Heart Rate Variability for Evaluating Psychological Stress Changes in Healthy Adults: A Scoping Review Open Access

With advances in wearable technologies, it has become much easier to monitor the general population’s physiological health and well-being objectively. Heart rate variability (HRV), referring to variations in the time intervals between consecutive heart beats [1, 2], has emerged as a non-invasive tool to estimate psychological states such as fatigue, stress, anxiety, and burnout and is considered an indirect indicator of general mental well-being [3, 4]. HRV refers to the fluctuations in time between consecutive heartbeat cycles. HRV biofeedback, widely adopted in various therapeutic interventions, uses real-time electronic feedback of the moment-to-moment changes in HRV to achieve therapeutic goals [5, 6]. Several studies in psychology and neuroscience have confirmed that HRV measurements reflect the continuous interplay between the sympathetic and parasympathetic branches of the autonomic nervous system that maintain homeostasis of physiological arousal [7‒9]. A strict periodicity of reduced HRV is not a sign of good health which often reflects the experience of physiological distress and can be associated with several pathological conditions [10]. As such, novel wearable devices increasingly incorporate non-invasive physiological monitoring methods to capture real-time HRV-based stress changes. Studies comparing the quality of HRV measurements acquired from conventional ECG and those obtained from wearable devices have revealed that HRV obtained from trackers and wearable devices resulted in a small acceptable error, the method is valid, more practical, and cost-effective in tracking stress and well-being [11]. Given the vast interest toward using HRV metrics in unobtrusive stress tracking, studies that report HRV and its associations with stress assessed with stress-screening tools warrant a thorough examination.

Any physical or psychological stimuli can invoke a stress response due to the disruption of homeostasis. Or in other words, an environmental demand (stressor) translates both into psychological (self-appraisal or perception) and biological responses (stress response) in the body [12]. Response to these stressors is shown to be mediated by a complex interplay of nervous, endocrine, and immune mechanisms, which in turn produce objective physiological changes such as altered HRV [13]. The adaptability of an individual to these stressors and their physiological responses are influenced by several factors including the nature of stressors, intensity, frequency, duration of exposure to the stressful stimuli, and other health conditions [14].

Over the last two decades, several original research and review articles have been published examining the relationship between HRV and different types of stressful situations. Heightened occupational stress is associated with lowered HRV, specifically with reduced parasympathetic activation [15‒17]. Low parasympathetic activity, characterized by a decrease in high-frequency power and an increase in low-frequency power, was reported in a review as the most common factor associated with changes in stress [18]. As a hallmark of depression, blunted stress reactivity manifested as reduced high-frequency HRV has also been reported [19]. Neuroimaging studies on cerebral blood flow have shown that threat and safety-related acute stress significantly modulates HRV and supports the idea that the vagus nerve serves as a structural and functional link between the brain and the heart [20].

Psychological stress occurs when a stressful condition causes negative affective states (e.g., feelings of anxiety and depression) in an individual where they perceive that the environment’s demand exceeds their capacity [12] and invoke physiological changes in the body. Studies have shown that psychological stressors such as public speaking or interpersonal conflict have been shown to evoke increases in blood pressure and heart rate [21]. While previous reviews support the notion that changes in HRV is a valuable physiological indicator of psychological stress, it is unclear whether psychometrically validated instruments underpin stress assessments in all the considered studies. If quantified using psychometrically validated instruments, the perceived subjective component of stress [12] is more likely to reduce errors and biases associated with self-reports [22]. Second, findings from studies based on different stress contexts or demands such as physiological (e.g., pain, hunger), psychosocial, cognitive load, anticipatory etc., (e.g., mathematical tasks, social exclusion, achievement, or competitive situations) and different protocols of stress inducement were all congregated in the synthesis and meta-analysis of reviews as though the autonomic responses experienced in all these scenarios are homogenous [18]. This aspect requires some dismantling to investigate how responses vary with contexts and to explore any specific associations between the contexts and certain HRV metrics. Such metrics could be further explored within those contexts [23]. Third, studies do not always include a comparison with a stress-assessed at baseline or restful condition, which is also important to investigate and address, again based on the above-mentioned reasons that emphasize the need for quantifying and validating the intra-individual perceived stress changes and statistically comparing them with their baseline. This scoping review was carried out to address these gaps by (i) reviewing HRV-based studies with a well-designed within-subjects design that associated HRV with psychological stress states, (ii) including studies that used a validated psychometric instrument to elicit stress levels to systematically compare stress and HRV responses between stressful states, and (iii) including studies that provide a comparison of stress state measures with a no-stress baseline or restful condition. In the review process, as secondary objectives, we also explored whether included studies reported the mediating and/or moderating effects of underlying chronic stress, short-term HRV analysis windows, and the impact of breathing on HRV.

A detailed stepwise search strategy was developed in collaboration with a research librarian at Flinders University. Two primary key terms (psychological stress and heart rate variability) were developed as “concepts ”based on the research question. Detailed alternate terms and synonyms for each key concept were also identified as free terms and used to search articles’ titles, abstracts, keywords, and key concept fields in the following electronic databases: Medline, PsycINFO, Scopus, and Web of Science. Detailed search strategies are provided in the online supplement (online suppl. Table. S3; for all online suppl. material, see https://doi.org/10.1159/000530376).

Studies that (i) examined the relationship between HRV and psychological stress levels, (ii) with measures of HRV acquired in two or more time points that are established or hypothesized within the protocol as different stress states, (iii) used validation instruments to quantify mental or psychological stress, (iv) included healthy adult population, and (v) written in English were included. Conference articles, any form of review articles, case studies, letters to the editors, dissertations, and books were excluded.

Studies that purposefully sampled clinical populations with a diagnosed condition (e.g., insomnia, dysphoria, clinically depressed states, metabolic syndrome, irritable bowel syndrome, neuroticism, psychosis, panic disorder, eating disorders, patients who underwent surgery, hypertensive women, borderlines) were excluded. Thus, “healthy adult population” here refers to populations recruited from the general community without using any sampling criteria based on clinical status, and not based on whether the “healthy” status was ascertained with a psychometric questionnaire.

A total of 15 journal articles evaluating HRV in response to psychological stress and meeting the eligibility criteria mentioned above were included in this review. Study characteristics and findings were extracted from every publication and listed. During the process of data acquisition and integration, this list was completed with additional variables that appeared relevant and eventually covered the following information: author, publication year, sample size, gender, age range, type of stress or experimental manipulation, duration of analyzed task period, the time between repeated measures, the context of the study, outcome measures, and key findings. A mixed methods appraisal tool was used to appraise the quality of the studies and is included in the online supplement (online suppl. Table S2).

Using the search criteria described in the methods section, electronic databases were searched for HRV research in psychological stress. The initial search fetched 8,297 studies as of December 2020; these were reduced to 5,638 once 2,659 duplicates were removed, to 90 after screening for titles and abstracts, and then to 15 after reviewing the full texts for study methods (Fig. 1). A detailed list of the studies included in the present review is shown in Table 1.

Sample sizes ranged between 10 (Cervantes Blasquez 2009) and 403 (Kanthak 2017) participants, with a mean of 67 participants. All but one study included both genders, ages ranging between 18 and 60. The nature of stress, whether it was real-life situational stress or experimentally induced stress, was also elicited from the studies. Ten studies involved a controlled setting where stress was induced and quantified using HRV, both at baseline and while participants performed cognitive control-based timed tasks such Stroop task, mental arithmetic, speech preparation and delivery, Trier Social Stress Test (TSST) etc. To collect HRV, four studies used single-lead ECG signals, while Dimitriev et al. (2016) acquired 12 lead ECGs. Wearable devices (n = 7) such as Polar, Pulse, Firstbeat, Holter (n = 1) etc., were also used as RR acquisition modes. The study characteristics are summarized in Table 1.

The most common stress assessment method used in the reviewed studies (n = 10) is the State-Trait Anxiety Inventory (STAI) which captures state anxiety [24]. Different versions of the STAI have been used within these studies, complemented by other instruments like the visual analog scale, and the perceived stress scale has also been used (Table 2). Some studies use specific symptom-focused instruments like the Maslach Burnout Inventory (MBI), Centre for Epidemiological Studies Depression Scale (CES-D), and Ruminative Response Scale (RRS), among others, to quantify psychological stress. One study (Schubert 2009) measured chronic stress using the hassles frequency subscale of the Combined Hassles and Uplifts Scale (CHUS).

In the reviewed studies, a wide range of linear and nonlinear HRV metrics recommended by the Task Force of the European Society of Cardiology [25] were used. The number of metrics used in individual studies varied from one parameter to 11 different parameters. The most used time-domain parameter was the RMSSD (n = 10), and the second most common time-domain parameter was the SDNN (n = 8). The most used frequency-domain parameters were the LF/HF ratio (n = 7) and HF power (n = 6). Overall, nonlinear metrics have been applied less frequently in stress assessment. However, within included studies based on nonlinear analysis, SD1 (n = 5), SampEn, and α1 (n = 3) have been shown to be associated with stress. The online supplementary material (online suppl. Table S1) shows a detailed summary table showing all HRV metrics analyzed in each study and highlighting the metrics reported to be significantly associated with stress in the context of each study.

Anticipatory, acute stress in real-life situations such as before sitting for a university exam [26, 27] and before swimming [28] was investigated to evaluate stress. Reviewed studies also addressed other real-life stressors, such as chronic stress due to depression and rumination, emotional exhaustion, and burnout. Experimental acute stressors were the most common form employed in nine reviewed studies (Table 1). Acute task-based stress was induced in these studies by making participants perform mental arithmetic, speech, verbal fluency, imaging, and Stroop test tasks.

In the included studies, a range of linear time-domain, frequency-domain, and nonlinear metrics derived from the RR interval time series have been used to quantify HRV. The definition of these metrics, the studies in which they have been included, and the physiological changes that these metrics have shown to elicit are summarized in Table 3. To integrate the findings on HRV changes associated with stress, we coded baseline-to-task changes for every HRV measure as either significantly increasing or directly related (↑), decreasing or inversely related (↓) as shown in Table 1. Similarly, different psychometric evaluation instruments were used, and measures were derived in each study (Table 2).

This study was conducted to review the available literature for studies investigating various HRV metrics in healthy adult populations under different psychological stress states, including a baseline or a restful state. The review thus presents a synthesized summary derived from the included studies on different categories of HRV metrics, associated stress types, stress elicitation methods used, and the psychometric validation instruments used to evaluate stress.

Common real-life stressors of moderate intensity, such as preparation for a major university examination, may alter basal hemodynamic indices of resting autonomic cardiac control. Such stressors were associated with frequency domain and nonlinear HRV metrics. Lucini et al. [26] evaluated psychological involvement in students with exam-related anticipatory stress using a battery of questionnaires providing self-rated scales that focus on the appraisal of stress, coping, and health. Anticipatory stress was characterized by significantly higher values of LF_n. In a similar study, analysis of state anxiety and stress in a cohort of students just before a university exam and their nonlinear HRV revealed that anxiety is associated with alterations in the complexity of HRV captured as an increase in the short-term fractal exponent α₁[27]. Competitive sports, where adaptation to training loads and anticipation to compete causes anxiety, are another context of real-life anticipatory stress. Precompetitive anxiety before a swimming competition was studied [28] using HRV metrics and a Competitive State Anxiety Inventory-2 (CSAI-2). High precompetitive anxiety elicited a shift toward sympathetic predominance due to parasympathetic withdrawal. This study also concludes that RMSSD seems to be the most valid indicator of emotional state in pre-competitive situations among time-domain parameters.

In addition to anticipatory stress, other aspects, such as chronic fatigue, emotional exhaustion, and burnout, were also involved in the reviewed studies. During a blood-sampling procedure, emotional stress was examined and correlated with vagally mediated HRV in a large population-based sample of the Dresden Burnout Study [29]. The emotional exhaustion component of burnout, assessed using a Maslach Burnout Inventory (MBI), was associated with reduced vagal cardiac control during an emotionally arousing situation, reflecting an individual’s capacity to respond flexibly and adapt to changing environmental needs. Rajcani et al. 2016, investigated psychophysiological changes in the stress reaction, both in an experimental setting using PSST based on a simulated speech task and in a real-life setting by comparing a stressful and a relaxed day. Significantly diminished SDNN, LF/HF, RMSSD, and high-frequency power, all pointing to reduced HRV during a stressful state, have been observed in both settings. However, the effects in naturalistic settings were weaker but analogous to results based on experimental protocol [30]. The reduced HRV correlated with significant anxiety and perceived stress differences based on STAI and PSS scores. Observations in real-life settings have also been used by Carnevali et al. to evaluate the interplay between HRV and depressive and rumination symptoms causing chronic stress in young, healthy adults. In this study, RMSSD was negatively correlated with both rumination and depressive symptoms at each time point, implying that autonomic dysfunction, predominately low vagal tone, characterizes individuals with higher rumination traits and is prospectively linked to the generation of depressive symptoms in a non-clinical setting [31].

Experimental acute task-based stress induced by performing mental arithmetic, speech task, verbal fluency components of the TSST, Montreal Imaging Stress Task, Stroop test etc., have been used to investigate HRV-based physiological changes in nine of the reviewed studies. TSST is a standardized laboratory social stressor that induces robust and reliable increases in psychological, physiological, and neuroendocrine measures of stress. It is a helpful alternative to physical stressors and reproduces the more naturalistic psychological stress of performance in the presence of an evaluative audience [32]. In this review, Torino et al. employed only the mental arithmetic phase of TSST to study the degree of stress-associated changes in sympathetic activity and vagal tone using LF and HF components of HRV [33]. Mohammadi et al. observed stress before, during, and after TSST and noticed that HRV metrics change during stress tasks, but the changes persisted during recovery, i.e., after removing stress. This persistence of reduced HR complexity after stress is speculated to represent lower adaptability and a functional restriction of the participating cardiovascular elements. Gender differences have also been shown in this study with some significant correlations, such as an increase in LF/HF, and SD1 observed only in women [34].

Logan et al. [35] used only a single HRV metric, the HF-HRV to study autonomic function where TSST-induced stress. To quantify the effect of TSST, baseline measures of HF-HRV and state anxiety using STAI were compared with measurements obtained after the test, and no significant effects were observed. It must be noted that in this analysis, HRV before and after were only considered and HRV during stress was not included in the study. In a similar approach toward using HF-HRV, Spangler et al. used HF-HRV to assess cardiac vagal influences before, during, and after stress tasks. They demonstrated an inverse relationship between stress during tasks and HF power. The commonality seen in the analysis and findings of all these laboratory stressor-based studies is that induced stress causes the ANS to shift toward sympathetic predominance because of parasympathetic withdrawal, which provokes a characteristic defense-arousal reaction in the physiological system. A similar result was validated by Delaney et al [36]. In addition to demonstrating stress responses in a task-performing group, an intervention effect was also shown based on comparisons with a control group.

Schubert et al. performed an interesting study that demonstrated the influence of baseline chronic stress on the association between acute stress and HRV metrics. Short-term stressor reactivity was assessed with a speech task and various HRV metrics. In addition to other conventional measures, D2, a nonlinear HR complexity measure, significantly decreased with acute stress. The higher the HR D2, the more degrees of freedom of the cardiac pacemaker and, therefore, the greater range of possible adaptive responses [37]. This HR complexity measure was significantly inversely correlated with baseline chronic stress levels and further reduced during acute stress, suggesting that long-term chronic stress levels mediate or may be influenced by HRV responses to short-term stressors. Studies have also demonstrated the role of laboratory-based acute HRV metrics in predicting autonomic modulation of ecological emotional stress [38].

The literature shows that people with high trait anxiety tend to exhibit increased HR and diminished HRV [39, 40]. Two studies in this review have investigated the interplay of these trait features in HRV stress associations. Effortful control, a trait measure of adaptability to stress, was used by Spangler et al. (2009) to investigate the extent to which such trait-based self-regulatory constructs influence autonomic control of cardiac responses to acute stress. Resting cardiac vagal control was significantly correlated with effortful control, further influencing the vagal recovery after the verbal fluency stress task. The study by Medica-Torino et al. [33] had a similar approach in testing the influence of chronic stress, assessed using trait anxiety measure at baseline, on associations between state anxiety-based acute stress and HRV during experimental stressors. Significant associations were found between the level of state anxiety and worry associated with stress tasks and the degree of sympathetic activation and vagal tone reduction; however, no significant moderating role of trait anxiety was observed.

Due to a reduction in vagal tone, acute stress is generally associated with a typical increase in heart rate, which has been observed and reported in five reviewed studies [28, 35, 36, 41, 42]. Elevated blood pressure due to an increase in sympathetic drive is another physiological phenomenon observed during acute stress conditions which were reported in one of the reviewed studies [26].

As discussed above, variations in HRV due to acute stress, along with the underlying non-clinical yet persistent trait anxiety, depression, burnout, and chronic stressors to which individuals might adapt differently, might make the interpretations of observed HRV changes more complex. There is an enormous scope of well-designed studies that untangle the various layers of these stress responses and mediating factors which is a recommendation for future researchers.

It has been argued that acute psychological stress acts on cardiac autonomic regulation in a way that may lead to non-stationarities in the interbeat interval series, making linear methods, primarily the frequency domain-based metrics, unreliable compared to nonlinear approaches [43]. Despite the robustness and reliability of nonlinear metrics in revealing useful additional information about HRV characteristics in different applications and patient groups [44], nonlinear methods are not very popular in HRV studies investigating psychological stress. A robust nonlinear approach termed binary symbolic dynamics analysis [45], which considers signal non-stationarity, has been used by one study to detect acute stress-associated changes in HRV and is thought to be more informative than linear methods [46]. It is recommended in this review that more studies be designed to compare and prove the usefulness of several nonlinear metrics and their analysis windows, which inherently can overcome the limitations of non-stationarities.

In addition to non-stationarity, fine-grained stress assessments are precluded by longer analysis windows, especially in real-time unobtrusive measurement using sensors, wearables etc. Short-term analysis (≤5 min) thus would be an advantage for stress studies due to the rapid physiological response time. Discrimination of short-term stress conditions using short-term HRV metrics was explored by Pereira et al. [47]. They used the TSST protocol and quantified HRV at five time points (baseline, silence, reading, presentation, and counting phases), allowing five short-term time windows for evaluation. In this study, several linear and nonlinear metrics demonstrated reduced HRV during stress, with much higher reduction strength in time-domain metrics. In particular, AVNN, RMSSD, and SDNN measures using windows as short as 50 s have been shown as the most discriminating HRV metrics distinguishing stress and non-stress states [47]. More studies demonstrating the discriminating power of such short-term HRV in real-life stress detection are recommended for future research on wearable smart device-based health trackers. Brugnera et al. performed a similar protocol in experimental stress but on five-minute analysis windows. Their results indicate that stress led to a sympathetic shift in sympathovagal balance and reduced the complexity of the cardiac signal. Montreal Imaging Stress Task induced the strongest cardiovascular response compared to the other speech and Stroop tasks and was associated with a specific profile of cardiovascular activity, characterized by sympathovagal co-inhibition, i.e., decreases in both HF and LF power [42].

In this review, spectral metrics such as HF power, LF power, and LF/HF have been claimed to be stress indicators in studies (n = 3) involving a speech task. Spectral analysis-based HRV changes, especially during the speech-based task, are argued in the literature to be interpreted with caution, as the respiratory changes produced by speech markedly alter variability and the spectral component of HRV without necessarily involving respective changes in autonomic activation [48, 49]. Respiratory patterns observed during speech tasks have been shown far from being sinusoidal, highly erratic, and associated with markedly broadband characteristics in the HR power spectrum, with considerable power present in both the LF and HF bands. In this review, we observed that the analysis was controlled for respiratory frequency in only one study [41, 42]. In contrast, two studies acknowledged respiratory influences on HRV during speech tasks [47] and mentioned not controlling for breathing and non-stationarities as their study limitation [46]. Future studies are recommended to rigorously account for and control analysis for the influence of respiratory patterns in speech-based stress tasks to improve the validity of findings. As with many other reviews, this paper included only published journal articles, and a possible publication bias may have affected our findings. Also, the length of recordings obtained and analyzed varied between the included papers, and this might have influenced our findings and interpretations. However, due to the heterogeneity, we did not further group studies based on these data and compare the findings.

In this review, we focused on identifying specific HRV metrics and their association with change in psychological stress under different stress states in studies using repeated sampling protocol in healthy populations. It was believed that these associations would allow us to make inferences about the metrics most suited to act as physiological markers to monitor and track the impacts of psychological stress in different stress contexts, especially real-life stressors. The review does highlight the impact of psychological stress on sympathovagal balance which is closely reflected by HRV metrics, thus evidencing that HRV can be used as an informative marker of the physiological effects of psychological stressors in healthy adult populations. The traditional time-domain RMSSD is still the most explored and reported measure. Most of the reviewed studies were performed under laboratory conditions instead of natural working life settings. Few novel studies have considered both scenarios, explored nonlinear novel metrics, and the influence of underlying chronic stressors and trait characteristics on the effects observed under acute stress conditions. However, of the few, little has been validated with large samples. In addition to short-term laboratory measurements based on conventional metrics, acute real-life stress contexts, novel HRV metrics, long-term HR monitoring, repeated sampling, and accounting for underlying chronic stressors are imperative for validating HRV metrics. Such validated metrics can accurately assess and predict stress and recovery patterns, and such validation approaches warrant future research and large-scale study designs. In addition, HRV measurements in real-life ambulatory settings often involve known confounders such as physical exertion and other unidentified confounding factors that can rarely be controlled entirely but must be carefully considered in interpreting study results. These effects can be reduced to some extent by cross-validation of findings using consistent, validated subjective screening instruments with objective HRV quantification methods in large-scale study settings. In addition, this review identified the diversity of instruments used in studies assessing psychological stress and the heterogeneity in methodological approaches of quantifying HRV, which makes the interpretation of measurements inconclusive in the context of this review. In summary, utilization of standard stress theories/models, uniform validated stress indicators, and standardized methods would improve the comparability of results in future studies. More unified HRV assessment and analysis methods utilizing both conventional and contemporary metrics and longitudinal real-life study settings are needed.

An ethics statement is not applicable because this study is based exclusively on the published literature.

The authors have no conflicts of interest to declare.

There were no funding sources.

Sarah Immanuel: conception and design of the work, search criteria formulation, data acquisition, and analysis and interpretation of data for the work. Meseret Teferra: data extraction and screening and drafting of the work. Niranjan Bidargaddi: study design and drafting and critical revision of the work. Mathias Baumert: drafting, providing context, data synthesis, and critical revision for intellectual content.

All data generated or analyzed during this study are included in this article and its supplementary materials. Further inquiries can be directed to the corresponding author.