ICF. Malawi Demographic and Health Survey 2024: Final Report. Zomba; Columbia, MD; Rockville, MD (2025).
22.
GariTLohaEDeressaWSolomonTLindtjørnB. Malaria increased the risk of stunting and wasting among young children in Ethiopia: results of a cohort study. PLoS ONE. (2018) 13:e0190983. doi: 10.1371/journal.pone.0190983
23.
DaiB. Multivariate Bernoulli distribution models (Doctoral dissertation). University of Wisconsin-Madison, Madison, WI, United States.
24.
IslamMA. A trivariate Bernoulli regression model. Cogent Math Stat. (2018) 5:1472519. doi: 10.1080/25742558.2018.1472519
25.
KaombeTM. Properties of model errors in logistic regression and their application to detect outliers in child mortality study. In: Statistical Modelling for Evidence-Based Public Health: Case Studies in HIV/AIDS Epidemiology and Child and Maternal Health.Cham: Springer Nature Switzerland (2025). p. 247–81. doi: 10.1007/978-3-031-97281-2_12
26.
KaombeTM. A bivariate Poisson regression to analyse impact of outlier women on correlation between female schooling and fertility in Malawi. BMC Women's Health. (2024) 24:55. doi: 10.1186/s12905-024-02891-w
27.
YeeTW. VGAM family functions for Bivariate Binomial Responses (2008).
28.
MyattMKharaTSchoenbuchnerSPietzschSDolanCLelijveldNet al. Children who are both wasted and stunted are also underweight and have a high risk of death: a descriptive epidemiology of multiple anthropometric deficits using data from 51 countries. Arch Public Health. (2018) 76:28. doi: 10.1186/s13690-018-0277-1
29.
WellsJCBriendABoydEMBerkelyJAHallAIsanakaSet al. Beyond wasted and stunted—a major shift to fight child undernutrition. Lancet Child Adolesc Health. (2019) 3:831–4. doi: 10.1016/S2352-4642(19)30244-5
30.
NigatuDAzageMMotbainorA. Effect of exclusive breastfeeding cessation time on childhood morbidity and adverse nutritional outcomes in Ethiopia: analysis of the demographic and health surveys. PLoS ONE. (2019) 14:e0223379. doi: 10.1371/journal.pone.0223379
31.
NtambaraJZhangWQiuAChengZChuM. Optimum birth interval (36-48 months) may reduce the risk of undernutrition in children: a meta-analysis. Front Nutr. (2023) 9:939747. doi: 10.3389/fnut.2022.939747
32.
SyedaBAghoKWilsonLMaheshwariGKRazaMQ. Relationship between breastfeeding duration and undernutrition conditions among children aged 0-3 Years in Pakistan. Int J Pediatr Adolesc Med. (2021) 8:10–7. doi: 10.1016/j.ijpam.2020.01.006
33.
RochaHACorreiaLLLeiteÁJRochaSGMachadoMMCamposJSet al. Undernutrition and short duration of breastfeeding association with child development: a population-based study. J Pediatria. (2022) 98:316–22. doi: 10.1016/j.jped.2021.07.003
34.
WoldeMBerhanYChalaA. Determinants of underweight, stunting and wasting among schoolchildren. BMC Public Health. (2015) 15:8. doi: 10.1186/s12889-014-1337-2
35.
LiHYuanSFangHHuangGHuangQWangHet al. Prevalence and associated factors for stunting, underweight and wasting among children under 6 years of age in rural Hunan Province, China: a community-based cross-sectional study. BMC Public Health. (2022) 22:483. doi: 10.1186/s12889-022-12875-w
36.
YuTChenCJinZYangYJiangYHongLet al. Association of number of siblings, birth order, and thinness in 3-to 12-year-old children: a population-based cross-sectional study in Shanghai, China. BMC Pediatr. (2020) 20:367. doi: 10.1186/s12887-020-02261-z
37.
HeldLOttM. How the maximal evidence of p-values against point null hypotheses depends on sample size. Am Stat. (2016) 70:335–41. doi: 10.1080/00031305.2016.1209128
Summary
Keywords
joint child undernutrition, risk factors, stunting, trivariate logistic regression, underweight, wasting
Citation
Kaombe TM (2026) Concurrent child undernutrition indicators in Malawi: application of trivariate logistic regression. Front. Nutr. 13:1848833. doi: 10.3389/fnut.2026.1848833
Received
06 April 2026
Revised
06 May 2026
Accepted
08 May 2026
Published
29 May 2026
Volume
13 - 2026
Edited by
Yoseph Leonardo Samodra, National Taiwan University, Taiwan
Reviewed by
Muh. Arfandi, Universitas Mulawarman, Indonesia
Ronaldo Theodorus Azikin Japari, Hainan Medical University, China
Updates
Copyright
© 2026 Kaombe.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Tsirizani Mwalimu Kaombe, tkaombe@unima.ac.mw
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.