Discussion
Our study showed that integration and decentralisation of child TB services into paediatric healthcare services increased TB symptoms screening and TB diagnostic workup among all child attendees, leading to a marked increase in TB case detection in Cameroon, whereas in Kenya, we did not observe a significant increase in TB case detection.
The study’s finding that integrating and decentralising TB services into child healthcare services has the ability to increase TB case detection and improve treatment outcomes was supported by previous research.13 21 However, the INPUT study represents a pioneering effort in using the stepped-wedge cluster-randomised design to demonstrate the efficacy of integrating paediatric TB services. This robust study design has gained traction in public health research and has proven effective in evaluating multifaceted interventions.22
In analysing the difference in the effect on TB case detection between the two countries, it is crucial to consider the contrasting baseline situations. In Kenya, prior to the study implementation, efforts had already been made at the MoH level to establish decentralised TB care services, resulting in a relatively efficient and established system for TB case identification. This pre-existing infrastructure and decentralised approach likely contributed to the higher case detection rate during the standard of care in Kenya. Conversely, in Cameroon, there had been no attempt to decentralise TB services before the roll-out of the CaP-TB intervention, hence the very low detection rate of TB found during the control phase, which markedly increased during the intervention phase.
These contextual differences across countries highlight the importance of considering the existing healthcare landscape when interpreting the outcomes of multicountry interventions.
Before conducting the study, we discussed the possibility of an initial increase in case finding at study initiation, even before the roll-out of the intervention.15 The increase could be due to the use of new TB registers and monitoring and evaluation forms, which by themselves could raise awareness about paediatric TB, as well as the capture of a reservoir of previously undiagnosed prevalent cases. While the paucity of TB diagnoses in standard of care rules out this hypothesis for Cameroon, we cannot exclude the possibility that this confounding effect may have contributed to decreasing our ability to show a difference in case detection in Kenya.
Another factor that could have impeded our ability to show an effect in Kenya is the occurrence of massive healthcare worker strikes, which significantly impacted the final step when all sites were implementing the CaP-TB intervention. In December 2020, study sites in Kenya faced a severe decline in attendees and were unable to enrol children with presumptive TB due to the strikes. We extended the duration of the last step by a few weeks to approximate a full 4 months of normal activity in these clusters. However, as shown in figure 1, the number of attendees in step 4 was still lower compared with other steps.
Unfortunately, another notable limitation of the INPUT study’s stepped-wedge design was the unforeseen onset of the COVID-19 pandemic during the later phases of the trial, which mainly encompassed the intervention periods. It is reasonable to hypothesise that the impact of the COVID-19 pandemic, weighing heavily on the intervention phases, could have impeded our ability to fully demonstrate the true impact of the intervention.23 We can observe from boxplots (online supplemental material 4) that during phase III, both control and intervention sites appeared to have lower case detection than what was achieved during phases I and II and pre-COVID. Indeed, the COVID-19 pandemic has had a profound disruptive effect on all healthcare services, including TB-related services.24–26
One of the key objectives of the CaP-TB intervention was to provide training to healthcare workers in primary healthcare facilities, enabling them to confidently diagnose and initiate treatment for paediatric TB, even in the absence of bacteriological confirmation. Even though specimen collection was more frequent during CaP-TB, bacteriological confirmation of the TB diagnosis was not more frequent. This study confirms that bacteriological confirmation is rare in children, despite the use of molecular tests and improved sample collection methods.27
In this context, CXR plays a crucial role. In our study, although the proportion of presumptive children undergoing CXR remained consistent between study phases (approximately half of them), the intervention phase saw a substantial increase in children with presumptive TB, elevating the absolute number of CXRs conducted from 65 to 308. During the intervention phase, there were fewer abnormal CXR findings attributed to TB, and conversely, more abnormal CXR findings attributed to other causes than TB, compared with the control phase. This may be due to the broader range of differential diagnoses considered during the intervention because of broader screening for respiratory symptoms, as well as due to the positive impact of the training provided by CaP-TB in improving the interpretation of X-rays, thus increasing their specificity to detect TB. Also, looking at the absolute numbers of CXR findings with abnormal images not consistent with TB (109 in the intervention compared with two in the control), we can expect that the use of CXR, in turn, contributed to a more accurate diagnosis of pulmonary infections in these children. Notably, before TB investigations, only half of the children had received antibiotics, contrary to guidelines recommending a 2-week course for all those with TB signs. Abnormal X-ray images likely prompted appropriate antibiotic courses.
These considerations become particularly relevant considering the new conditional recommendation by the WHO to use treatment decision algorithms based on clinical symptoms and CXR when available and the concern that this approach may lead to overtreating children who do not have TB.8 28 However, given the potential consequences of missing a TB diagnosis and the low frequency of serious adverse reactions to anti-TB treatment in children,29 it is considered reasonable to prioritise high sensitivity and accept a lower specificity26 even though that implies a degree of overtreatment.30
While a larger number of children with presumptive TB were identified during the intervention and underwent TB investigations, the intervention decreased the probability of a child investigated for TB being ultimately diagnosed with TB. This occurred because the increase in the number of children with presumptive TB undergoing TB investigations (denominator) outpaced the increase in actual TB diagnoses (numerator). Although the lower yield of TB diagnosis among those investigated may seem counterintuitive, we view this as a positive outcome of the intervention, suggesting that investigating more children led to more accurate diagnoses and appropriate treatments, even when TB was ruled out. Our study follow-up focused on confirmed TB cases; therefore, we were unable to document the outcomes of children who were not diagnosed with TB. We strongly advocate for studies investigating the outcomes of all children with presumptive TB, irrespective of the final diagnosis. We do believe that increasing the number of children identified with pulmonary symptoms and undergoing TB diagnostic examination should in turn contribute to improving their health outcomes, whether the diagnosis was TB or not.
It is important to acknowledge that the number of children diagnosed with TB in our study was below our expectations, which has limited our power to show significant differences between the intervention and control periods. It was especially the case for treatment outcomes, for which we failed to find a significant association, although the proportion of favourable treatment outcomes rose from 82% to 92% between the control and intervention phases, a 10% increase that can certainly be considered clinically important.
Nevertheless, our findings provide important evidence of the potential benefits of integrating TB services into child healthcare services. In the view of scaling up such interventions, it will be critical to take logistical challenges and cost implications into account. The main logistical challenges identified within CaP-TB were to ensure adequate training and retraining of healthcare workers, including at the low level of care, and to ensure the availability of TB diagnostic tools and medications. Costing considerations, both from a patient and from a health system perspective, as well as cost-effectiveness, have also been studied in INPUT.31 32 This analysis showed that the cost-effectiveness of integrating TB services into child healthcare services depends on baseline service coverage, TB detection and treatment outcomes. These additional findings are crucial for reinforcing the importance of continued investment in TB control efforts in Africa.
link