Contactless Assessment of Patient Vital Signs for Triage Using Remote Photoplethysmography in the Emergency Department

Remote photoplethysmography (rPPG) is a camera-based method that enables contactless measurement of variation in light absorbance by haemoglobin as reflected by the changes in skin colour/pixel intensity unperceivable by human eyes. Beat-to-beat variation of light-absorbance can be used to estimate the change in the arterial blood volume underneath the skin and different vital signs including the blood pressure, heart rate, respiratory rate and oxygen saturation (SpO2) of the patient.

The investigators propose an observational prospective study to independently evaluate the accuracy of contactless vital sign measurements using a cell phone-based remote photoplethysmography (rPPG) technology. This study will recruit adult patients presenting to the Accident and Emergency Department (A&E) of Queen Mary Hospital (QMH) between 1 August 2024 and 30 October 2024. The study will be divided into 2 stages. In the initial run-in period, the investigators will collect facial video data from 200 patients to calibrate the VitalsTM system in the A&E environment. Then, the investigators will evaluate the accuracy of the VitalsTM in estimating vital signs of 1,000 ambulatory patients in the A&E setting.

A trained research nurse and research assistant will be deployed from 09:00 to 17:00 on weekdays to screen for eligible patients in the A&E waiting hall and to recruit patient participants throughout the study period. In this study, the investigators will only recruit triage category 3 (semi-urgent) to 5 (non-urgent) patients who are clinically stable and evaluate the performance of the VitalsTM in a designated room in A&E. Written informed consent will be obtained from all patient participants after an explanation of the details of the study, including the rationale, benefits and risks of participation.

After informed consent, contactless and manual measurement of patient blood pressure, heart rate, respiratory rate, body temperature and SpO2 will be carried out simultaneously in a designated room in A&E. A mounted light emitting diode (LED) will be used as the light source. Facial video of the patient participant will be captured using an iPhone, an iPad and a thermal camera placed at around 50 cm from the patient. A reference object for thermal and visual imaging will be placed at the background. Also, ambient temperature and light intensity will be measured.

The patient will be asked to sit down and remain still during video-recording. Manual measurement will be performed simultaneously by a trained research nurse or research assistant using standard hospital device or other medical-grade devices. A software program will be used to log vital sign measurements and recordings simultaneously so that data can be synchronised to reduce variations.

The primary outcome is the accuracy of the VitalsTM platform in estimating patient heart rate measured in intraclass correlation coefficient (ICC). The investigators will also measure of accuracy of the VitalsTM platform in estimating other vital signs, including blood pressure, respiratory rate, SpO2 and body temperature. The investigators will also evaluate patient satisfaction and comfort with the contactless and manual measurement techniques.

The accuracy of contactless measurement will be determined by calculating the ICC from the manual measurement readings as the ground truths. The investigators will also calculate the root mean square error (RMSE) and Pearson correlation between the contactless and manual measurement readings for individual vital signs. Altman Bland plot will be used to evaluate the bias and limits of agreement between two vital sign measurement methods.