Ambulatory Physiological Assessment of Posttraumatic Stress Disorder: Integrating Passive Sensing with Ecological Momentary Assessment to Measure Trauma Reactivity

Abstract

The development of wearable technology affords objective measurement of physiological states outside the laboratory. We used ambulatory physiological assessment to measure overall arousal and reactivity to trauma reminders, a hallmark symptom of posttraumatic stress disorder (PTSD). Ambulatory assessment improves upon lab-based tests by measuring actual trauma reminders as they occur in everyday life. In this study, we recruited a mixed-trauma sample of 80 participants (39 diagnosed with PTSD) who completed three days of ambulatory physiological assessment time-synced to self-reported ecological momentary assessment of trauma reminders and contextual factors. We assessed heart rate (interbeat interval or IBI) as a nonspecific marker of overall physiological arousal, skin conductance and pre-ejection period (PEP) as markers of sympathetic activity, and respiratory sinus arrhythmia (RSA) as a marker of parasympathetic activity. We found that individuals with and without PTSD did not significantly differ on average levels of any physiological marker. Among individuals diagnosed with PTSD, IBI was significantly lower, indicating higher arousal, when participants were reminded of their trauma. Trauma reminders were not significantly associated with RSA or PEP. Skin conductance was significantly lower (indicating lower arousal) in the presence of trauma reminders, counter to predictions. For time-varying predictors, we found that trauma reminders were associated with less strong physiological responses, indexed by IBI, when reminders were perceived as controllable, and when they were experienced in the presence of social support. Our findings support heart rate as an inexpensive and accessible marker that can elucidate the role of contextual factors affecting PTSD symptom expression.

General Scientific Summary:

This study found that ambulatory heart rate is a promising marker of responding to trauma reminders, a key target of posttraumatic stress disorder treatment.

With advances in wearable technology, researchers can now noninvasively, passively monitor participants’ physiological activity as they go about their daily lives. The availability of such technology creates opportunities for a host of clinical and research applications with greater ecological validity than lab-based testing. Here we present the results of the first application of ambulatory assessment to the study of physiological reactivity to trauma reminders, a hallmark symptom of posttraumatic stress disorder (PTSD).

Overall Physiological Arousal

Like all psychological disorders, PTSD is assessed via subjective reports of symptoms, either through self-report questionnaires or clinical interviews (e.g., Bovin et al., 2016; Weathers et al., 2018). Physiological markers offer a potentially more objective assessment tool less influenced by common limitations of self-report, such as response biases, lack of insight, or demand characteristics (Furnham & Henderson, 1982). Prior laboratory research has found that individuals with PTSD show altered patterns of autonomic nervous system activity (Campbell et al., 2019; Pole, 2007). Specifically, people with PTSD show higher resting heart rate (a nonspecific marker of autonomic arousal), higher skin conductance levels (sympathetic arousal), and lower resting respiratory sinus arrhythmia (RSA, a marker of parasympathetic activity), with variability across studies but aggregate effects that are significant albeit small (Campbell et al., 2019; Pole, 2007). PTSD-linked differences in resting physiology are thought to reflect long-terms alterations in the autonomic nervous system caused by hyperarousal symptoms. Specifically, frequent activation of the sympathetic nervous system due to hypervigilance and in response to perceived threats leads to persistent alterations in resting physiology that are present even in the absence of environmental threats (Buckley & Kaloupek, 2001; Pole, 2007)

Ambulatory assessment can offer a more valid measure of average autonomic activity across contexts, compared to clinic or lab-based samples. A well-known phenomenon in clinical settings is the “white coat” effect, in which physiological recordings are artificially inflated due to stress related to the presence of a physician (Pioli et al., 2018). In the context of PTSD research, physiological recordings are frequently taken right before participants complete lab tasks related to their trauma, possibly leading to situation-specific increases in arousal (Buckley & Kaloupek, 2001). Limited research has used ambulatory heart rate recording to compare military veterans with and without PTSD (Beckham et al., 2000; Buckley et al., 2004), but findings have been mixed and have not included other ANS markers. Thus, one goal of the current study is to examine established autonomic markers of PTSD (heart rate, skin conductance, RSA) using ambulatory assessment to determine whether there are PTSD-linked differences similar to those seen in the lab. Another goal of the current study is to test a novel marker, pre-ejection period (PEP), which is a well-established cardiovascular marker of sympathetic activity, but rarely studied in PTSD (for exceptions, see Meyer et al., 2016; Sheikh et al., 2024).

Physiological Reactivity to Trauma Cues

In addition to alterations in resting levels, PTSD is marked by strong physiological reactions to internal and external trauma reminders (Pineles & Orr, 2018; Pole, 2007). These physiological reactions are believed to represent conditioned fear responses to cues that remind individuals of their trauma (Foa et al., 2006; Lang, 1979). That is, cues that would not evoke a fear response prior to the traumatic event begin to elicit subjective feelings of distress as well as physiological reactions (increased heart rate, sweating) following the trauma, through associative learning. For example, a rape survivor might develop a physiological fear response to the smell of the perpetrator’s cologne, an external reminder, or to an image of the perpetrator popping into her mind, an internal reminder. Indeed, a large body of research has demonstrated that individuals with PTSD show increases in physiological markers including heart rate and skin conductance in response to trauma cues (Pole, 2007). In fact, physiological reactions to trauma reminders are such an integral part of PTSD phenomenonology that they have been officially recognized as a symptom since DSM-III-R (American Psychiatric Association, 1987). Prominent theories of PTSD, such as the revised Emotional Processing Theory, argue that PTSD is caused by “fear structures” or “fear networks” that include previously neutral stimuli that now evoke strong fear responses, and these conditioned responses are maintained by avoidance of feared stimuli, which prevents corrective learning (i.e., learning that these stimuli are not in fact dangerous) from occurring (Foa et al., 2006). Exposure-based treatment for PTSD, such as Prolonged Exposure and Written Exposure Treatment, are based on facilitating corrective learning by exposing individuals with PTSD to feared stimuli (Foa, 2011; Sloan & Marx, 2019).

There has been great interest in using physiological responses to trauma cues as objective markers of fear responding, which could be used as a diagnostic aid or to track treatment progress. Despite the strong research base supporting physiological reactivity to trauma cues in PTSD, current research is limited by an exclusive focus on symptom evocation protocols in which trauma reminders are provided in the lab to evoke a physiological response (Pole, 2007). Such lab-based tasks are limited to a small number of stimuli that may differ from the actual experience of trauma reminders in the “real world,” which may be cued or uncued, internal or external, and may involve a wider array of stimuli (Kleim et al., 2013). An important question is whether PTSD is marked by physiological reactivity to trauma reminders when assessed in naturalistic environments, outside tightly controlled lab settings, which would greatly increase the clinical applications. For example, the primary goal of exposure-based therapies, frontline treatments for PTSD (Foa, 2011), is extinction of conditioned fear responses (Craske et al., 2014). For treatment to be successful, learning that occurs during therapy must generalize to other stimuli and settings. Ambulatory physiological assessment offers an ideal means of assessing to extent to which a client’s learning has generalized beyond the therapy room, but it has yet to be tested as a measure of trauma reactivity.

Contextual Factors Affecting Trauma Reactivity

Other limitations of symptom evocation tasks are that the type and number of trauma-related stimuli are typically limited. Symptom evocation tasks use either standardized trauma cues (e.g., combat pictures for a sample of combat veterans) or idiographic trauma cues (e.g., a trauma script) (Pole, 2007). Standardized trauma cues can be challenging to develop given the heterogeneity of cues that become conditioned stimuli, particularly for mixed trauma samples, and effect sizes are typically smaller for standardized than for idiographic trauma cues (Pole, 2007). Due to the limited range of available stimuli, it can be challenging to examine stimulus-related factors that moderate trauma reactivity. The studies that have examined moderation effects, using standardized trauma cues, have found that uncontrollable and unpredictable trauma cues elicit stronger physiological reactions relative to controllable and predictable trauma cues, suggesting that these factors may be important to consider when examining “real world” reminders (Grillon et al., 2008; Simmons et al., 2013; see also Foa et al., 1992).

Further, symptom evocation tasks are only presented in a single environment (the laboratory), precluding assessment of contextual factors. The basic fear conditioning literature has identified factors that influence conditioned fear responses, such as social context (i.e., the presence of others) and learning history with the environment (i.e., whether testing occurs in a similar or different environment to fear acquisition). However, it is unclear whether these basic findings, which typically use associative learning paradigms pairing aversive stimuli with simple, easily distinguishable cues, often in animals, will generalize to conditioned fear reactions in humans to complex trauma cues. Passive sensing offers the ideal opportunity to noninvasively monitor physiological reactions as individuals with PTSD encounter a range of trauma cues in a variety of contexts, allowing assessment of stimulus-related and environmental factors that moderate physiological reactivity to trauma cues in this important clinical population.

The Current Study

Thus, the purpose of this study was to (1) examine whether trauma-exposed individuals with and without PTSD differ in average overall (IBI), sympathetic (PEP and SC), and parasympathetic (RSA) arousal levels using ambulatory assessment in naturalistic settings, (2) examine whether “real world” trauma cues are associated with heightened physiological arousal among individuals with PTSD, and (3) probe different contextual factors (predictability, controllability, safety, and social support) that might affect physiological reactivity among individuals with PTSD. Specifically, we predicted that trauma reminders that are more predictable and controllable, experienced in safer environments, and experienced in the presence of more social support, will be associated with lower physiological arousal.

Method

Participants and Procedure

Eight-five trauma-exposed individuals were recruited and consented to participate in the Ambulatory Physiological Assessment of PTSD study. Participants were recruited from a midsized city in the southeastern United States using a variety of sources, including flyers posted in the community, advertisements posted in local newspapers and online community boards, and from a psychology clinic. Five participants were excluded due to reporting an exclusion criterion after consenting (n=1) or not completing all study procedures (n=4), yielding a final sample of 80 participants (39 diagnosed with PTSD, 41 without current PTSD). Table 1 presents demographic and trauma-related characteristics.

Table 1.

Demographic Characteristics in the Full Sample and by Diagnostic Status

Variable	Full Sample (n = 80)	TEC (n = 41)	PTSD (n = 39)
Demographic Characteristics
Age (M, SD)	21.79 (4.21)	22.22 (5.06)	21.33 (3.07)
Gender (n, % woman)	60 (75%)	31 (75.6%)	29 (74.4%)
Race/Ethnicity (n, %)
White (Not Hispanic)	28 (35%)	18 (43.9%)	10 (25.9%)
Black/African American	23 (28.7%)	9 (22%)	14 (35.9%)
Hispanic/Latino	17 (21.3%)	9 (22%)	8 (20.5%)
Asian/Pacific Islander	2 (2.5%)	0	2 (5.1%)
Biracial	4 (5.0%)	2 (4.9%)	2 (5.1%)
Other/Multiracial	6 (7.5%)	3 (7.3%)	3 (7.7%)
Education (n, %)
High School Diploma	14 (17.5%)	7 (17.1%)	7 (17.9%)
Associate’s Degree	9 (11.3%)	5 (12.2%)	4 (10.3%)
Some College	41 (51.2%)	22 (53.7%)	19 (48.7%)
Bachelor’s Degree	8 (10.0%)	4 (9.8%)	4 (10.3%)
Graduate Degree	7 (8.8%)	3 (7.3%)	4 (10.3%)
Trauma Characteristics
Index Trauma Type (n, %)
Sexual Assault	34 (42.5%)	17 (41.5%)	17 (43.6%)
Physical Assault	19 (23.8%)	8 (19.5%)	11 (28.2%)
Natural Disaster/Accident/Fire	13 (16.3%)	8 (19.5%)	5 (12.8%)
Serious Illness/Injury/Death	13 (16.3%)	7 (17.1%)	6 (15.4%)
Combat Exposure	1 (1.3%)	1 (2.4%)	0
Months Since Trauma (M, SD)	58.11 (53.19)	59.72 (50.63)	56.21 (56.79)
PTSD Severity (CAPS-5; M, SD)	19.81 (12.07)	10.09 (7.07)	30.08 (6.25)*

Note. TEC = Trauma-Exposed Control; PTSD = Posttraumatic Stress Disorder; CAPS-5=Clinician Administered PTSD Scale for DSM-5 (Weathers et al., 2013c); Education level is missing for one participant, and months since trauma could not be determined for 13 participants.

^*denotes a statistically significant difference between TEC and PTSD groups (p < .05).

Inclusion and exclusion criteria were guided by best practices for psychophysiological data acquisition. Participants were eligible if they were between 18 and 40 years old, endorsed exposure to at least one traumatic event (meeting DSM-5 Criterion A for PTSD) that occurred more than one month prior to participation, and had a body mass index (BMI) between 18.5 and 34.9. Exclusion criteria were psychosis, trauma exposure within the past month, pregnancy, history of cardiovascular disease, and use of medications that affect cardiovascular functioning (including antidepressants and antihistamines). Participants who reported high levels of dissociative symptoms on a prescreening measure were also excluded, due to evidence that physiological reactivity may differ among individuals with the dissociative subtype of PTSD (Lanius et al., 2006).

Participants were first screened to determine inclusion/exclusion criteria and to oversample individuals with PTSD. The online Qualtrics prescreen included demographic and health-related questions, the Life Events Checklist-5 (LEC-5), used to verify trauma exposure, the PRIME screen for psychosis, and the PTSD Checklist-5 with two items added to assess dissociative symptoms (Tsai et al., 2015). Participants who met eligibility based on their responses then completed a follow-up phone screening to verify that the index trauma endorsed on the LEC-5 met the DSM-5 Criterion A definition of trauma. Participants with probable PTSD (PCL-5 scores of 33 or higher (Bovin et al., 2016)) were oversampled to ensure an approximately even ratio of participants with and without PTSD, stratified by trauma type, age, race/ethnicity, and BMI to ensure matching on these variables.

Eligible participants came into the lab for their first session, where they provided informed consent, completed structured clinical interviews including the CAPS-5, and completed two standardized study scripts. Following the first laboratory session, participants completed three days of ambulatory assessment. Each ambulatory assessment day started with a visit to the lab to be connected to the Mindware Mobile device and receive a Lenovo Android tablet to collect their ecological momentary assessment (EMA) data. We provided participants with experimenter-owned devices, rather than having them use their own devices, to reduce possible data loss and facilitate time-syncing with the Mindware Mobile device. The use of tablets rather than smaller devices, like smartphones, can also increase participant compliance because they are more noticeable to participants. On the first ambulatory assessment visit, they completed a practice survey and had the opportunity to ask questions. Tablets were pseudorandomly configured to administer prompts within 90-minute windows starting at 9:00 am and ending at 11:30 pm each evening, for up to 17 prompts per day (the number of administered prompts varied depending on the time participants started their study day and went to sleep, the mean number of administered prompts was 42.76 per participant). Participants had 20 minutes after the prompt to complete the survey. Mean time between prompts was 53.67 min (SD=13.34, range 31–73 min). Participants were shown how to remove and store the Mindware device, which they were instructed to do on their own at home when they were ready to go to bed that night. The participants then returned to the lab on the morning of the next two ambulatory assessment days, returned their equipment, and received freshly charged devices for their next day of ambulatory assessment. Finally, participants returned to the lab for their final lab session which consisted of a script-driven imagery procedure and additional questionnaires (not analyzed here). They were compensated $150 USD.

Measures

Experience Sampling Questions.

All experience sampling items were preceded by the prompt “during the past ten minutes…” and rated on Likert-type scale from 1=“not at all” to 7=“very much.” Trauma reminders were assessed via four items designed to index internal and external cues: “I was thinking about the trauma,” “I had unwanted memories of the trauma,” “I relived the trauma as though it were actually happening again,” and “something reminded me of the trauma.” These four items were recoded to range from 0–6 each and then summed to create a trauma reminder composite (range of 0–24). In this sample, reliability values were ω_between=.95 and ω_within=.85, calculated using the approach described by Geldhof et al., (2014). Because the composite variable was highly skewed, it was recoded as a dichotomous variable (scores of 0 coded as 0=no trauma reminders, any score >0 coded as 1=trauma reminders). Contextual factors were assessed with a single item each: predictability (“I was expecting to think about the trauma”), controllability (“I could control my thoughts about the trauma”), social support (“I was with supportive people”) and location safety (“I was in a place where I usually feel safe”). Additionally, caffeine use and physical activity were assessed with yes/no responses to single items (“Since the last signal, have you had caffeine?” and “Since the last signal, have you been physically active?”).

Clinician Administered PTSD Scale-5 (CAPS-5).

PTSD symptom severity and diagnostic status was assessed using the CAPS-5 (Weathers et al., 2018). The CAPS-5 assesses exposure to a Criterion A event and the severity of each of 20 PTSD symptoms on a scale from 0 (absent) to 4 (extreme/incapacitating). PTSD diagnostic status was determined using DSM-5 diagnostic criteria with items scored as “2” or higher counting as symptoms. Symptom scores are summed to give an overall PTSD symptom severity score ranging from 0 to 80. PTSD chronicity was assessed by asking participants how long their symptoms have lasted (in months). All interviews were administered by a trained graduate student and audiorecordings were reviewed by a second coder for reliability coding, with discrepancies reviewed by the principal investigator. Interrater reliability for total CAPS-5 score was very high (ICC=.99).

Traumatic Life Events Questionnaire (TLEQ).

The TLEQ was administered to assess lifetime trauma history. The TLEQ queries 21 potentially traumatic events and, for each event, asks how many times the event occurred. The number of event occurrences are summed to give the total number of prior traumas.

Psychophysiological Data Collection.

We had four primary physiological outcomes: interbeat interval (IBI), RSA, PEP, and skin conductance. IBI refers to the number of milliseconds between R peaks on an electrocardiogram and is inversely proportional to heart rate in beats per minute. We analyzed IBI instead of heart rate consistent with the Society for Psychophysiological Research guidelines (Quigley et al., 2024). RSA is the linear transform of high frequency power calculated using spectral analysis with high frequency band of 0.12–0.40 Hz. PEP refers to the number of milliseconds between electrical activation of the ventricle and opening of the aortic valve. Skin conductance refers to total skin conductance level in microsiemens. Due to the nature of ambulatory recording, we selected skin conductance level rather than skin conductance responses (SCRs) as our outcome variable. Thresholding standards for SCRs have yet to be developed for ambulatory recording, and we did not have enough information about the environment to determine whether SCRs are stimulus-related or nonspecific. Respiration rate was also derived from the impedance signal (Ernst et al., 1999) for use as a covariate in some analyses.

Data collection procedures.

Ambulatory physiological data were collected continuously across multiple channels with a sampling rate of 500 Hz using Mindware Mobile impedance cardiograph devices (Mindware Technologies, Gahanna, OH). Heart rate and high-frequency heart rate variability data were collected via electrocardiogram using disposable 1.5” diameter Ag/AgCl electrodes with 7% chloride wet gal connected to the participant’s chest in a standard lead III configuration. Impedance cardiography (PEP) was collected via four additional electrodes on the participant’s chest and back. Skin conductance was collected via disposable 1.5”×1” inch rectangular Ag/AgCl electrodes with 0.5% chloride wet gel attached to the back of the participant’s shoulder. Placement site for skin conductance was determined based on impracticality of finger/palmar recordings for ambulatory assessment and strong correspondence between shoulder and finger readings in prior research (van Dooren et al., 2012). All leads were individually taped to the participant’s body to minimize interference. We selected a chest-worn device over more portable wrist-worn devices for two reasons: 1) to allow measurement of impedance cardiography (not available via wrist-worn devices at the time of data collection) and 2) to optimize cleaning of RSA data (the Mindware Mobile device allows visualization of the ECG waveform unlike most wrist-worn devices).

Data Preprocessing.

Data were edited and analyzed using the Mindware suite of software (HRV, IMP, and EDA Analysis). Physiological data were collected continuously throughout the day, but only data corresponding to the ten minutes prior to each completed survey were analyzed. Physiological data were split into 60 second epochs for cleaning and artifact correction, and the ten epochs prior to each beep were averaged to give one observation per completed survey (i.e., one value each for IBI, RSA, SC, and PEP per beep).

For IBI and RSA, R peaks were identified and potential artifacts were flagged using the Mindware IBI and MAD/MED algorithms. All R peaks were then visually inspected and any misspecified R peaks were manually corrected by a trained rater. Artifacts (e.g., missed beats) were corrected using the midbeat function for RSA. Epochs requiring artifact correction for 10% or more of R peaks or with less than 30 seconds of continuous useable signal were excluded from analysis. For cardiac impedance, data were manually inspected and edited so that only clearly identified R peaks and dZ/dt cycles were retained in the ensemble averages (artifacts in RR intervals were dropped rather than corrected); only epochs that included 50% or more useable data were retained. B Points were calculated using the method “percent of dZ/dt Time + C” (Lozano et al., 2007). Skin conductance signals were manually inspected for signal loss (i.e., a sudden drop in amplitude) and we excluded any epochs with values less than 1 or greater than 20. Any epochs with questionable signal quality were first reviewed by the principal investigator and then by an expert psychophysiology consultant from Mindware Technologies prior to exclusion.

Data Analysis Plan

Data management was done using IBM SPSS v18; multilevel models were run using MPlus version 8.6. All multilevel models were run with observations (level 1) nested within people (level 2) using maximum likelihood with robust standard errors. All regression weights are unstandardized. We planned separate means-as-outcomes models examining PTSD diagnostic status as a level 2 (between person) predictor of average physiological arousal. We ran separate models for each physiological indicator (IBI, PEP, RSA, and skin conductance), and ran robustness checks controlling for key demographic and health-related factors (age, sex, BMI). For RSA, we also included respiratory rate as a covariate.

We then analyzed the effects of trauma reminders as a level 1 (time-varying) predictor of physiological arousal in the subsample of participants diagnosed with PTSD. We again ran separate models for each physiological outcome (4 total); all were run with random intercepts and fixed slopes. Finally, we examined the four contextual variables (predictability, control, safety, and social support) as potential moderators of the association between trauma reminders and each physiological outcome in participants with PTSD. The moderation analyses were run with trauma reminders, the moderator, and their interaction as level 1 predictors. Separate models were run for each moderator and for each physiological outcome (16 total); all predictors were person-mean-centered and all models were run with random intercepts and fixed slopes. We decided a priori to limit the analyses concerning trauma reminders to participants with PTSD because we are most interested in generalizing these findings to individuals diagnosed with PTSD. We anticipated that reactions of people with and without PTSD may be qualitatively different, and that the base rate of trauma reminders among individuals without PTSD might be too low for these analyses. We planned robustness checks controlling for day, time of day, caffeine use, and activity level (and respiration rate for RSA) at Level 1 (within-participant) and age, sex, BMI, number of past traumas (from the TLEQ), and PTSD chronicity (from the CAPS-5) at Level 2 (between-participant).

An a priori power analysis was run prior to data collection to determine necessary sample size. We used Monte Carlo power simulations in MPlus to estimate power to detect significant effects at α=.05 (Muthén & Muthén, 2002). Pilot data was collected from ten participants to use as seed data for parameter estimates (pilot data not presented here). All coefficients of interest were powered at .80 or higher with a sample size of 80 (half with PTSD) and 24 completed surveys per participant; analyses involving the subset of participants diagnosed with PTSD were powered at .80 or higher with a sample size of 40.

Transparency and Openness

We report how we determined our sample size, all data exclusions, and all manipulations in this study, which was not preregistered. All data collected in the study will be publicly available from the National Data Archive (www.nda.nih.gov) following an embargo period, and output files for primary analyses are included with this article as online supplemental material. Other publications present secondary analyses, addressing different research questions, of some of the self-report data presented here; the physiological data have not been published elsewhere (Nester & Wisco, 2024; Pugach, May, et al., 2023; Pugach, Starr, et al., 2023). The first author presented a talk as part of a symposium at the 2024 annual meeting of the International Society for Traumatic Stress Studies that included these findings.

Results

Participants completed 2,158 surveys, or an average of 26.98 surveys per participant, for an overall compliance rate of 63.1% (SD=19.7). After excluding data due to lack of signal or excessive artifact, 1884 (87%) useable observations remained for IBI and RSA, 1641 (76%) for PEP, and 1753 (81%) for skin conductance. Descriptive statistics are presented in Table 2 and bivariate correlations are presented in Table 3. Participants with PTSD reported significantly more trauma reminders than trauma-exposed control participants, B=0.22, SE=0.05, p<.001. Intraclass correlations (ICCs) of the empty models indicated that the amount of variance attributable to within-person effects ranged from 43.4% (skin conductance) to 63.4% (IBI).

Table 2.

Ambulatory Assessment Descriptive Statistics

Variable	Full Sample (n = 80)		TEC (n = 41)		PTSD (n = 39)		ICC
Level 1 (Within-Person)	M	SD	M	SD	M	SD
Interbeat Interval	707.12	116.11	706.27	114.73	708.16	117.92	.37
RSA	5.81	1.31	5.75	1.30	5.88	1.32	.36
PEP	94.10	14.71	95.04	15.92	93.08	13.22	.51
Skin Conductance	6.98	6.03	7.22	6.50	6.69	5.43	.57
Trauma Reminders	0.28	0.45	0.18	0.38	0.39	0.49	.30
Predictability	1.48	1.17	1.37	1.09	1.60	1.23	.38
Control	3.99	2.47	4.11	2.56	3.85	2.37	.48
Safety	5.51	1.76	5.65	1.77	5.36	1.74	.40
Social Support	3.55	2.49	4.13	2.49	2.92	2.33	.33

Note. ICC = intraclass correlation coefficient. The trauma reminders variable was a dichotomous variable scored a 0=no reminders, 1=reminders present; predictability, control, safety, and social support were indexed via single items with a possible range from 1 to 7.

Table 3.

Bivariate Correlation Matrix

	1	2	3	4	5	6	7	8	9
1. Interbeat Interval	-	.49*	.26*	.11	.08	−.05	.22*	−.05	−.03
2. RSA	.79*	-	.13	.07	.13	−.03	.12	−.03	−.20
3. PEP	.48*	.38*	-	.01	−.08	.02	.16	.33*	.24*
4. Skin Conductance	−.13*	−.11*	−.10*	-	−.03	−.20*	−.11	−.06	−.04
5. Trauma Reminders	−.07*	−.04	−.03	−.05*	-	.53*	−.04	−.20	−.32*
6. Predictability	−.08*	−.05*	−.01	−.05*	.28*	-	−.05	−.11	−.07
7. Control	−.05	−.05*	−.02	−.02	.02	.13*	-	.39*	.52*
8. Safety	.14*	.11*	.10*	−.02	−.06*	−.03	.10*	-	.54*
9. Social Support	−.10*	−.02	−.05	.02	−.04*	.05	.04	.14*	-

Note. Correlations above the diagonal represent between-person correlations, below the diagonal represent within-person correlations.

^*p < .05

Average Physiological Arousal

The means-as-outcomes models indicated that individuals with and without PTSD did not significantly differ on average levels of any physiological indicator. Specifically, PTSD status was not significantly associated with IBI, β=−3.73, SE=16.74, p=.824, skin conductance, β=−0.28, SE=0.66, p=.668, PEP, β=−0.71, SE=2.54, p=.780, or RSA, β=0.11, SE=0.18, p=.539. Robustness checks controlling for age, sex, BMI, and respiratory rate (for RSA) showed the same pattern of results.

Effects of Trauma Reminders on Physiological Arousal

We next examined the effects of trauma reminders on each of our physiological indicators among the subsample diagnosed with PTSD. We found that trauma reminders were significantly associated with IBI, β=−16.10, SE=6.74, p=.017. Consistent with predictions, in moments when participants with PTSD were thinking about their trauma, their physiological arousal was greater (indicated by shorter IBIs corresponding to higher heart rate). In contrast, trauma reminders were not significantly associated with either of the other cardiovascular markers: RSA, β=−0.07, SE=0.08, p=0.405; PEP, β=−1.21, SE=0.87, p=0.164. Surprisingly, trauma reminders were associated with significantly lower skin conductance levels, β=−0.44, SE=0.18, p=.016, the opposite of predictions. However, this effect did not hold up during the robustness check adjusting for day, time of day, caffeine use, and activity level at Level 1 and age, sex, BMI, number of past traumas, and PTSD chronicity at Level 2, adjusted β=−0.20, SE=0.17, p=0.239. Including these covariates (and respiration for RSA) did not affect statistical significance of any other findings.

Contextual Moderators

We then examined whether any of the contextual moderators affected the strength of the association between trauma reminders and physiological arousal. First, we examined moderation effects for IBI. Both the controllability of the trauma reminders and the presence of social support significantly moderated the association between trauma reminders and IBI, indicated by significant interactions in these models (Table 4). The more that participants reported being around supportive others, the weaker the association between trauma reminders and IBI (Figure 1). Similarly, the more control over trauma-related thoughts the participant perceived, the weaker the association between those thoughts and IBI (Figure 2). Neither the predictability of the reminders nor situation safety significantly moderated this association. However, there was a significant association between situation safety and IBI, such that the safer the participant felt, the lower their physiological arousal (indicated by a positive association, see Table 4). Robustness checks indicated that both interactions survived correction for covariates.

Table 4.

Trauma Reminders and Contextual Moderators

	Interbeat Interval		RSA		PEP		Skin Conductance
	R2		R2		R2		R2
	B (SE)	p	B (SE)	p	B (SE)	p	B (SE)	p
Controllability Models	.015		.005		.004		.013
Trauma Reminders	−15.99 (6.73)	.018	−0.07 (0.08)	.411	−1.24 (0.88)	.158	−0.44 (0.18)	.013
Controllability	−5.66 (1.81)	.002	−0.05 (0.03)	.082	0.12 (0.25)	.625	0.13 (0.13)	.345
Trauma x Controllability	9.06 (3.20)	.005	0.07 (0.04)	.107	0.04 (0.42)	.928	−0.25 (0.18)	.171
Predictability Models	.012		.003		.009		.015
Trauma Reminders	−12.81 (7.05)	.069	−0.04 (0.09)	.634	−1.51 (0.90)	.092	−0.30 (0.18)	.099
Predictability	0.97 (5.41)	.858	−0.02 (0.06)	.800	1.32 (0.47)	.005	−0.09 (0.13)	.480
Trauma x Predictability	−11.51 (7.91)	.145	−0.06 (0.09)	.504	−1.30 (0.64)	.044	−0.25 (0.17)	.134
Safety Models	.032		.027		.012		.017
Trauma Reminders	−15.75 (6.87)	.022	−0.06 (0.08)	.444	−1.21 (0.85)	.155	−0.72 (0.41)	.074
Safety	9.52 (3.53)	.007	0.11 (0.04)	.010	0.71 (0.39)	.069	−0.31 (0.16)	.061
Trauma x Safety	3.24 (4.51)	.473	0.04 (0.06)	.475	−0.09 (0.57)	.870	−0.03 (0.18)	.865
Support Models	.022		.003		.011		.007
Trauma Reminders	−16.86 (6.68)	.012	−0.07 (0.08)	.400	−1.31 (0.88)	.137	−0.72 (0.39)	.064
Support	−7.05 (2.26)	.002	−0.03 (0.03)	.385	−0.41 (0.21)	.051	−0.05 (0.07)	.445
Trauma x Support	9.22 (3.33)	.006	0.05 (0.04)	.187	−0.07 (0.49)	.880	0.02 (0.13)	.856

Note. R² values reflect the proportion of the variance explained by the model at the within-level. RSA = respiratory sinus arrhythmia, PEP = pre-ejection period. Parameter estimates are unstandardized and models are not adjusted for covariates.

Figure 1. Momentary Association Between Trauma Reminders and Social Support on Interbeat Interval.

Note. Interbeat interval (IBI) is inversely associated with heart rate, such that shorter IBIs reflect greater physiological arousal. Error bars reflect 95% confidence intervals.

Figure 2. Momentary Association Between Trauma Reminders and Controllability on Interbeat Inverval (IBI).

Note. Interbeat interval (IBI) is inversely associated with heart rate, such that shorter IBIs reflect greater physiological arousal. Error bars reflect 95% confidence intervals.

None of the contextual variables significantly moderated the associations between trauma reminders and RSA or between trauma reminders and skin conductance. A significant interaction between trauma reminders and predictability did emerge for PEP, such that the effect of trauma reminders on PEP was stronger for more predictable reminders, contrary to our hypotheses. No other significant interactions emerged for PEP.

Situation safety also emerged as a significant predictor of RSA, such that the safer the participant felt, the greater their parasympathetic activity (indicated significantly higher RSA values). Situation safety did not significantly predict PEP or skin conductance, but observed effects were in the expected direction for both variables (greater safety associated with lower sympathetic arousal).

Discussion

The goals of this study were to examine PTSD-linked differences in average levels of physiological arousal and physiological reactivity to trauma reminders in PTSD using ambulatory assessment in naturalistic settings. We found that individuals with and without PTSD did not differ in average levels of physiological arousal for any indicator, but heart rate emerged as a promising marker of trauma reactivity among those diagnosed with PTSD. Further, controllability and social support were important moderators of heart rate reactivity to trauma reminders in PTSD, highlighting the importance of assessing these factors in future work.

Overall Physiological Arousal

Contrary to our predictions, trauma-exposed individuals with and without PTSD did not differ significantly on mean levels of any autonomic marker in daily life. This stands in contrast to lab-based results, which have shown small but significant differences between trauma-exposed individuals with and without PTSD on several autonomic indicators, including resting heart rate, skin conductance level, and respiratory sinus arrhythmia (Campbell et al., 2019; Pole, 2007). The discrepancy between our finding and prior work could be due to the greater noise associated with ambulatory assessment compared to recordings taken in tightly controlled laboratory settings, possibly limiting our ability to detect significant effects. Our findings are also consistent with the possibility that PTSD-linked differences in lab-based assessment are driven more by anticipatory anxiety related to upcoming trauma-related content, rather than true differences in “resting” physiology. Further research is needed to examine these different possibilities, but our findings suggest that focusing on trauma reactivity may be more fruitful than average arousal levels in future ambulatory studies in this population.

Effects of trauma reminders on physiological indicators

Among individuals diagnosed with PTSD, trauma reminders were significantly associated with ambulatory IBI. This effect is consistent with predictions from the revised emotional processing theory of PTSD, which posits that trauma reminders elicit conditioned fear responses including strong physiological reactions, such as increased heart rate, among those with PTSD. Past research has supported this theory with retrospective self-report measures and with symptom evocation tasks meant to mimic trauma intrusions in the lab. To our knowledge, this is the first demonstration of objectively-measured physiological reactivity to naturally occurring trauma reminders among individuals with PTSD, offering strong support for the clinical relevance of this phenomenon. In contrast to IBI, trauma reminders were not significantly associated with either RSA or PEP, although both effects were in thedirections expected by theory. The null findings for RSA and PEP leave the physiological basis for the observed change in IBI unclear (i.e., whether the increase in heart rate was driven by sympathetic activity, parasympathetic activity, or both) and suggest that future research should examine these questions in larger samples to adequately power analyses of ambulatory RSA and PEP. In contrast, skin conductance was significantly associated with trauma reminders, but the observed effect was in the opposite direction as expected and the effect did not survive adjustment for relevant covariates. We do not interpret these results as contradicting emotional processing theory, but rather interpret these results as initial evidence that ambulatory skin conductance may not be a reliable marker of trauma reactivity among individuals diagnosed with PTSD, perhaps due to measurement difficulties outside tightly controlled laboratory settings.

Therefore, of the four indicators of autonomic activity examined in this study, heart rate emerged as the best marker of trauma reactivity among individuals diagnosed with PTSD. Heart rate also has the benefits of being inexpensive to monitor, easily accessible via a range of devices, and highly scalable. As mentioned, this is the first study to examine heart rate as a marker of trauma reactivity via passive sensing in a sample of individuals diagnosed with PTSD. The highly promising results open up numerous opportunities for future research, as discussed below.

Moderation of trauma reactivity by context

Because heart rate (IBI) emerged as the most robust physiological marker of trauma reactivity, we focus our discussion of contextual factors on this outcome. We found that both controllability and social support moderated heart rate response to trauma reminders among those with PTSD, whereas predictability and location safety did not. In terms of controllability, the less controllable the trauma reminders were perceived to be, the greater the heart rate reactivity. This finding corresponds with basic fear conditioning research showing that uncontrollable stressors lead to greater distress and enhanced fear conditioning (Foa et al., 1992; Grillon et al., 2008; Simmons et al., 2013). This finding also corresponds to research on the phenomenology of trauma intrusions, in which their uncontrollable nature is a key component (Brewin et al., 2010; Ehlers et al., 2004).

Social support also significantly moderated the association between trauma reminders and heart rate. During moments when participants were around supportive others, the association between trauma reminders and IBI was weaker. This finding has important implications for our understanding of trauma reactivity in PTSD. Social support is one of the most robust protective factors for PTSD (Charuvastra & Cloitre, 2008). Our findings suggest that the presence of supportive others reduces heart rate reactivity to trauma cues in one’s day-to-day life, suggesting a potential physiological mechanism through which social support might exert its protective effects. Paradoxically, this finding also suggests that overreliance on social support to manage trauma reactivity could maintain PTSD symptoms. The use of safety behaviors, including the presence of a supportive person, can maintain anxiety by serving as a subtle form of avoidance (Helbig-Lang & Petermann, 2010). If trauma-related situations can only be approached in the presence of supportive others, then PTSD symptoms are likely to be maintained. Interestingly, the presence of social support (or a “safe person”) is widely recognized as a potential safety signal in certain disorders, such as panic disorder and agoraphobia (Carter et al., 1995), but has received less attention in the context of PTSD (Blakey et al., 2020). This finding can also inform the implementation of exposure-based therapies for PTSD (Foa, 2011), as therapists might consider whether clients are relying upon social supports as safety signals during exposure exercises.

Contrary to predictions, neither the predictability of trauma reminders nor the perceived safety of the environment moderated the effects of trauma reminders on heart rate. These results indicate that participants with PTSD showed similar heart rate reactivity to trauma reminders whether or not they were expecting to think about the trauma, and whether or not they were in places where they usually feel safe. However, these null findings should be interpreted in light of some measurement considerations. Examination of mean predictability scores (Table 1) show that participants with PTSD reported infrequently expecting to think about their trauma during their daily lives, with a mean of 1.60 on a scale with “1” reflecting “not at all.” This is somewhat surprising because these participants were enrolled in a study which asked them to answer trauma-related questions multiple times per day, yet they still reported rarely expecting to think about the trauma. In terms of location safety, a limitation is that we only asked participants whether they typically feel safe in the location, not how similar the location is to the place where the trauma occurred, which might better capture conditioned fear responses. The range of location safety ratings may also have been restricted due to participants’ avoidance, a cardinal feature of PTSD. Future research examining these contextual factors should sample across a larger number of days to capture more instances of predictable trauma reminders and collect more fine-grained information about location, perhaps assessing a range of locations relevant to the index trauma (e.g., highways or parking lots in cases of PTSD resulting from motor vehicle accidents), or incorporating other passive sensing techniques such as GPS tracking.

Although there were no significant interactions between location safety and trauma reminders, location safety itself was a fairly robust predictor of physiological arousal. Location safety was significantly associated with higher levels of overall arousal (IBI) and with lower levels of parasympathetic activity (RSA). This finding suggests an effect of self-perceptions of safety on physiological arousal assessed via ambulatory methods and indicates that this variable should be included in future ambulatory physiological assessment studies.

Constraints on Generality, Limitations, and Future Directions

The generalizability of these results is limited by the nature of our sample, which was predominantly female, civilian, and with sexual assault as the most frequently reported index trauma. Results may not generalize to men, military and veteran populations, or other index trauma types. We also excluded participants older than 40 and those taking antidepressant medications, further limiting generalizability. In this study, we relied on a mix of passive sensing (of physiological responding) and ecological momentary assessment (of trauma reminders) to measure trauma reactivity. It is possible that being queried multiple times per day about trauma-related symptoms prompted trauma reminders (i.e., experimental reactivity). This concern is somewhat mitigated by the fact that participants typically reported that they were “not at all” expecting to think about their trauma, which we would expect to be higher if the questionnaires themselves were eliciting trauma intrusions. But, ideally, assessment of trauma reminders could be respondent-driven or even fully automated (e.g., through digital phenotyping) to reduce the possibility of reactivity, participant burden, and other limitations of self-report. Another limitation of our study was that participants completed surveys on experimenter-provided tablets, rather than their personal smartphones. Tablets have reduced portability and may have changed compliance or affected participant behavior in unknown ways. Collection of physiological data in uncontrolled environments leads to noise (e.g., effects of ambient temperature or movement) that likely reduced our ability to detect significant effects.Finally, participants had to come into the lab to be fitted to the ambulatory device each morning, reducing scalability of our specific methods. This was necessary given best practices for ambulatory impedance cardiography, used to measure PEP. Given our finding that heart rate was the most robust correlate of trauma reactivity, future research could focus exclusively on heart rate reactivity to trauma cues, enhancing scalability by using wearable technology that is already widely available and inexpensive.

The availability of heart rate as a scalable, easily accessible ambulatory marker of trauma reactivity in PTSD opens exciting new avenues for future research. Research could focus on characteristics of trauma intrusions (e.g., vividness, “here and now” quality) that affect heart rate reactivity, the effectiveness of different emotion regulation strategies in downregulating heart rate reactivity to trauma reminders, or correspondence between self-reported and physiological markers of trauma reactivity. These findings could be replicated in military veterans and could be extended to include ambulatory assessment of combat-related constructs such as moral injury and their interactions with heart-rate reactivity. Ambulatory heart rate monitoring of trauma reactivity could also be used to measure PTSD treatment progress. A key goal of exposure-based therapies for PTSD is for learning to generalize beyond the therapy room to the participant’s daily life. The ambulatory assessment approach tested in this study could measure generalization of learning directly with an objective marker of fear extinction.

Conclusions

We found that individuals with PTSD experienced higher heart rate during moments with trauma reminders, as assessed via ambulatory physiological assessment as they went about their daily lives. Controllability of the trauma reminders and social support significantly moderated heart rate reactivity, such that heart rate reactivity was lower for controllable reminders and for reminders experienced in the presence of social support. Heart rate is an inexpensive, easily accessible, and highly scalable marker, and we hope that this demonstration of ambulatory heart rate monitoring as a marker of trauma reactivity in PTSD will encourage future research using this method to test important clinical questions with greater ecological validity.