This was a cross-sectional, retrospective study. Medical records from Mbabane Government Hospital and Piggs Peak Government Hospital were used to extract data for TB cases (Figure 1). These hospitals were selected because they are the main referral facilities for their respective regions and maintain complete paediatric TB records. Health centres do not routinely maintain comparable admission registers.

A purposive, feasibility-based sampling technique was employed for the questionnaire component of the study (Figure 1). Probability or random sampling was not employed since the medical records are retrospective in nature and facility-based and the sampling frame was limited to children with confirmed TB diagnoses recorded at Mbabane Government Hospital and Piggs Peak Government Hospital who had complete medical records and available caregiver contact information. Incomplete contact details, loss to follow-up, and caregiver unavailability further constrained the use of random sampling methods.

A pilot study with 12 participants was conducted to guarantee the reliability of the study questionnaires. The participants were asked the same questions that were asked to study participants on one day and then after a day, the same 12 subjects were asked the same questions again. The researcher then checked whether they answered the same answers and any questions where answers were confusing or inconsistent were revised.

Proper and inclusive response options were made available for each question to be certain that they cover every possible response needed to answer all research questions. The content of the questions was checked against the reviewed literature and the research questions. In addition, the researcher made sure that the language used when writing questions was clear and unambiguous to minimise misinterpretation and response errors. The instrument was piloted to check that all questions were valid in relation to the study population.

The inclusion criteria were parents/guardians/caregivers of children below the age of 15 years diagnosed with tuberculosis in the Hhohho region, Eswatini; parents/ guardians/caregivers of children diagnosed with tuberculosis between 2022 and 2023; and parents/guardians/caregivers of children who had adequate information on the child and household conditions. Exclusion criteria were parents/guardians/caregivers with children above 15 years old; parents/guardians/caregivers of children below the age of 15 years who were not diagnosed with tuberculosis between 2022 and 2023 in the Hhohho region; and participants who did not have adequate information on the child and household conditions.

The population size is the total population of children under 15 years old in the Hhohho region, which is 115 642 (Eswatini Statistics, 2023). Using EPINFO 7.2, with an estimated 2023 population of 115 642 in the Hhohho region of Eswatini, the acceptable error margin is 5% (95% confidence), and with one cluster the estimated sample size at 95% confidence level is N = 383 (Table 1).

In this study, the actual sample size was 50 participants, determined based on the number of cases with available contact information identified during the review of hospital registers at Mbabane Government Hospital and Piggs Peak Government Hospital. The 50 participants were a total of all cases with available contact information from the number of cases identified during medical records review at Mbabane Government hospital and Piggs Peak Government hospital.

For review and approval, the research proposal was sent to the Higher Degrees and Research Ethics Committee and the Departmental Research Committee of the University of Johannesburg. Approval was given from the Eswatini Health and Human Research Review Board to conduct the study (EHHRRB 105/2024). Participants were made aware of the purpose of the study as well as the requirements for their participation. We informed them that they could withdraw from the study at any time without facing any consequences and that participation was completely voluntary. Informed consent forms were given to participants who agreed to participate in the study, asking them to sign them to confirm their participation.

In the analysis a social vulnerability index (SVI) was computed. An SVI is a composite measure designed to capture multiple dimensions of social and environmental disadvantage that may predispose individuals or households to adverse health outcomes. This approach is grounded in the US CDC framework and was adapted using local relevance, as guided by the (8). model, which applied an SVI to the South African context. The objective was to use available variables from our dataset to calculate an SVI score and investigate whether higher vulnerability correlates with TB symptom prevalence.

From the dataset, 13 variables identified as directly relevant to social and environmental vulnerability were selected. These variables were unemployment (caregiver), low education level (primary or no education), food insecurity, household crowding (>4 people), lack of basic sanitation (no flush toilet), unsafe water source (river, well, borehole), stored water in large containers, condensation/mould, dust exposure, lack of windows (poor ventilation), indoor smoking, presence of animals/pets indoors, and polluting fuel use (firewood, paraffin). Each variable was recoded as a binary indicator (1 = vulnerability present, 0 = absent). The 13 binary indicators were added for each participant to obtain a total SVI score ranging from 0 to 13. The threshold for high vulnerability was set at SVI Score ≥ 7, based on the median distribution and precedent from similar studies. For each variable percentile ranks were calculated. A new binary classification was created: socially vulnerable expanded = 1 if score ≥ 7 and socially vulnerable expanded = 0 if score < 7. Out of 50 participants: 32 (64%) were classified as socially vulnerable (SVI ≥ 7) and 18 (36%) were classified as not socially vulnerable (SVI < 7).

To examine the associations between vulnerability indicators and TB symptoms more formally, a Fisher's Exact Test was conducted, two sided, alpha 0.05, in IBM SPSS Statistics version 29. This test compared TB symptom status (Yes vs. No) across each indicator: caregiver unemployment, household crowding, defined as 5 or more people per household, polluting cooking fuel, absence of basic sanitation, no flush toilet, indoor smoking, and poor ventilation, fewer openable windows.

The sample consisted of 50 children. A total of 46% of participants were between 0 and 4 years of age, 30% were aged 5 to 9 years, and 24% were between 10 and 14 years. This skew toward younger children reflects higher biological susceptibility to TB in early childhood. The gender distribution showed a predominance of males (60%) compared to females (40%).

Regarding caregiver employment, a higher proportion of caregivers in the vulnerable group were unemployed (65.2%) compared to 48.1% in the non-vulnerable group. Similarly, 43.5% of vulnerable caregivers had only primary education, and only 13% had tertiary education, in contrast to 40.7% of caregivers in the non-vulnerable group with tertiary education. These disparities suggest a link between higher social vulnerability and poorer socio-economic status.

Overall, 88% of the children presented with TB symptoms, with symptom prevalence being slightly higher among the non-vulnerable group (88.9%) compared to the vulnerable group (87.0%), possibly due to random variation or unmeasured protective factors. Table 2 presents a composite summary of these characteristics.

To provide epidemiological context for the vulnerability and symptom analyses, the prevalence of tuberculosis among children younger than 15 years was assessed using outpatient and inpatient TB registers at Mbabane Government Hospital and Piggs Peak Government Hospital for the years 2022 and 2023. The results, summarized in Table 3, show the proportion of TB cases among all paediatric (under 15 years of age) hospital attendees.

Across both facilities and years, the overall prevalence of tuberculosis among children under 15 was 0.2%. The prevalence was slightly higher at Mbabane Government Hospital and increased in 2023 compared to 2022. This facility-based prevalence provides important context for interpreting the much higher burden of TB symptoms identified in the subsequent SVI-targeted analysis.

SVI scores in this population revealed a marked concentration of vulnerability. While scores ranged from 0 to 10, the distribution was right-skewed, with most children exhibiting moderate to high cumulative exposure to risk factors. Nearly two-thirds of participants scored 7 or above-the defined threshold for high vulnerability in this analysis.

Figure 2 illustrates the distribution of vulnerability levels among participants, showing a noticeable skew toward higher vulnerability. A significant proportion of children meet 6–9 of the criteria. Additionally, the figure highlights that most participants scored between 6 and 9, with a peak at score of 7. This supports the use of 7 as a cutoff point for defining social vulnerability.

SVI scores in this population revealed a marked concentration of vulnerability. While scores ranged from 0 to 10, the distribution was right-skewed, with most children exhibiting moderate to high cumulative exposure to risk factors. Nearly two-thirds of participants scored 7 or above-the defined threshold for high vulnerability in this analysis (Figure 2).

This figure illustrates the distribution of vulnerability levels among participants, showing a noticeable skew toward higher vulnerability. A significant proportion of children meet 6–9 of the criteria. Additionally, the figure highlights that most participants scored between 6 and 9, with a peak at score of 7. This supports the use of 7 as a cutoff point for defining social vulnerability. These results suggest that the SVI is not only a valid tool for summarising risk, but also a potentially powerful instrument for guiding targeted TB screening and response strategies in resource-constrained settings (9).

A clear disparity emerged in TB symptom prevalence between vulnerable and non-vulnerable children. Among those classified as vulnerable (SVI ≥ 7), 94% exhibited TB symptoms. In contrast, 78% of children with lower vulnerability scores were symptomatic. While TB symptoms were common across both groups, the gap suggests that children exposed to greater cumulative disadvantage are substantially more likely to be symptomatic.

Although overall prevalence was high, the data reveal an intensifying effect of cumulative vulnerability (Figure 3). This distinction underlines the value of the SVI in differentiating levels of risk within a uniformly high-burden population.

When examined across the full range of SVI scores, TB symptom prevalence displayed a non-linear but consistently elevated pattern. Children with mid-range scores (2–6) were universally symptomatic. A modest dip was observed at score 7, where symptom prevalence fell to 77.8%, but rose again at scores 9 and 10.

Figure 4 visualises the percentage of children with and without TB symptoms across different SVI scores (0–10). First it reveals high TB symptom prevalence across all SVI scores: All children with SVI scores from 2 through 6 show 100% prevalence of TB symptoms; and even children with SVI scores of 0 and 1 still show notable symptom rates (83% and 67%, respectively). Secondly, it reveals that there is no linear pattern, but consistently high vulnerability: while one might expect increasing SVI scores to correspond with increasing TB symptoms, the relationship is not strictly linear, but symptom prevalence is consistently high (above 80%) for most scores; and this suggests that even low to moderate social vulnerability may significantly contribute to TB symptom burden. Thirdly, it reveals that score 7 + contains a mix: At SVI Score 7, there's a visible drop in symptomatic percentage (to ∼75%), indicating the presence of asymptomatic children even in vulnerable settings; and this might reflect noise due to small sample size or other mediating factors (e.g., care-seeking behaviour, nutrition, immunity).

Several individual SVI indicators were particularly associated with elevated symptom prevalence (Table 4). These included caregiver unemployment, household crowding, use of polluting fuels, and poor sanitation—each contributing to symptom rates exceeding 90%. Indoor smoking, although less common, was universally associated with TB symptoms in the small subset of children affected.

The high prevalence associated with specific indicators reinforces the rationale for composite vulnerability measurement. These findings support targeting interventions not just based on disease incidence but also on upstream predictors of exposure and delayed care-seeking.

As summarised in Table 5 below, none of the differences reached statistical significance, p greater than 0.05. This is consistent with the small sample size (n = 50) and the uniformly high symptom prevalence across exposure categories, which reduced statistical power. Even so, the descriptive patterns align directionally with known social determinants of childhood TB, indicating that children living in socioeconomically disadvantaged or environmentally constrained households experienced higher symptom burden.

A limitation of the study is that the cross-sectional design precludes establishing causality. Associations observed between vulnerability and TB symptoms cannot confirm temporal or causal relationships. Another limitation of the study was the small sample size. With only 50 participants, the study could not perform multivariable logistic regression or explore subgroup analysis (e.g., by HIV status or age group). Data were collected from the two main referral hospitals for their respective regions and maintain complete paediatric TB records and not other health centres.

This study highlights critical links between social vulnerability and paediatric TB in Eswatini, yet it also underscores several areas where further research is urgently needed. First, longitudinal studies are required to better understand the causal pathways through which social disadvantage, household crowding, poor nutrition, and environmental exposures interact to elevate childhood TB risk. Cross-sectional designs, while informative, cannot fully capture the temporal dynamics of exposure, infection, and progression to active disease.

Future work should refine and validate SVI tools for low-resource African settings. Although the adapted SVI proved useful in quantifying disadvantage, locally contextualised indices—incorporating factors such as informal settlement characteristics, caregiver HIV status, biomass fuel exposure, and food insecurity—may provide greater predictive accuracy. Mixed-methods studies involving community stakeholders could strengthen the cultural relevance and interpretability of these indices. Spatial epidemiological studies integrating geocoded SVI data with environmental, climatic, and housing-quality indicators would help identify high-risk neighbourhoods and micro-environments. Such analyses could guide targeted interventions, including ventilation improvements, social protection programmes, and enhanced TB screening in vulnerable communities.