Data

The study examined the secondary data obtained from the Tanzania Demographic and Health Survey and Malaria Indicator Survey (TDHS-MIS) 2022 dataset. The survey was a cross-sectional conducted by the Tanzania National Bureau of Statistics in collaboration with other government partners. The survey used a multi-stage sampling design, where the first stage involved selecting sample points (clusters), consisting of enumeration areas (EAs) delineated for the 2012 Tanzania Population and Housing Census (PHC). A total of 629 clusters were selected. Among the 629 EAs, 211 EAs were from urban areas and 418 EAs were from rural areas. In the second stage, 26 households were to be systematically selected from each cluster, for a total anticipated sample size of 16,354 households for the 2022 TDHS-MIS. The survey involved 4,935 children aged 0–59 months and their mothers or caregivers. The present study was restricted to only 2,158 children aged 0–23 months and their anthropometric data on weights. The weight measurements of children were collected by trained enumerators using a SECA Uni scale with a precision of 100 g.

The child was considered underweight/stunting/ wasting if his WAZ/HAZ/WHZ score was below or equal to − 2 SD, and Normal if he/she WAZ/HAZ/WHZ greater than -2. The calculated WAZ, HAZ, and WHZ were based on 2006 WHO standards (WHO,2006).

Inclusion and exclusion criteria

Children were included if their caregivers provided complete information during the TDHS data collection. Parental or guardian consent was obtained for all participants. Children who had incomplete information or were two years or older were excluded.

Study variablesOutcome variable

In this study, the primary outcome variable is "Coexisted undernutrition," which is derived from three child anthropometric variables: stunting (height-for-age z-scores), underweight (weight-for-age z-scores), and wasting (weight-for-height z-scores). Each of these three variables was dichotomized into two categories: "0" for normal and "1" for the presence of the condition (stunting, underweight, or wasting).

These dichotomized variables were then summed to create the "Coexisted undernutrition" variable, resulting in a discrete score ranging from 0 to 3:

0 = the child does not suffer from any of the three conditions (normal).

1 = The child suffers from one of the three conditions (single burden).

2 = The child suffers from two of the three conditions (double burden).

3 = The child suffers from all three conditions of undernutrition (coexistence).

For analytical purposes, the "Coexisted undernutrition" variable was further categorized into three response levels: However, from a clinical perspective, children with one or two dimensions of malnutrition often require similar interventions and exhibit comparable health outcomes.

0 = normal (none of the conditions).

1 = single or double burden (one or two conditions).

3 = coexistence of all three conditions.

From a clinical perspective, children with one or two dimensions of malnutrition often require similar interventions and exhibit comparable health outcomes.

Given this outcome variable structure, we employed a multinomial logistic regression model to examine the associations between various predictors and the three levels of "Coexisted undernutrition." The reference category for the multinomial regression was the "normal" group (score 0).

Independent variables

The independent variables are divided into three groups: Child, Mother, and Household characteristics. Children's characteristics include the following: Sex of a child (Male, Female), Age of a child (0–5, 6–11, 12–17, 18–23 months), birth weight of a child (less than 2501g, 2501-4000g, greater than 4000g), size of a child (Small/less than 2.5kg, average/2.5kg-3.5kg, large), ever experienced fever/diarrhea/received vaccination 2 weeks prior the survey (Yes, No), breast-feeding status (ever, not currently breast-feeding, never breastfeeding, still breastfeeding), birth order (1st, 2–4, 5 and above). Also, mother characteristics include the following: Mother’s age (< 25, 25–29, 30–34, 35–49 years), Highest education level (No education, primary, secondary and above), Marital status (single, married, divorced/separated/widow), Mother working status (Yes, No), Number of ANC visit (< 4, > = 4 visits). The study also includes household characteristics: sex of household head (male, female), type of place of residence (rural, urban), source of drinking water (improved, unimproved); type of toilet facility (improved, unimproved, open defecation), electricity source (Yes, No) and household wealth (poor, middle and richer).

Model

A multinomial regression model was run to identify the risk factors associated with the coexistence of three child anthropometric measurements (stunting, wasting, and underweight) among children aged between 0–23 months.

Let \(_\) Be the categorical outcome variable representing the nutrition status, taking values in the set (stunting, wasting, and underweight). The examined model is shown below:

$$P\left(_=j\right)=\frac^_+_}_+_}_+_}_}}^^_+_}_+_}_+_}_}}$$

Here:

\(P\left(_=j\right)\) Is the probability of observation I belong to category j.

\(__, _, _\) Are the parameters specific to category j.

\(}_, }_. }_\) Are the values of the dimensions (wasting, stunting, and underweight) for observation “i”?

\(K\) Is the total number of categories (in this case, 3)

Data analysis

The study employed both descriptive and inferential statistical methods for data analysis, using STATA version 16. To assess the association between the coexistence of undernutrition and various risk factors, we first conducted a multicollinearity diagnosis on the independent variables. Collinearity tests included variance inflation factors (VIF), square VIF, tolerance, and R2. These tests aimed to verify that the independent variables were indeed “independent,” meaning they were not highly correlated and did not share variance in explaining the outcome variable. Results from the multicollinearity tests (mean VIF = 1.59; maximum VIF = 3.53; minimum VIF = 1.01; see Appendix 1) confirmed that collinearity was not a concern.

Following this, bivariate analysis identified explanatory variables significantly associated with the coexistence of undernutrition at a significance level of α = 5%. These statistically significant variables were then included in a multinomial logistic regression model to investigate the factors associated with the coexistence of the three undernutrition indicators (stunting, wasting, and underweight) among Tanzanian children aged 0–23 months. In the multinomial model, normal nutritional status served as the reference category, while the response levels included single or double burden and coexistence of all three dimensions.