Abstract
Background: Digital adherence technologies (DATs), such as phone-based technologies, and digital pillboxes, can potentially provide more patient-centric approaches for supporting tuberculosis (TB) medication adherence and improving outcomes. My project focused on the diagnostic accuracy of dose reporting generated by DATs for TB treatment support.
Design/Methods: A systematic review was conducted to identify relevant literature from January 2000 through April 2022 in MEDLINE, Embase, CENTRAL, CINAHL, Web of Science, and preprints from medRxiv, Europe PMC, and clinicaltrials.gov. The systematic review protocol was registered a priori as PROSPERO-CRD42022313526. Studies reporting quantitative data on the accuracy (sensitivity/specificity) of DATs for TB infection or disease treatment, against a reference standard (urinalysis test, pill count, or directly observed therapy [DOT]), with at least 20 participants using the DAT were included. Study characteristics and accuracy outcomes (i.e., sensitivity, specificity, PPV, NPV, and number of events) were extracted. This systematic review is part of a series of linked reviews that aim to highlight the accuracy of dose reporting, health outcomes, cost and cost-effectiveness, and implementation considerations of DATs for TB treatment support.
Results: Of 4,803 studies identified by our systematic search, 8 met our inclusion criteria. The accuracy of DATs was reported for phone-based technologies, specifically phone calls with medication sleeves [N=3], digital pillboxes [N=3], and ingestible sensors [N=2] for a total of 1000 persons treated for TB disease. Two studies used a digital pillbox as their reference standard, so the accuracy of the pillbox was calculated against pill count from their primary data. When ingestible sensors were compared to DOT, the probability of detecting a dose truly taken (sensitivity) was ≥95%. Across studies, the diagnostic accuracy of digital pillboxes varied extremely, while sensitivity and specificity were 60-80% [95% CI 57-81%] for phone calls with medication sleeves using patient-reported doses.
Conclusion: The limited available evidence suggests variable accuracy of DATs. Future research should aim to better understand how the technology, setting, and patient characteristics, and patient-technology interactions impact DAT accuracy. Additionally, there remain evidence gaps around the accuracy of some widespread DATs such as SMS-based interventions, and around the digital support of patients with TB infection.