Global paediatric drug-resistant tuberculosis surveillance after the removal of age-disaggregated WHO reporting fields: a cross-national analysis

Paediatric drug-resistant tuberculosis (DR-TB) is one of the most poorly documented dimensions of the global TB epidemic. In 2020, the World Health Organization removed age-disaggregated rifampicin-resistant TB (RR-TB) fields from its standard reporting format. No country has since resubmitted these fields to WHO, although some may continue to collect paediatric age-disaggregation through national or sub-national systems that are not reflected in the global dataset. This study analyses 2017–2019, the only three years for which age-disaggregated data exist in the WHO database, to quantify what has been lost from routine global surveillance and to inform discussion of reinstating standardized age-disaggregated reporting. We analysed WHO drug resistance surveillance data for 215 countries and territories over 2017–2019. Generalised estimating equations (GEE) with a Poisson family were used to describe year-on-year paediatric and adult notification patterns; because only three annual data points are available, these are descriptive comparisons rather than evidence of an underlying temporal trend, and no inferential trend claim is drawn from them. We constructed an Under-Reporting Index (URI) as a measure of each country’s deviation from the reporting pattern of its own WHO region, not as a direct or externally validated measure of the true extent of under-reporting, to identify countries whose notified paediatric caseload deviates most from regional norms. A gradient-boosted ensemble trained on approximately 87 country observations was interpreted with SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME); given the small sample size relative to the number of candidate predictors and the absence of an independent external validation dataset, the machine-learning component of this study is exploratory and hypothesis-generating throughout, and none of its outputs should be read as confirmatory or as demonstrating causal relationships. Among 87 countries with any paediatric RR-TB data, 6,846 children under 15 were notified with RR-TB during 2017–2019. The global paediatric fraction rose from 1.82% in 2017 to 3.28% in 2018 before falling to 2.52% in 2019 (three-year mean: 2.54%; a descriptive pattern across three annual observations, not a statistically evaluated trend). Africa (3.64%) and South-East Asia (3.57%) carried the highest regional fractions. Notified paediatric counts fell well short of regional benchmarks in many countries: 209 country-years recorded zero paediatric RR-TB notifications despite non-zero adult caseloads. Among the exploratory machine-learning features, log-transformed total RR-TB burden had the largest mean absolute SHAP value (2.81; bootstrap 95% CI: 1.94–3.60), followed by retreatment fraction (1.17; 95% CI: 0.61–1.79); confidence intervals across features overlapped substantially, so these rankings are exploratory correlational associations from a small-sample model and should be read as directional and hypothesis-generating rather than definitive or causal. Notified paediatric RR-TB fell well below regional reporting benchmarks before 2020, indicating a substantial gap between the paediatric and adult evidence base rather than establishing the true size of undetected paediatric disease. WHO’s removal of age-disaggregated RR-TB fields from the standard reporting format that year has left a surveillance data gap that, as of this analysis, spans five consecutive years. Reinstating standardized age-stratified reporting, extending drug susceptibility testing to paediatric patients, and investing in community-based case-findingan approach supported by evidence from multiple African settings, not only the single country highlighted in our own data are among the actionable responses this descriptive and exploratory analysis can support.