Linear discriminant analysis of heart rate variability differentiates between autonomic modulation induced by acute physical and psychophysiological stress

Purpose: Heart rate variability (HRV) is increasingly used to study the psychophysiology of stress in laboratory environment. However, in the real life, it can be difficult to differentiate the relative contribution of psychophysiological stress (PS) and physical stress (PhS) when their ffects may overlap, such as during military or police tactical operations, or critical incidents. This study aimed to assess the predictive accuracy of HRV parameters in differentiating autonomic modulations induced by maximal PhS and by PS during realistic operational training (OT).\r\nMethods: 12-lead ECG of 40 police officers (POs) was continuously monitored during rest daily activity (control state) and during highly demanding OT scenarios implying high PS without or with minimal PhS. All POs underwent, in a separate session, an exercise test with bicycle ergometer until muscle exhaustion. Baseline\r\nclinical and psychological assessment was performed on the days of the\r\nOT session. Behavior and tactical outcome were monitored with multiple video cameras. Quantitative estimates of time-domain (TD), frequency-domain (FD), and nonlinear HRV parameters were computed from standard (300-seconds),\r\nshort-term (120-seconds) and very short-term (60-seconds) intervals. Preliminarily time-varying (TV) spectral HRV analysis was used to visualize transient fluctuations\r\nof FD components. Discriminant Analysis (DA) was applied to evaluate\r\nwhich HRV parameters (or their combination) were efficient to provide adequate separation between PhS and PS.\r\nResults: TV HRV analysis provided dynamic imaging of transient autonomic\r\nadaptation induced by PS and/or PhS. Quantitative estimation of the majority\r\nof TD and FD HRV parameters was not significantly affected by shortening\r\nthe length of the explored time-segments (from 300 to 60 seconds), as demonstrated by the intraclass correlation coefficient >0.70. DA differentiated PhS from PS with best predictive accuracy of single HRV parameters ranging between 82% and 92% (SD1/SD2 and SDNN/RMSSD, measured from 60-seconds and 120-seconds intervals, respectively). The highest classification accuracy (93%) was obtained with the combination of NL SD1/SD2 (Poincaré plot), rplmean and\r\nrpadet (recurrence plot).\r\nConclusions: TV HRVa is efficient in evidencing transient changes of autonomic modulation of the heart rate during realistic OT. Quantitative estimate of HRV parameters is not significantly affected by shortening the duration of the explored time-window from 300 to 60 seconds. DA of HRV features may be useful to evaluate PS induced by realistic police OT and to distinguish it from overlapping PhS.