Evaluation of the tuberculosis notification impact from an intermediary-facilitated public–private mix intervention in Viet Nam: A population-level quasi-experimental study

Background: Treatment coverage gaps remain a major barrier to ending tuberculosis (TB). Public–private mix (PPM) entails the engagement of healthcare providers outside of national TB programs (NTP). PPM strategies are promoted to improve TB notification, particularly in high-burden settings with fragmented care systems, and intermediary agencies are recognized as effective mechanisms to operationalize PPM by bridging NTPs and out-of-network healthcare providers. However, population-level evidence is rare, as most evaluations have been limited to pilot studies, small geographic areas, or uncontrolled designs. This study documents the population-level association of a PPM intervention with TB notifications through the use of unbiased analytical methods to account for heterogeneity from the intervention’s staggered deployment and dynamic intensity. Methods and findings: We present a non-randomized, quasi-experimental evaluation of a large-scale intermediary-facilitated PPM intervention conducted between 2020 and 2023 in 15 provinces of Viet Nam, covering 40.6 million people equivalent to 40% of the population. The intervention engaged 3,154 non-NTP providers in TB care, linking them to the NTP through systematic screening, referral, and reporting mechanisms. The model emphasized active case finding, provider incentives, patient support and integrated data systems to raise notifications as the primary outcome. We analyzed quarterly province-level TB notification data from 2016 to 2025 using a heterogeneity-robust extended two-way fixed-effects framework to estimate the average treatment effect on the treated. The analysis adjusted for province and calendar-quarter fixed effects, and for COVID-19 disruptions. Using Poisson pseudo-maximum-likelihood fixed-effects models, we estimated dose–response effects and post-intervention patterns and triangulated effects with an interrupted time-series analysis of national notifications and repeated all analyses using the intermediate outcome of PPM notifications. Lastly, we calculated marginal cost per additional notification from a funder perspective. The intervention was associated with 316.8 additional notifications per province-quarter (95% confidence interval [CI] [23.4, 610.2]; p = 0.034). Regarding dose–response effects, we observed a 1.4% increase in notifications per 100 verbal assessments per 100,000 population (incidence rate ratio [IRR]=1.00014; 95% CI [1.00002, 1.00025]; p = 0.018). National models suggested 3,206 (95% CI [1,658, 4,763]; p