The prospective observational cohort study was approved by the Ethics Committee of Medicine in Helsinki Uusimaa Hospital District and was performed without external funding. No written consent was required by the Ethics Committee. All patients over 18 years of age admitted to the ED for any reason during 72 hour data collection period were included in the study. Patients with incomplete data were excluded from the final analysis. The study was conducted in a tertiary referral university teaching hospital ED covering following specialties: general medicine, respiratory medicine, neurology, gastroenterological surgery, vascular surgery and thoracic surgery. The study did not affect the treatment or triage classification of the patients. Only the two research nurses responsible for the data recording were aware of the pulse photoplethymography values.
Basic vital signs including blood pressure, pulse rate, respiratory rate, capillary refill (over or under 2 seconds), oxygen saturation (SpO2), level of consciousness (on a 4 level AVPU scale: alert, responds to voice, responds to pain, unresponsive), and body temperature (both core and peripheral, from tympanum and index finger, respectively) were measured and recorded on arrival to ED by either of the two research nurses. Possible use of vasoactive medications before data collection was recorded.
The pulse photoplethysmograph signal was collected using an AS/3 monitor (GE Healthcare, Little Chalfont, United Kingdom) for one minute from an index finger during the triage within ten minutes of presentation to the ED immediately before measuring the basic vital signs. Nail polish was removed with acetone if present. Data was recorded with Collect S/5 software, version 4.0 (GE Healthcare, Little Chalfont, United Kingdom) and PPGA was automatically averaged every 10 seconds as previously described . No calibration for the hardware or software was necessary during the measurements. The research nurses received hands-on training in the use of the equipment and their performance in using it was frequently evaluated by one of the researchers.
To describe the intensity of care needed in the ED, we registered interventions performed (invasive hemodynamic monitoring and intubation) and total volume of intravenous fluid received. Based on patient records we collected data on highest level of follow-up care during the hospital stay [discharged from ED, ward, high dependency unit (HDU), intensive care unit (ICU), operating theatre (OR), transfer to another hospital] as well as survival to hospital discharge.
Maximum, minimum and mean of PPGA (PPGAmax, PPGAmin, PPGAmean) were determined for each patient from collected data after the data collection period. To approximate variation in the PPGA signal, PPGAvar was derived from the equation used by Broch et al.  to determine pleth variability index (PVI): PPGAvar = (PPGAmax – PPGAmin) / PPGAmax. To classify critically ill and non-critically ill patients, modified Early Warning Score (mEWS) was calculated from the basic vital functions for each patient and mEWS > 3 was used as cut-off as shown to predict higher requirement for intensive care in an earlier study . MEWS as a proven predictor of death and critical illness [5, 6] was used as a surrogate due to the pilot nature and limited number of patients in the study and thus death as an outcome was deemed to be unlikely to reach statistical significance.
Statistical analysis was performed using GraphPad Prism, version 5.0d (GraphPad Software Inc., San Diego, CA, USA). We determined standard deviations, interquartile ranges, mean values and 95% confidence intervals (CI) for different variables. The variables were analyzed for normality using D’Agostino & Pearson test. Because all other data collected except for heart rate were not normally distributed, we analyzed the data using Mann–Whitney test for other variables and unpaired T test for heart rate. For correlation analysis we used Spearman correlation where applicable. We also calculated receiver operating characteristic (ROC) curves for the different photoplethysmography variables to detect criticall illness. P-value of < 0.05 was considered statistically significant.