This study was a secondary analysis of a randomised controlled trial (RCT) titled ‘Sustainable household contact tracing and screening for tuberculosis patients and families’. The RCT (Trial registration number: ISRCTN81659509) aimed to implement and assess the effectiveness of a home-based TB screening strategy delivered at TB patients’ households. In the RCT, the study participants were randomly assigned to¹ a standard of care group that followed standard and/or routine household screening, in which TB index patients were told to bring their household contacts for TB screening to the facility or² the intervention group, called patient-delivered household active case finding for TB (PACTS), which involved the home-based TB screening strategy. Participants assigned to the intervention group were given a screening and/or triage tool and sputum collection pots for each household contact that was reported. The screening tool comprised information such as symptom screening, IPT eligibility for household contacts and sputum collection for the contacts. This screening was carried out by either the index patient or the parent and/or caregiver in the household. The TB screening was performed for both adult and child contacts. All child contacts were followed up at 3 months and 6 months for uptake and completion of IPT.²² Isoniazid preventive therapy completion was ascertained through clinical records, caregiver reports and inspection of medication bottles at 3- and 6-month visits. The study was conducted from May 2014 to November 2015 at Queen Elizabeth Central Hospital (QECH) in Blantyre, Malawi. The QECH is one of the tertiary-level facilities and one of the largest hospitals in Malawi. Most TB cases are diagnosed and managed at this hospital.

The study population for this study was all under-five child contacts identified from TB cases in the RCT, irrespective of the trial arm in which the index case was in the RCT. These children were 128 in total.

The dependent variable of this study was the completion of 6 months of IPT among all child contacts. The treatment outcome variable was coded as binary ‘IPT completed’ and ‘IPT not completed’.

The independent variables were at various levels. Contact characteristics included age, sex and relationship with the household head. Variables assessed also included index patient characteristics such as index case TB type and human immunodeficiency virus (HIV) status, which were used in the same way that they were originally coded in the primary study. Household characteristics such as household head employment, household head education level, health-seeking decision maker and wealth quintile consisted of the following categories in the primary study: least poor, less poor than average, average, poorer than average and poorest. These were re-categorised into three categories: high wealth status, medium wealth status and low-wealth status. Thus, ‘least poor’ and ‘less than poor than average’ were grouped as high wealth status, average was grouped as medium wealth status, and ‘poorer than average’ and ‘poorest’ were grouped as low-wealth status. For health access characteristics, the variables used were distance to the health facility (in kilometres) and the type of contact tracing and screening used. For distance, in the primary study, the distance variable had four categories, namely, ‘less than 1 km’, ‘more than 1 km but less than 5 km’, ‘more than 5 km but less than 10 km’ and ‘more than 10 km’. For the purpose of this study, the variable was categorised into two categories: ‘5 km or less’ and ‘more than 5 km’. The variable type of contact tracing and screening used a binary variable (‘patient-conducted tracing’ and ‘routine contact tracing’).

A descriptive analysis of the distribution of baseline characteristics was performed. For categorical variables, percentages and frequencies were reported. Numerical variables were summarised as means and standard deviations if normally distributed and as median and interquartile range if not normally distributed. The prevalence of completion of IPT was estimated across all explanatory variables. This prevalence was reported as percentages for categorical variables, and t-test and Wilcoxon rank sum test were used for numerical variables in which medians or interquartile ranges (IQR) were compared. Binary logistic regression models were used to investigate factors associated with completion of IPT. Univariable and multivariable analyses were conducted. Unadjusted and adjusted odds ratios and 95% confidence intervals (CI) were reported. For building the multivariable regression model, stepwise regression was used: sex, age of the child and index HIV status were included a priori, and other variables were included if found significant from the univariable logistic regression model. Variables with p-values < 0.20 in univariable analysis or those supported by literature were included in the multivariable model. For handling of missing data, we used complete case analysis.

PACTS trial data were used (data from the primary study) in this study. Permission to use the data was sought and was approved by the principal investigator of the primary study, Dr Kruger Kaswaswa from the Kamuzu University of Health Sciences (KUHES). Data were obtained from the principal investigator of the primary study.

Ethical clearance to conduct this study was obtained from the University of the Witwatersrand Human Research Ethics Committee (No. M230121 MED22-10-172). The primary study was approved by the Institutional Review Board for Dr Kruger Kaswaswa from the Kamuzu University of Health Sciences (KUHES) in Malawi. The primary participants provided consent for the primary study, including the use of anonymised data for secondary data analyses. The data shared were anonymised, and there was no identifying information. This study was a secondary data analysis; thus, the ethics review did not require individual informed consents.

Table 1 shows the baseline characteristics of the child contacts, index case characteristics, household characteristics and health access characteristics. The female child contacts constituted 52.0% of the study population. The median age of the child contacts was 3 (IQR: 1–4), and 71.9% were children of the household head. About two-thirds of the related-index cases had smear-positive TB, and 52.3% were HIV-positive. About half of the child contacts came from households with low-wealth class (52.3%), 73.4% of their household heads were employed, and 53.9% had no school and/or primary level of education. The median number of children under five in the households was two (IQR: 1–3). The majority of the child contacts had their parents (68.8%) as the main healthcare-seeking decision maker, and the median number of people living in households was six (IQR: 4–8). Over four to five (82.0%) of the children’s contacts lived at a distance of 5 km or less from the health facility, and 51.6% were identified through routine contact tracing.

Table 2 presents IPT completion among child contacts by different characteristics. Overall, 45.3% of the child contacts completed IPT. Completion of IPT was high among female child contacts (48.5%) compared to male child contacts (41.9%), but this difference was insignificant (p = 0.46). The child contacts that were younger had a slightly higher completion rate with a median age of 2 (IQR: 1–3), even though it was not statistically significant (p = 0.26). Child contacts who had the head of the household as their parent had a higher completion of 47.8% compared to that for child contacts who were not children of the household head (38.9%) (p = 0.36).

Isoniazid preventive therapy completion was slightly higher for child contacts whose related-index case had smear-negative TB type (46.5%) than (44.7%) for the smear-positive TB (p = 0.85). Isoniazid preventive therapy completion was high among child contacts whose index case was HIV-negative (56.0%) (p = 0.07) compared to that of child contacts whose index case was HIV-positive (38.8%).

Children from high-wealth status households had a significantly higher completion of 64.9% compared to that of children from low-wealth status (34.3%) (p = 0.01). Completion was slightly higher in households in which the household head was employed (46.8%) and had a secondary and/or higher level of education (50.8%). The median number of under-five children in the household who completed IPT was two (IQR: 1–3). Completion was higher among child contacts whose parent was the health-seeking decision maker (53.4%), compared to other relatives (27.5%) (p = 0.01). The median number of people living in the household who completed IPT was five (IQR: 4–6) (p = 0.01).

Child contacts that lived at a distance of 5 km or less from the health facility had significantly higher prevalence of IPT completion (49.5%), compared to those that lived further away (26.1%) (p = 0.04). There was higher completion among child contacts who were identified through the patient-conducted tracing (53.2%) compared to those who were identified through routine contact tracing (37.9%) (p = 0.08).

Table 3 shows the univariable and multivariable analyses for the completion of IPT. For wealth status, the odds of completing IPT among child contacts who came from high wealth status were 4.18 times the odds of completing IPT among child contacts who came from low-wealth status (odds ratio [OR] = 4.18, 95% CI: 1.70–10.25, p = 0.01). The role of the household healthcare decision maker affected IPT completion; the odds of completing IPT among child contacts whose parents were the household health-seeking decision maker were 2.58 times the odds of completing IPT among child contacts whose other guardian was the household health-seeking decision maker (OR = 2.58, 95% CI: 1.12–5.95, p = 0.02). For the number of people living in the household, the odds of completing IPT were 0.81 (OR = 0.81, 95% CI: 0.69–0.96, p = 0.01) for each unit increase in the number of people in the household.

Distance to a health facility was also important. The odds of completing IPT among child contacts whose distance to the health facility was more than 5 km were 0.30 times the odds of completing IPT among child contacts whose distance to the health facility was < 5 km (OR = 0.30, 95% CI: 0.10–0.88, p = 0.05). For the type of contact tracing, the odds of completing IPT among child contacts who were identified through patient-conducted tracing were 2.41 times the odds of completing IPT among child contacts who were identified through routine contact tracing (OR = 2.41, 95% CI: 1.14–5.11, p = 0.02).

On multivariable analysis, wealth status was significantly associated with IPT completion. Higher wealth status was significantly associated with higher completion of IPT among children, in whom the adjusted OR (aOR) for IPT completion was 3.42 (95% CI: 1.19–9.88, p = 0.02) for children coming from households of high wealth status, compared to those coming from the lowest wealth status. A significant association was noted after multivariate analysis between the HIV status of the child contacts’ index case and the completion of IPT; child contacts with HIV-positive index cases had significantly lower odds of IPT completion, with aOR = 0.39 (95% CI: 0.16–0.94, p = 0.02). Household health-seeking decision maker was still significantly associated with completion of IPT, with aOR = 3.17 (95% CI: 1.19–8.42, p = 0.03) for child contacts whose parents were the household health-seeking decision maker, compared to child contacts who had other guardians as the household healthcare decision maker.

After the multivariable analysis, the number of people living in the household was not significantly associated with completion of IPT, with aOR = 0.87 (95% CI: 0.71–1.06, p = 0.61) for each unit increase in the number of people living in the household. The distance to a health facility was also still significantly associated with IPT completion, with aOR = 0.25 (95% CI: 0.07–0.89, p = 0.03) for child contacts living at a distance of more than 5 km, compared to those at a distance of less than 5 km. Similarly, the type of contact tracing was still significantly associated with the completion of IPT, with aOR of 3.13 (95% CI: 1.25–7.84, p = 0.01) for child contacts who were identified through patient-conducted tracing, compared to child contacts who were identified through routine contact tracing.

Although this study provides valuable insights into factors associated with the completion of IPT, the study had some limitations. The primary study had information on uptake and completion of IPT available, but information on actual drug adherence (such as pill count) was not available, which may have resulted in an overestimation of the number of participants who received an effective course of IPT and completed the whole course of IPT. Literature has also found that other relevant factors associated with completion of IPT, such as duration of IPT, presence of side effects, pill burden, pill size, pill taste, health factors (drug stock outs) and direct observed therapy; however, these were not included in this study, because the dataset did not have the variables. In addition, this study used a small sample size, which may have increased the risk of model overfitting. We did not account for household clustering in the model, which could have introduced bias if outcomes were correlated among children from the same household. There is potential for residual confounding given the observational nature of the data for some variables, such as distance to the health facility (if self-reported) and the healthcare-seeking decision maker, which may be subject to measurement error. Furthermore, the use of a complete case analysis to handle missing data may have introduced bias.

Our findings suggest that IPT completion among child contacts is shaped by both health facility access and household factors. Distance to the clinic highlights the need to strengthen peripheral delivery or provide transport support. Especially in rural areas, this can be achieved through community health posts, mobile clinics and outreach clinics. We recommend the need for counselling, awareness, family-centred engagement and/or support and health education for guardians other than parents who make healthcare decisions in the households. This will assist in ensuring that the guardians are empowered and are fully informed about IPT and the importance of IPT completion, thus facilitating the care for the child contacts. Closer integration of IPT with HIV services and offering of targeted adherence support may help to mitigate the challenge of stigma and balance competing health demands when the index case is HIV-positive, thereby improving IPT completion. Supporting patient contact tracing and exploring support targeted for low-wealth status households could also improve IPT uptake and completion. Future studies should investigate other factors that may affect the completion of IPT, for instance, how household dynamics and different community-based delivery models affect IPT completion. This effort will guide more effective policies and practices.