Do electronic medication monitors improve tuberculosis treatment outcomes? Programmatic experience from China

Background In China, an indigenously developed electronic medication monitor (EMM) was used. EMM recorded each time the device was opened (no real time data), offering an indirect measure of tuberculosis treatment adherence. Previous study in China showed that the EMM uptake was satisfactory, missing adherence data were common in the information management system (25%) and shift to directly observed therapy (DOT) based on poor adherence documented by EMMs were seldom. Objectives Among people with tuberculosis notified in 30 counties (July-December 2018) where EMM supported self-administered therapy (SAT) was suggested to all eligible (no communication impairment, ambulatory), we assessed the relative differences in unfavourable outcomes and deaths among those started on EMM at baseline (within first month of diagnosis) when compared to SAT alone. Methods This was a cohort study using secondary data. We employed an intention to treat analysis, and used modified Poisson regression with robust variance estimates to assess the association. Results Of 1810 eligible people, 1047 used EMM at baseline and of them, 216 (20.1%) stopped using EMM midway. Of 763 people who did not use EMM at baseline, 267 (35.0%) started using EMM later during the treatment. Among those who started using EMM at baseline, 6.3% [95% CI: 4.9, 8.0] had unfavourable outcomes compared to 6.7% [95% CI: 5.1, 8.8] among those who did not (p = 0.746). Lesser deaths were observed in people who started EMM at baseline when compared to those who did not: 2.5% [95% CI: 1.7, 3.7] versus 3.5% [95% CI: 2.4, 5.2], p = 0.191. The lack of association remained after adjusting for potential confounders (occupation, TB classification and TB category). Conclusion Under programmatic settings, we did not find significant differences in the outcomes. Optimization of EMMs by shifting to DOT when indicated, addressing the issue of missing data and ensuring continuous use is required.