Background: Direct observation has been widely used to assess interactions between healthcare workers (HCWs) and patients but is time-consuming and feasible only over short periods. We used a Radio Frequency Identification Device (RFID) system to automatically measure HCW-patient interactions. Methods: We equipped 50 patient rooms with fixed sensors and 111 HCW volunteers with mobile sensors in two clinical wards of two hospitals. For 3 months, we recorded all interactions between HCWs and 54 patients under airborne precautions for suspected (n=40) or confirmed (n=14) tuberculosis. Number and duration of HCW entries into patient rooms were collected daily. Concomitantly, we directly observed room entries and interviewed HCWs to evaluate their self- perception of the number and duration of contacts with tuberculosis patients. Results: After signal reconstruction, 5490 interactions were recorded between 82 HCWs and 54 tuberculosis patients during 404 days of airborne isolation. Median (interquartile range) interaction duration was 2.1 (0.8-4.4) min overall, 2.3 (0.8-5.0) in the mornings, 1.8 (0.8-3.7) in the afternoons, and 2.0 (0.7-4.3) at night (P,1024). Number of interactions/day/HCW was 3.0 (1.0-6.0) and total daily duration was 7.6 (2.4-22.5) min. Durations estimated from 28 direct observations and 26 interviews were not significantly different from those recorded by the network. Conclusions: The RFID was well accepted by HCWs. This original technique holds promise for accurately and continuously measuring interactions between HCWs and patients, as a less resource-consuming substitute for direct observation. The results could be used to model the transmission of significant pathogens. HCW perceptions of interactions with patients accurately reflected reality.