Digital tools cut tuberculosis treatment losses in Kenya

Tuberculosis (TB) remains a serious global health problem because successful treatment is essential to protect patients, prevent transmission, and reduce costs. To test whether affordable digital tools could help more people finish treatment, researchers led by Erez Yoeli ran a large randomized trial in Kenya. The study enrolled people with TB who still had at least two months of treatment to go and compared the usual clinic care with three different digital add-ons. One group received daily SMS medication reminders. A second group received access to an interactive platform that asked people each day to confirm they had taken their medicine and provided disease information, motivational messages, and an adherence game. A third group used that same platform but also got extra help from a team of trained supporters; this version was called Keheala. The trial was open-label and run at 902 clinics, with participants randomized to these four arms to see whether increasing levels of digital support could improve completion of TB therapy.

The trial ran from April 13, 2018 to December 20, 2019 and randomized 16,753 people, producing a modified intention-to-treat (mITT) population of 14,962: 1,997 in the control arm, 1,475 in the SMS arm, 6,057 in the platform arm, and 5,433 in the Keheala arm. Randomization used a 4:3:12:12 ratio across the four groups. The study’s primary outcomes were the proportion with an unsuccessful treatment outcome (a composite of died, failed treatment, or loss to follow-up) and loss to follow-up (LTFU) specifically. A secondary outcome measured medication non-adherence using unannounced urine isoniazid tests in a random sample of 731 people from the control and Keheala groups. In the control group the absolute risk of an unsuccessful outcome was 12.4%. Compared with control, that risk fell by 1.9 percentage points in the SMS group (95% C.I.: -0.1–4.0), by 1.9 percentage points in the platform group (95% C.I.: 0.3–3.4), and by 2.6 percentage points in the Keheala group (95% C.I.: 1.0–4.2). Medication non-adherence was 12.4% in control and was reduced by 7.5 percentage points in Keheala (95% C.I.: 2.6-12.5). The authors report that most of the improvements were due to reductions in loss to follow-up.

These findings show that relatively simple digital supports can nudge TB programs toward better outcomes without replacing standard care. All three digital interventions—daily SMS reminders, the self-verification and information platform, and the platform plus human support (Keheala)—produced modest reductions in unsuccessful treatment outcomes, and the most supported package (Keheala) also led to a measurable drop in missed doses as confirmed by unannounced urine isoniazid tests. Because the trial was large, ran across many clinics, and focused on people already in care with at least two months left in treatment, the results are directly relevant to program managers deciding how to improve completion rates. The authors suggest these digital tools could be considered as supplements to routine services, especially in settings where in-person support is hard to provide. The data point to a scalable approach: combining technology with human encouragement can lower loss to follow-up and help people stay on potentially lifesaving TB treatment.