Heart rate variability (HRV) has been well established to measure instantaneous levels of mental stress. Circadian patterns of HRV features have been reported but their use to estimate levels of mental stress were not studied thoroughly. In this study, we investigated time dependent variations of HRV features to detect subjects under chronic mental stress. Sixty eight subjects were divided into high (n=10) and low stress group (n=43) depending on their self-reporting stress scores. HRV features were calculated during three different time periods of the day. High stress group showed decreased patterns of HRV features compared to low stress group. When logistic regression analysis was performed with raw multiple HRV features, the classification was 63.2% accurate. A new % deviance score reflecting the degree of difference from normal reference patterns increased the accuracy to 66.1%. Our data suggested that HRV patterns obtained at multiple time points of the day could provide useful data to monitor subjects under chronic stress.

No abstract

Heart rate variability (HRV) has been well established to measure instantaneous levels of mental stress. Circadian patterns of HRV features have been reported but their relationships to mental stress were not studied explicitly for estimating stress levels. In this study, we investigated long term variations of HRV features to provide a reliable measure of chronic stress levels. Twenty three subjects were divided into high (n=10) and low stress group (n=13) depending their selfreporting stress scores. HRV features were calculated during five different time periods of the day. High stress group showed decreased overall variations of HRV features but similar median values to low stress group. Compared to normal sinus rhythm data during each time period, high stress group showed significantly less % difference of HRV patterns than low stress group. Our data suggested that long term variations of HRV features might be more useful to detect subjects under chronic stress.

Circadian variations of cardiac diseases have been well known. For example, atrial fibrillation (AF) episodes show nocturnal predominance. In this study, we have developed multiple formulas that detect AF episodes in different times of the day. Heart rate variability features were calculated from randomly sampled three min ECG data. Logistic regression analyses were performed to generate three formulas for the entire day, daytime, and evening time. Compared to the first formula that disregarded the time of the day, the second formula for the daytime detection detected AF episodes more accurately (95.2% vs. 99.3%), whereas third formula for the evening time detection did less accurately (93.8%). These results suggest the detection of AF episodes might become more accurate by considering the time-dependent changes of HRV features. In addition, the detection method for the evening time requires further investigation.

Long term patterns of heart rate variability (HRV) features were decreased in subjects with higher self reporting stress scores. For mobile applications, short term analysis of HRV features may be ideal since conventional heartbeat recordings (3~5 min) might be inadequately long. In this study, short term analysis has been performed for heartbeat data obtained at five different time points from two subject groups (15 under high and 18 under low mental stress). The reliability of short term heartbeat data was demonstrated by detecting significant differences in long term patterns of HRV features between two groups.Fifteen to thirty second heartbeat measurements were long enough to produce reliable long term patterns of HRV features. Thus, short and intermittent recordings of heartbeats could be used to detect long term HRV patterns and offer a convenient method to monitor mental stress in mobile environments.