Improving Tuberculosis Case Detection in Children: Summary of Innovations and Findings From 18 Countries

Abstract

Despite a growing focus on the plight of tuberculosis (TB) among children, 56% of the 1.2 million children who develop TB annually are not detected and notified. TB REACH is a platform of the Stop TB Partnership that supports innovative interventions to improve TB case detection and preventative treatment. We present summary findings from 27 TB REACH-supported projects in 18 countries. Interventions were designed around intensified case-finding approaches (facility-based systematic screening and contact investigation), capacity building (including decentralized care delivery and supported decision-making), and improving diagnostic methods (ie, introduction of alternative respiratory specimens and new tools to aid the diagnosis). These interventions were evaluated on how they worked to identify children with TB, prevent further transmission of TB among children, and strengthen the health system involved with childhood TB care. Overall, 13 715 children were detected with TB, improving case notifications by 34%. In addition, nearly 5000 eligible contacts were enrolled on TB preventive treatment through these interventions. Focusing efforts and funding on childhood TB can produce marked improvements in case detection.

Despite being treatable and preventable, tuberculosis (TB) kills more people every year than any other infectious disease, replaced transiently by Covid-19 [1]. The main reason for the mortality is the gap between incident TB and those who are detected, notified, and treated. Globally, the detection gap has been 30%–40% among adults; however, among children and adolescents the gap is even greater. The World Health Organization (WHO) estimates that 1.2 million children (12% of all cases) developed TB in 2020, yet only 44% of them were diagnosed, and 16% died [1]. Other modeling suggests up to 25% of children who develop TB die from the disease [2, 3]. The diagnostic gap for children under 5 years of age is greater still at 72%, and they have the highest risk to develop severe forms of TB, often leading to disability and death [2–5].

CHALLENGES IN CHILDHOOD TB DETECTION

As detailed in this supplement and elsewhere, a complex set of factors contributes to this large proportion of children with TB who are missed [1, 6, 7]. Factors include lack of awareness among health care staff and in the communities, absence of integrated approaches to screen children for TB, nonspecific presenting symptoms, the difficulty children experience in expectorating sputum, nonavailability of tools with adequate diagnostic accuracy, scarcity of diagnostic services for children at decentralized settings, and low-bacterial load in respiratory secretions from children with TB. Community and/or primary care facilities in high-burden settings are often operated by less-experienced physicians who have limited access to experienced clinicians with whom to discuss the clinical presentations of sick children. Many physicians are also reluctant to initiate children on 6–12 months anti-TB treatment without confirmatory diagnosis. Despite numerous pleas for more attention to TB among children [8–10], progress globally has been slow.

TB REACH OVERVIEW

The Stop TB Partnership’s TB REACH is a grant-making platform designed to improve TB case detection. The initiative receives foundational support from Global Affairs Canada but is supported by multiple donors. TB REACH funds local innovators with grants up to USD 1 million through competitive calls for proposals (3%–7% grant award rate) to use innovative approaches and technologies to find and treat people with TB disease, drug-resistant TB, and/or TB infection. Projects are focused on health care delivery as opposed to research, but undergo rigorous monitoring and evaluation (M&E) [11] which aims to document experiences and link impactful interventions to long-term sustainable funding [12]. TB REACH can fund new, untested ideas, and approaches which donors like Global Fund cannot given the restrictions to use only WHO-recommended tools and approaches.

There have been several reviews of childhood TB burden [13], global and national policies [14], and some on different risk factors and outcomes [15–17], but the question of how to improve case detection among children has not been included in these studies. Several thematic reviews of TB REACH projects have been published [18–20], but none on childhood TB. Here, we present key findings from TB REACH projects focused on childhood TB that may serve as models for adoption or scale-up to improve childhood case detection in high TB burden countries.

METHODS

TB REACH has completed 7 funding waves to date, with waves 8, and 9 ongoing at the time of writing. TB REACH uses an online grant management system to collect detailed technical and financial data quarterly from all projects and encourages its grantees to review, analyze, and publish their results. M. T. R. and J. C. reviewed the reports from Wave 1 to 7 for projects focused on childhood TB, as well as published results from TB REACH projects, to document and group interventions based on the approaches used. Our goals were to report on the results of the TB screening cascade for children, document the additional number of children with TB that were found through the interventions, and distill key findings from different types of interventions to inform future approaches for improving childhood TB detection.

RESULTS AND LEARNINGS FROM TB REACH INTERVENTIONS

We reviewed 297 projects from 7 funding waves, and dozens included components of childhood TB interventions. We identified 27 projects in 18 countries (Table 1) focused on childhood TB or included children in interventions and reported child TB-specific data. Among those 27 projects, 18 focused exclusively on increasing childhood TB detection/prevention and 9 included children and adults. Only projects with age-disaggregated screening data were included in the cascade analysis and the additionality calculations [11], limiting that number to 18. However, all 27 projects were reviewed to extract learnings. Projects had a wide range of scope and reach, and interventions were grouped into the following areas: facility-based (both out-patient and in-patient) systematic screening among children, systematic contact investigation, use of new tools (ie, Xpert MTB/RIF and Ultra), use of alternative specimens (stool and gastric aspirate), decentralized care delivery models, supported diagnostic decision-making (video consultation, medical board, and mentorship program), and TB preventive treatment (TPT). We further grouped the interventions and findings in 3 main areas: case finding based on screening approaches, diagnostics or diagnostic methods, and capacity-building measures to improve TB detection. Most of the projects provided diagnostic and transportation support to facilitate and provide quality to the evaluation and treatment process (including adherence and completion); see Table 1 for an overview of interventions and relevant references.

Table 1.

TB REACH-Supported Interventions

Projects	Country	Interventions
		Facility-based ­Systematic Screening	Systematic ­Contact Investigation	Introduction of New Tools (eg, Xpert MTB/RIF Ultra)	Use of Alternative Respiratory Specimen	Decentralized Care Delivery Model	Supported Decision-making (Video Consultation, Medical Board)	TB Preventive Treatment
Afghan Community Research & Empowerment Organization for Development (1 and 2)	Afghanistan	☑		Digital CXR, LED microscope
Baylor College of Medicine Children’s Foundation	Swaziland	☑	☑
Baylor College of Medicine	Swaziland		☑					☑
Center for Health Solutions[21, 22]	Kenya	☑	☑		☑	☑		☑
Centre For Infectious Disease Research in Zambia	Zambia	☑	☑	AI algorithm for CXR among children		☑
Friends Affected and Infected Together in hand/Aastha	Nepal	☑	☑			☑
German Leprosy and TB Relief Association	Nigeria	☑	☑
Health and Development Alliance	Cambodia	Community-based screening	☑
IRD Bangladesh (1 and 2)	Bangladesh	☑	☑	AI algorithm for CXR among children	☑	☑	☑	☑
icddr,b	Bangladesh			Xpert MTB/RIF Ultra	☑
IRD South Africa	South Africa	☑	☑					☑
KNCV	Ethiopia		☑					☑
Medical Research Council [23]	Gambia		☑					☑
National TB Program & Baylor College of Medicine	Swaziland	☑	☑
The Indus Hospital [24]	Pakistan	☑	☑		☑	☑	☑	☑
The Aurum Institute	South Africa	Door to door screening	☑
Wellbody Alliance	Sierra Leone	Door to door screening	☑	Xpert MTB/RIF

Abbreviations: AI, Artificial Intelligence; CXR, Chest X-ray; LED, Light Emitting Diode; MTB, Mycobacterium tuberculosis; RIF, Rifampicin.

Overall, the 18 projects reported screening about 2.76 million children, of whom 13 715 were diagnosed with TB. There was great variation in scale as the largest projects screened 803 669 children (Afghanistan) and 1 034 729 children (Bangladesh), while more focused projects reported screening 4067 children (Zambia). Similarly, a project from Bangladesh identified 2692 children with TB, while some projects detected fewer than 50 children.

TYPES OF CASE-FINDING APPROACHES

Household Contact Investigation

Contact investigation is often the main approach for childhood TB interventions, and 15 TB REACH projects employed it. Overall, 134 854 child contacts of index infectious pulmonary TB patients were verbally screened (Table 2). Children with presumptive TB among the contacts were then linked to the health facilities for diagnostic evaluation, and 1921 (1.4%) children were identified with TB disease. One project evaluated a quantified TB exposure tool using a 10-point scoring system, and the tool was found effective in predicting TB among household contacts [25]. Notably, results from the Gambia showed that symptom screening alone would miss almost half of the TB cases among exposed children, so additional tools should be used [23].

Table 2.

Yield From Systematic Screening at Health Facilities and Household Contact Investigation (Data From 18 TB REACH Projects)

Indicators	Facility-based Screening (n = 15)	Household Contact Investigation (n = 15)
	n (%)	n (%)	Total
Children verbally screened	2 625 079	134 372	2 759 933
Children with presumptive TB	164 191 (7%)	20 129 (15%)	184 417 (7%)
Children tested/investigated for TB	121 977 (74%)	16 461 (82%)	138 536 (75%)
Children diagnosed with bacteriologically positive pulmonary TB	2680 (2.2%)	214 (1.3%)	2909 (2.1%)
Children diagnosed with all-forms TB	11 794 (10%)	1913 (12%)	13 715 (10%)
Children with all-forms TB started on treatment	11 360 (96%)	1810 (95%)	13 180 (96%)
Number needed to screen to find one child with TB*	223 (Range: 9–2153)	70 (Range: 3–721)	201
Number tested/evaluated to find one child with TB (NNT)†	10 (Range: 2–89)	9 (Range: 2–31)	10

NNS = Children verbally screened/Children diagnosed with all-forms TB.

NNT = Children investigated for TB/Children diagnosed with all-forms TB.

In addition to prospective contact investigations, some projects reported conducting reverse contact investigation or source case investigation (among household contacts of newly diagnosed children with TB) and retrospective contact investigation (among index patients identified 2 years prior to the inception of intervention) [26]. These approaches identified a considerable number of children with TB. The average number needed to screen (NNS) to identify 1 child contact with TB was 70 (range: 3–721) from 15 projects (Table 2). To improve participation, parents of child contacts received diagnostic and transportation support (Kenya, Nigeria, and others) when they were required to perform a radiological or histopathological evaluation.

Facility-Based Systematic Screening

While studies have shown facility staff routinely fail to diagnose adults with symptomatic TB seeking care [27], underdiagnosis is worse in children as they suffer the dual disadvantage of no routine TB screening and nonspecific symptoms that overlap with common childhood illnesses [24]. Overall, 15 projects implemented systematic facility-based screening ranging from peripheral level primary care facilities to high-volume tertiary care public and private facilities [21, 22]. These projects reported 2 625 079 children were verbally screened and 11 794 children with TB were diagnosed based on a composite criterion of clinical presentation (cough, fever, and weight loss), exposure to household contacts with TB, bacteriological evidence (if possible), radiological changes, and tuberculin skin test (TST). The average NNS to identify 1 child with TB was 223 (range: 9–2153) (Table 2).

TPT

Although enrolling more children onto TPT is one of the United Nations High-level Meeting goals, progress is severely limited [1]. Six projects enrolled 4942 eligible contacts on TPT. TB REACH projects introduced newer, shorter TPT regimens (3HP—weekly isoniazid and rifapentine for 3 months and 3HR—daily isoniazid and rifampicin for 3 months), and expanded TPT coverage to children, adolescent, and adult contacts. A risk stratification tool which was validated earlier in other settings [28], was implemented in Peru and tested under programmatic conditions while enrolling those eligible on TPT. A project in Ethiopia engaged women-led Iddirs (local social support groups), resulting in a significant increase in TPT initiation rates and enrollment, and high-completion rates [29]. Pre/post analysis in the 6 projects showed that TPT enrollment quintupled in the intervention area, with more than 95% treatment completion. Findings from the Bangladesh, Ethiopia, and Kenya projects showed that acceptability/uptake of TPT improved with a shorter, convenient regimen, and involvement of a medical doctor and community workers in the process.

STRUCTURAL INTERVENTIONS AND CAPACITY BUILDING

Decentralized Care Delivery Model

In many settings, children with TB are overwhelmingly diagnosed at large hospitals, as the services to aid bacteriological diagnosis and clinical expertise are centralized at the tertiary level [30]. Five projects encapsulated the experience of diagnosing children with TB in centralized/tertiary facilities and translated the learning to decentralize child TB programs. Projects in Bangladesh, Kenya, Pakistan, and Zambia trained health workers at primary-level sites (subdistrict and below) in child TB screening and triage, which increased childhood TB notifications by 100% or more (numerous sites had no child TB diagnosed and notified at baseline). Facility physicians, pediatricians, nurses, and other health staff were trained in child TB diagnosis and management based on national guidance. Specific avenues for clinical oversight and consultation for difficult clinical scenarios were provided, and clear referral mechanisms for specialized testing (including scanning and histopathology services) and care at district and sentinel sites were provided to families free of charge. Several projects conducted facility assessments to ensure peripheral facilities had required reagents (TST solution), tools (measurement of height, weight, and mid-upper arm circumference), and national TB program recommended diagnostics like chest X-ray (CXR), and histopathology. Some also linked to private laboratories for performing TST and CXR. Following decentralization of child TB services, community awareness (when and where to seek care), and coordination with local stakeholders were keys to ensure uptake of new services and improve healthcare-seeking behavior. Other projects in Cambodia and Nigeria decentralized screening for TB among children as part of community-based active case-finding activities and were able to link children to diagnostic testing at hospitals.

Supported Decision-Making (Video Consultation, Medical Board, and Mentorship)

Electronic mobile health technology has great potential to train and support clinicians at peripheral sites for diagnosing childhood TB [24]. Projects in Pakistan and Bangladesh used supported decision-making to link the decentralized health facilities (less-experienced physicians) with experienced physicians through weekly video consultations (contributing 25% of the project yield in Bangladesh), to guide the diagnosis and management of children nonconclusive diagnostic evaluation findings. Supported decision-making, including opinions from clinicians with different subject specialties, was employed. A weekly medical board was formed in Bangladesh (contributed to 14% of the project yield) comprised of physicians from several specialties, including pediatrics, infectious diseases, medicine, otolaryngology, pediatric surgery, and oncology. This helped with rapid and quality diagnosis (decisions taken upon agreement), reduced loss to follow-up, decreased out-of-pocket travel and testing expenses, empowered local physicians with confidence and knowledge, and helped create the next tier of trained trainers. Projects employed a novel-mentorship program using a WhatsApp (Meta Inc., Menlo Park, CA) group whereby less-experienced physicians requested a second opinion on diagnostic and treatment approaches from experienced physicians, receiving instant support to help diagnosis. A project in Kenya adopted a free online TB learning program using the Extension for Community Healthcare Outcomes Project [21].

IMPROVING DIAGNOSTIC METHODS

Efforts to develop rapid, sensitive diagnostics to identify TB closer to the point of care and tests that can be used to screen out TB including tests for TB infection are ongoing. Bacteriological confirmation of TB in children remains a challenge, especially among the very young, as their secretions are typically paucibacillary and they are often unable to produce adequate sputum. These factors affect the performance and use of the available laboratory tests and complicate diagnosis, as a negative test does not exclude the presence of TB disease [4].

Alternative Specimens

TB REACH projects in Bangladesh, Pakistan, and Kenya used alternative pediatric specimens (gastric aspirate, stool, and urine) to aid microbiological diagnosis. Evidence from the projects supported the recent recommendations from the WHO to include the use of Xpert assay (Cepheid, Sunnyvale, CA) in alternative specimens as an initial test [31]. Pediatricians were oriented on alternative specimen collection and received hands-on training on the collection procedure (gastric aspirate). In Bangladesh, a public facility was assisted in setting up a child-friendly out-patient unit. The unit was supported with staff trained on specimen collection procedures (gastric aspirate and sputum induction) and linked with an in-house Xpert testing facility which replaced the requirement of overnight in-patient admission. The unit also served as a referral hub for other public and private facilities, thus bringing more patients to the facility. Many children unable to produce sputum were tested through this approach and the Xpert positivity rates were 3%.

Experience With Xpert MTB/RIF Ultra

The lower limit of detection offered by Xpert MTB/RIF Ultra has opened an avenue to confirm paucibacillary TB in children [31, 32]. One of the projects in Bangladesh introduced Xpert MTB/RIF Ultra and tested a total of 2504 children with presumptive TB (pulmonary-2328; extra pulmonary-166) from 15 health facilities in the intervention area, using respiratory and stool specimens with an 11% positivity rate (n = 250) on Ultra including trace results. Results from this intervention were used in a meta-analysis to inform the WHO child TB guidance update [33].

Use of CXR and Need of Tailored Artificial Intelligence Algorithm for Younger Age Group

Radiological evidence plays a key role in diagnosing children with TB [24]. In the absence of microbiological confirmation, clinicians often rely on radiological evidence—coupled with clinical presentation and TB exposure. Lack of access to both rapid molecular testing and supportive tools such as CXR was addressed by transporting children for CXR (Nigeria and others) and providing CXR vouchers for referrals (Kenya), which improved CXR use in the diagnostic pathway. Other projects purchased new digital CXR machines and repurposed old machines (Afghanistan) to help with case finding among children. In Afghanistan the proportion of children investigated for TB doubled after investigation was encouraged by the provision of CXR as well as TST reagents, with diagnosis increasing more than 300%. Access to CXR and additional training improved clinicians’ confidence in diagnosing TB.

The expertise of CXR reading varies with experience levels among clinicians and radiologists. Moreover, the interreader variability is a diagnostic challenge. Automated CXR reading was recently recommended by WHO for adults [34]. Given the dearth of pediatric pulmonologists/radiologists, AI could be even more useful for children. However, evidence is lacking on the performance of AI on pediatric CXR images, and noted as a knowledge gap in the WHO guidelines. TB REACH projects in Bangladesh and Zambia are working to evaluate the performance of AI compared to human readers, providing evidence to inform future WHO reviews.

POPULATION-LEVEL IMPACT (CHANGE IN CHILD TB NOTIFICATION)

Similar to previously published results on active case finding [18, 35], TB REACH projects focused on improving childhood TB had positive outcomes. Using a standardized M&E approach to document impact on TB notifications [11, 36, 37] 16 TB REACH projects used age-disaggregated data. Results showed a total of 9127 additional (range −155 to + 1512) children with all-forms TB detected and treated, an increase of 34% in the intervention areas over expectations. During the same time, the control areas experienced a 7% decline (1609 fewer cases than baseline) (Table 3). The ratio between project yield and overall TB notifications in the intervention areas showed that 40% (range: 1%–100%) of child TB notifications were contributed by the projects (Table 3). Together, all projects reported a treatment success rate (cured and treatment completed) of over 92%.

Table 3.

Changes in All-forms Childhood TB Notifications in Intervention Area and Control Area, and Contribution of Project Yield to Notifications (n = 16 projects)

	Evaluation area All-forms	Control area All-forms
Baseline notifications	30 402	23 206
Implementation period notifications	35 637	21 597
Additional notifications	5235	(−1609)
% Change from baseline	17%	(−7%)
Predicted notification during intervention period *	26 510	21 319
Additional notifications	9127	278
% Change from predicted notifications	34%	1%
Project yield from intervention	14266	NA
Percentage of implementation period notifications from project yield	40%

Trend-adjusted using 3 years of historical childhood TB notifications.

Notification data comes from the TB programs. Project yield refers to children with TB directly identified by the project. All are collected through quarterly M&E reports through the TB REACH grant management system and reviewed by the external M&E team.

FUTURE DIRECTIONS FOR TB PROGRAMS

Diagnosing, treating, and preventing TB among children has well-documented barriers. The results of 27 TB REACH projects demonstrate that focusing efforts and funding on enhancing childhood TB diagnosis can produce marked improvements. Large increases in childhood TB notifications can be made by providing clinicians with the skills and confidence to diagnose the disease, conducting systematic screening at different facilities, and employing new tools and approaches to screen and test children while ensuring the process does not entail undue financial stress. Funding is a critical component of any health intervention, and, accordingly, as money to conduct activities for childhood TB runs out, the impact will drop as well [38]. We found that successful interventions screened more children for TB, brought the diagnostic process closer to where people sought care, and employed different technologies and tools to improve the diagnostic process.

The entry point for TB screening in many children will continue to be contact investigation. The clear lack of progress on global targets for TPT for children [1] as well as some published evidence on policy [39, 40] suggests that contact tracing implementation is not optimal. Our review demonstrated many different approaches to contact tracing, including “reverse” contact tracing, periodic visits and retrospective approaches, sample collection at the household, and transport assistance to help reduce dropouts. Active approaches to contact investigation were found to be more cost effective compared to passively inviting people to the health facility [41]. However, despite the higher risk for children who are contacts of people with TB, larger-scale approaches, often facility-based, tended to identify more children with TB despite higher NNS. Facility-based screening can also be less expensive than actively visiting homes, so making direct comparisons to NNS is not always valid [42]. Both approaches should be part of comprehensive plans to improve childhood TB diagnosis and management.

Given the lack of bacteriological confirmation in many children with TB, overdiagnosis—highlighted as a concern for TB among adults [43]—is also a concern in children [44, 45]. One approach we documented is to use audit or review panels that can assess decisions, provide support to clinicians, and gain consensus. However, missed TB diagnosis in children is a far bigger challenge, particularly in the very young child which carries grave consequences (morbidity and high mortality). Therefore, until improved diagnostic methods are developed, overdiagnosis, and over treatment in this age group should be considered acceptable to save lives. The forthcoming WHO child and adolescent TB guidelines update and accompanying operational handbook are expected to include clinical algorithm guidance for pulmonary TB among children under 10 years of age [46]. Application of the algorithm should help increase the specificity of clinical diagnosis, particularly in peripheral settings where all diagnostic tools may not be available.

Improved urine-based lipoarabinomannan assays [47, 48], use of nonrespiratory samples, and investments in radiology and AI [49] may enhance information clinicians have at their disposal to improve childhood TB diagnosis. As the software develops, AI reading of CXRs in children may help bring diagnosis closer to people’s homes and assist in TB screening at peripheral sites where lack of quality CXR and interpretation capacity are obstacles to child TB diagnosis. Across TB REACH projects and in general globally, bacteriological confirmation continues to be difficult. With more sensitive diagnostics as well as alternative specimen collection, more efforts must be made to confirm TB in children. Although new rapid diagnostics are more expensive, pooling specimens could save testing costs, especially where low yield is expected [50].

Enhancing awareness among health care staff and communities about TB symptoms, risk factors in young children, and the need to screen children for TB will be critical to improve existing case-finding approaches. To ensure that every child with TB is diagnosed and treated, all key frontline locations in the health facility, such as in-patient wards; malnutrition, immunization, antenatal, and human immunodeficiency virus clinics; and Integrated Management of Newborn and Childhood Illnesses units, should integrate screening for TB during baseline evaluations of children.

CONCLUSION

Despite improvements in recent years, TB case detection among children remains a huge challenge globally. The large gaps in pediatric TB case detection and the impactful results achieved by several interventions targeted to children across many countries demonstrate the need for more support for such activities as part of national strategic plans and for increased funding. This can only be achieved when strong political will acknowledges and funds the initiatives that care for vulnerable children in TB high burden, low- and middle-income countries. Much of the work must come from nonstandardized approaches, employing new technology in innovative ways, and ensuring that a complicated diagnostic process is both facilitated and simplified by the health system. The knowledge gained from TB REACH and other recent approaches can help national TB programs and partners achieve much more and reduce the child TB case detection gap.

Notes

Acknowledgments. The authors acknowledge the work of our grantees and the M&E team to help document the results and appreciates their reports, publications and data submitted to TB REACH from which these results have been derived. TB REACH is generously supported by Global Affairs Canada and additional support for several projects was provided by USAID. We thank Pamela Scofield for her careful review and editing.

Supplement sponsorship. This article appears as part of the supplement “What’s New in Childhood Tuberculosis?” sponsored by the Stop TB Partnership.

Potential conflicts of interest. M. T. R. and J. C. work for TB REACH at Stop TB Partnership, M. T. R., A. M., and F. A. have received TB REACH grants in the past.