Smartphone cough monitoring linked to TB but not yet reliable

Tuberculosis screening often depends on patients reporting cough, a symptom that can be missed or underreported. To reduce that subjectivity, researchers led by Patrick Biché tested whether passive cough monitoring with smartphones could help identify people likely to have TB. The study enrolled adults aged 15 and older who were either screened for TB in community settings or evaluated at health facilities in Kampala, Uganda. Participants had microbiological testing and wore a smartphone app that recorded coughs for 48 hours. The app used for monitoring was the Hyfe Research app. TB status in the study was determined using Xpert MTB/RIF Ultra and culture. The team also used inverse probability weighting to adjust for differences in who enrolled, and they interviewed clinic staff to understand practical challenges. By combining device-recorded data with lab tests and staff feedback, the study aimed to measure both diagnostic accuracy and whether phone-based monitoring could realistically be implemented in a low-resource setting.

In total 884 people were enrolled, of whom 197 had both valid cough recordings and confirmed TB status and were included in the main analysis (101 from community screening and 96 from health facilities). People with TB had higher median cough frequency than those without. In community screening the median was 2.2 coughs per hour (IQR 0.8–6.1) for those with TB versus 0.9 (IQR 0.4–2.0) for those without; in facilities the medians were 6.7 (IQR 2.6–27.5) versus 2.4 (IQR 0.9–4.8). These differences were statistically significant (Wilcoxon P < 0.0001 for both settings). Using weighted receiver operating characteristic curves, the area under the curve (AUC) was 0.69 (95% CI: 0.58–0.79) in the community and 0.76 (95% CI: 0.6–0.88) in facilities. At a sensitivity of 90%, specificity was low: 20% (95% CI: 0–22%) in the community and 46% (95% CI: 1–32%) in facilities. Smartphone-recorded cough frequency correlated moderately with self-reported cough severity, Saint George’s Respiratory Questionnaire scores, and staff-observed coughs. Staff interviews highlighted barriers to implementation including device visibility, stigma, and security concerns, and these operational challenges limited data collection.

The study shows that smartphone-recorded cough frequency is associated with TB status but, as tested here, does not reach the accuracy needed to act as a stand-alone screening or triage tool. The Hyfe Research app picked up more coughs among people whose Xpert MTB/RIF Ultra and culture results confirmed TB, but the overlap between groups and low specificity at high sensitivity means many people without TB would be flagged if the tool were used alone. Implementation problems reported by staff — concerns about the phones being visible, stigma tied to being seen with monitoring devices, and worries about device security — further limit real-world use. The authors conclude that future work must address these operational barriers and improve diagnostic performance before cough-monitoring apps can be widely deployed for TB screening. For now, smartphone cough monitoring may be a useful adjunct to existing pathways but not a replacement for laboratory testing and established clinical assessments.