Feasibility of wearable sensors in the NICU: Psychophysiological measures of parental stress

Abstract

Background:

Parents report elevated stress during their infant’s NICU hospitalization. Real-time measures may improve our understanding of parental stress in the NICU.

Aim:

Examine the feasibility of wearable sensors to describe parental stress in the NICU.

Study design:

In this prospective feasibility study of 12 parent-infant dyads, parents wore an Empatica E4 wristband to measure psychophysiological stress via electrodermal activity (EDA) during sensory interventions (holding, massage, reading, touch, etc) with their babies. Baseline and intervention periods were delineated during which concurrent monitoring and clinical observations of infant behavior and environmental factors were recorded. Feasibility was assessed by investigating recruitment/enrollment, retention/adherence, acceptability, sensor usability, and changes in EDA waveforms based on potential stressors. For the latter, independent samples t-tests and ANOVA were used to examine differences in EDA from baseline to intervention, and the impact of environmental and infant factors on parent stress were visually analyzed against EDA waveforms.

Results:

Wearable sensor use in the NICU appeared feasible as assessed by all methods. Preliminary data analysis indicated that overall parent EDA levels during parent-infant interactions were low, and engagement in sensory intervention(s) led to a non-significant increase in parental EDA, measured by both skin conductance levels and non-specific skin conductance responses. Three main patterns of EDA emerged: a temporary increase in EDA at the beginning of the intervention followed by a decrease and plateau, a plateau in EDA from baseline to intervention, and a gradual rise in EDA throughout intervention. Specific environmental and infant factors, such as infant stress and health care providers entering the room, appeared to impact parent stress levels.

Conclusion:

Although these preliminary findings provide support for use of EDA in the NICU, future studies are needed.

1. Introduction

Becoming a parent is a significant developmental transition in a person’s life, which is inevitably accompanied by changes to one’s roles and routines. Despite the positive expectations and excitement of many parents-to-be, parenting can be challenging and stressful [1]. However, the preterm birth of an infant and subsequent hospitalization in the neonatal intensive care unit (NICU) can be even more distressing for new parents [2], impacting parental mental health and infant developmental outcomes [3].

Studies have shown that the NICU journey for parents is stressful regardless of the gestational age their preterm infants were born at or the medical comorbidities of the infant [4]. In the acute care phase, parents may be vigilant and anxious as the illness trajectory and prognosis remain uncertain, with rapid alternation between deterioration and progress [5]. The perceived loss of the parental role and competency in their ability to hold, care for, and protect their infant from pain increases parent stressors during and after NICU hospitalization [6]. In addition to internal experiences of stress, environmental factors such as a perceived lack of privacy and the chaotic activity within the highly medicalized NICU can also contribute to parental stress during their infant’s hospitalization [7–9].

As part of family-centered and family-integrated care practices, parents are increasingly encouraged to actively engage within the NICU setting [10]. Parents play an important role in comforting and caring for their infant in the NICU, and their presence is associated with infant neurobehavioral outcomes at term equivalent age [11]. While there is recent evidence that at-risk infants benefit from receiving parent-implemented, appropriate, positive sensory experiences [12], little is known about parental motivation and choice in the selection of positive sensory intervention or the effect of providing sensory interventions on the stress experiences of parents in the NICU.

Parental stress is typically assessed using a variety of parental self-report questionnaires that often consist of quantitative and qualitative questions to assess stress experiences (e.g., parental role alteration, infant appearance, NICU sights and sounds), anxiety, depression, coping, and resourcefulness among parents of infants admitted to the NICU [4]. These measures rely on the parents’ ability to accurately identify and report changes in their stress level and may vary depending on individual and cultural perceptions of stress. Further, these measure stress at one point in time, and the event causing the stress can be unclear. Other studies have documented psychophysiological measures of stress such as pre- and post-measures of salivary cortisol during skin-to-skin experiences with infants [13]. Measurement techniques that allow for the continuous collection of psychophysiological indicators of stress in real-time may be an important supplementary measure in the NICU environment, as understanding the stress parents experience in the NICU has the potential to impact family-centered care. Therefore, the purpose of this study was to examine the feasibility of utilizing wearable sensors to describe parental stress in the NICU. Specifically, feasibility was examined by investigating (1) recruitment and enrollment, (2) retention and adherence, (3) sensor usability, (4) acceptability, and (5) change in EDA waveforms in context of potential stressors.

2. Methods

2.1. Study design and participants

This prospective feasibility study was approved by the Human Research Protection Office at Washington University and the Institutional Review Board of the University of Southern California. Parents signed informed consent which permitted the research team to access the infant’s medical data, monitor the infant and parent during sensory intervention(s), and conduct the study. The initial enrollment target was 20 parent-infant dyads, which was believed to be adequate to assess feasibility across a sample of infants with different needs and with families from diverse backgrounds, as feasibility studies can have smaller sample sizes than other investigations that involve statistical hypothesis testing [14]. All infants were hospitalized in a 132-bed Level IV NICU at St. Louis Children’s Hospital. Eighteen dyads of infants who were born ≤32 weeks estimated gestational age (EGA) were enrolled. Infants were included if they had no congenital anomalies, were medically stable [off invasive respiratory support, >32 weeks postmenstrual age (PMA), tolerated interaction without physiological consequences], and at least one parent was present for interactions with them. The study was conducted from October 2019 to March 2020. Enrollment windows lasting 1–2 days within the study period were defined based on doctoral student availability to conduct study procedures. During these windows of research team member availability, all eligible families in the NICU were recruited to minimize selection bias.

2.2. Procedures

Study procedures were conducted within a one-hour period on one day, which was often on the same day as enrollment. Parents were given a choice of sensory interventions (e.g., skin-to-skin, holding, massage, reading, talking, singing) to engage in with their infant while monitored by a member of the research team who observed from outside the infant’s private room. The sensory interventions aligned with the recommendations of the Supporting and Enhancing NICU Sensory Experiences (SENSE) program, which provides parents with a choice of age-appropriate sensory activities to perform with their infants across hospitalization [12]. The SENSE program also includes parent education on conducting sensory exposures to ensure infant safety and enhance parent confidence [12]. Prior to the study procedures, some parents had already been educated about, or already felt confident in, providing sensory interventions to their infants. However, the research assistant provided education as needed/requested. The Empatica E4 wristband [15] was placed on the parent’s non-dominant wrist and covered with a lightweight fabric band for approximately 30 min prior to the observation period to allow for sufficient buildup of moisture between the electrodes and the skin. This was followed by a baseline period of 10 min in which the parent was asked to sit and relax at the infant’s bedside. The parent then engaged in a minimum of 10 min of sensory interventions with their infant. EDA was collected continuously during the study procedures. Clinical observations of the parent, infant, and environment were made by a member of the research team throughout the study procedures. Although some parents engaged in sensory interventions longer than 10 min, only the first 10 min of EDA were evaluated to standardize the timing across participants.

2.3. Measures

In order to determine feasibility, we examined recruitment and enrollment, retention and adherence, sensor usability, acceptability, and change in EDA waveforms in context of potential stressors.

2.3.1. Recruitment and enrollment

To determine if feasible to enroll NICU parents to participate in EDA research, we tracked the number of parents approached versus those that were enrolled (i.e., consented).

2.3.2. Retention and adherence

To determine the continued participation of parents who enrolled, we tracked the number of enrolled participants who completed all study activities as described in the protocol in addition to the number of participants who withdrew from the study.

2.3.3. Sensor usability

The success of the sensor to detect and collect a usable signal (e.g., not an untraditionally low signal with minimal or no variability) was documented for each participant, based on examination of the EDA waveforms following participation in the study.

2.3.4. Acceptability

Post-hoc evaluation of participant recruitment/enrollment and retention/adherence was assessed as a proxy measure for acceptability. No formal interviews with the parents were conducted, however, any spontaneous parent comments made during research procedures that related to perceptions about wearing the wristband were noted. In addition, occurrences of participation duration beyond the required 10-min sensory intervention protocol were documented.

2.3.5. Change in EDA waveforms in context of potential stressors

To evaluate the feasibility of utilizing wearable EDA sensors in the NICU to examine stress and inform future studies using this technology, we assessed the usefulness and interpretability of the collected psychophysiological data. Data collected included descriptors of the sensory intervention, infant/parent characteristics, and environmental factors as well as psychophysiological data.

2.3.5.1. Sensory intervention.

Parent choice of sensory intervention was documented from clinical observations made by the research assistant for each minute of the 10-min intervention period. The interventions were labeled as tactile (gentle human touch, holding, kissing, stroking, skin-to-skin care, massage), auditory (talking to the infant and/the other parent, reading, singing), and/or vestibular (rocking/swiveling).

2.3.5.2. Infant and parent descriptive information.

The following infant factors were collected from the electronic medical record: infant sex, EGA at birth, mode of delivery, birth weight, Apgar score at 1 and 5 min, PMA at the time of study, and PMA at discharge from NICU. Further, whether parents were exclusively breastfeeding their infants at discharge was collected as a measure to identify parents who likely had significant contact with their infants in the NICU. Parent and sociodemographic factors collected included parent gender, race (White/ Caucasian or Not White/Caucasian), parent age, if the infant was the product of a first pregnancy for the parent, and whether the infant was part of a multiple birth.

2.3.5.3. Infant states and stress signs.

A member of the research team made clinical observations during study procedures to document behavioral signs of infant stress and arousal states. The following infant factors were documented: periods of infant irritability (fussing for >3 s), changes in infant vitals (oxygen desaturation < 90 % SpO2, heart rate < 100 or >200 bpm, respiratory rate > 60 breaths per min), any transition between states of arousal (deep sleep, light sleep, drowsy, quiet alert, active alert, crying) with a transition defined as >15 s, and any behavioral stress signs (finger splaying, labored breathing, hiccupping, color changes, arching, gaze aversion, stooling, spit up, facial grimace, yawning, sneezing). Each observation was timestamped (per minute) to allow examination of its relationship to concurrent EDA output.

2.3.5.4. Environmental factors.

In order to assess the ability of EDA to detect potential stress responses, we examined the relationship between psychophysiological stress and common environmental stimuli during the sensory encounters by documenting any changes in environmental factors. These included monitor alarms, other infants crying, and interruptions (e.g., people entering the room or anything that appeared to distract the parent implementing the sensory intervention). In addition, we tracked baseline EDA measures as well as EDA measures during the sensory intervention to examine if we could detect changes in EDA after the parent began engaging in sensory-based activities.

2.3.5.5. Psychophysiological markers of parental stress.

EDA is a non-invasive way to measure activation of the sympathetic “fight or flight” nervous system. It is well-documented that measures of EDA increase due to stress [16]. Traditionally, EDA involves use of wired electrodes applied to a participant’s fingertips in laboratory settings. Innovative technologies now allow real-time wireless data collection from alternative sites, such as the ankle or wrist [17]; recordings from these sites are reported to be correlated with standard measurement locations [18].

This study utilized the Empatica E4 wristband, which was placed on the parent’s non-dominant wrist and covered with a lightweight fabric band to ensure continuous contact between skin and electrodes. The E4 includes two small dry Ag/AgCl electrodes located on the ventral side of the wrist and samples data at 4 Hz. Skin conductance level (SCL) and frequency of non-specific skin conductance responses (NS-SCRs) were collected for each participant continuously throughout the study, with NS-SCRs only counted when an amplitude of 0.05μS was reached [16]. Both SCL and NS-SCR frequency are considered to be measures of tonic EDA, with increases in each widely understood to indicate an increase in stress and arousal [16,19].

2.4. Data analysis

2.4.1. Recruitment and enrollment; retention and adherence; acceptability

Descriptive statistics were used to report recruitment, enrollment, withdrawal, and study completion data. Any comments indicating acceptability or lack of acceptability of test procedures or of the E4 wristband were documented.

2.4.2. Usability

EDA data were downloaded in CSV format from the E4 Connect online portal and imported into the BIOPAC AcqKnowledge program for data visualization and analysis. A low-pass filter was applied to all data to filter out noise and reduce artifacts. Both SCL and frequency of NS-SCRs were scored offline using AcqKnowledge and then hand-checked to ensure no NS-SCRs were missed or incorrectly marked.

Visualizations of EDA waveforms and examination of SCL and NS-SCR frequency determined usability, defined for this study as an average SCL ≥ 0.1μS, average NS-SCR frequency ≥ 0.2 (e.g., only 2 NS-SCRs in the entire 10-min recording), and no periods of flatline activity. These values were chosen to reflect the attenuated signal from dry electrodes at a decreased density sweat gland site (i.e., wrist), compared to the typical values obtained with wet electrodes at the fingertips (SCL: 2–20μS; NS-SCR frequency: 1–3 per minute) [16]. Descriptive statistics were used to report the proportion of usable versus unusable data across participants. Differences in participant characteristics for those with usable vs. unusable data were also evaluated using independent samples t-tests, chi-square/Fisher’s Exact test, or Mann Whitney U test, depending on variable type and normality.

2.4.3. Change in EDA waveforms in context of potential stressors

To determine feasibility of the EDA in detecting changes in psychophysiological stress related to providing sensory interventions, one-sample t-tests and ANOVA were used to determine differences from baseline to intervention in parent stress levels (EDA values). Visualizations of EDA waveforms were conducted in order to preliminarily examine the relationship of time-stamped environmental factors and/or behavioral indicators of infant distress with the EDA data.

3. Results

Our sample consisted of 17 parent-infant dyads. Most parents were female (93 %) with a mean age of 29.9 years (±6.1). Most (53 %) infants were female and, on average, 36.5 weeks (±3.7) PMA at the time of the study. See Table 1 for the characteristics of the sample.

Table 1.

Participant characteristics, stratified by usability of electrodermal data collected.

	Participants with usable EDA Data	Participants with unusable EDA Data	p-Value*
	Mean ± SD or N (%) or Median (IQR)	Mean ± SD or N (%) or Median (IQR)
Parental and environmental factors	(n = 9)a	(n = 5)
Sex: female	9 (100 %)	4 (80 %)	0.47
Race: White/Caucasian	9 (100 %)	3 (60 %)	0.21
Participating parent age	30.7 (±4.6)	28.6 (±8.7)	0.89
Multiple birth	(22 %)	0 (0 %)	0.62
First pregnancy	(33 %)	2 (50 %)b	1.0
Time spent in sensory intervention	10.0 (10.0–14.5)	12.0 (9.5–17.5)	0.91
Room temperature (°F)	72.6 (±1.1)	71.2 (±1.7)	0.06
Infant factors	(n = 12)	(n = 5)
Sex: female	6 (50 %)	3 (60 %)	1.0
Estimated gestational age (wks)	29.58 (±2.6)	32.2 (±2.2)	0.07
Birth weight (g)	1507.75 (±661.2)	1864.0 (±653.5)	0.33
Apgar score at 1 min	5.3 (±2.8)	7.0 (±1.4)	0.13
Apgar score at 5 min	6.8 (±2.1)	8.8 (±0.5)	0.01
Postmenstrual age at time of the sensory intervention	37.2 (±4.0)	34.8 (±1.9)	0.24
Mode of delivery: vaginal	4 (44 %)	3 (60 %)	0.59
Postmenstrual age at discharge	40.75 (±4.0)	39.0 (±4.6)	0.45
Exclusively breastfed at discharge	2 (17 %)	2 (40 %)	0.44
Infant medical diagnoses	(n = 12)	(n = 5)
Bronchopulmonary dysplasia	5 (42 %)	0 (0 %)	0.25
Patent ductus arteriosus	3 (25 %)	0 (0 %)	0.52

EDA: electrodermal activity.

^*P-value derived from t-tests or Chi-Square/Fishers Exact test or Mann Whitney U.

^aAlthough there were n = 12 parent-infant dyads included in the usable data group, only n = 9 parental factors are reported here due to two instances of multiple births (one set of twins and one set of triplets).

^bMissing data on one participant.

3.1. Recruitment and enrollment

Twenty-two parents were approached to participate in the study; 18 were enrolled (82 %), and four (18 %) declined due to anticipated challenges with scheduling (n = 2) or without reason (n = 2). Study procedures could not be conducted on one enrolled parent-infant dyad due to newly instituted COVID-19 restrictions (enrollment occurred in March 2020). No further attempts at recruitment were made due to the COVID-19 restrictions.

3.2. Retention and adherence

The parent-infant dyad who could not complete study procedures due to COVID-19 restrictions was not included in retention analyses due to the circumstances. Of the 17 parent-infant dyads successfully enrolled prior to the pandemic, 100 % completed the study procedures as defined by the protocol, and no participants withdrew from the study.

3.3. Acceptability

All proxy measures for acceptability indicated that the use of wireless sensors in the NICU environment was acceptable to parent participants (e.g., recruitment and enrollment: 82 % successful, retention and adherence: 100 % successful). One hundred percent of parents verbally reported the use of wireless sensors to be acceptable in the NICU environment while engaging with their infants. They were enthusiastic about the focus on parents and welcomed a non-invasive, quantitative method to examine their stress. No parents voiced concern or negative comments related to the study procedures or to wearing the E4 wristband during the study, and none requested that study procedures be stopped or that the wristband be removed. Of the 17 enrolled dyads, 41 % (n = 7) chose to continue wearing the sensor beyond the required 10-min sensory intervention, indicating the potential acceptability of wearing the sensor for a longer duration.

3.4. Sensor usability

Of the 17 dyads, 71 % (n = 12) of the EDA data collected from parents was usable. The unusable data had atypically low tonic SCL with no or few NS-SCRs present across the recording. We were unable to determine if these data were unusable due to equipment or user error, if the participants were electrodermal non-responders, or if participants were taking anti-cholinergic medications which would have negatively impacted EDA recordings. There were no observed differences in parent factors (see Table 1) or choice of sensory intervention(s) (see Table 2) between participants in the usable data group as compared to the unusable group. For infant factors, Apgar scores at 5 min were lower in the usable data group compared to the unusable group (p = 0.009).

Table 2.

Sensory intervention engaged in by parent-infant dyad, stratified by usability of electrodermal data collected.

Participant number		Tactile						Auditory			Vestibular
		Gentle human touch	Hold	Kiss	Stroke	Skin-to-skin	Mass-age	Talk to infant	Read	Sing	Rock/Swivel
Usable EDA data	1	x	x	x	x			x
	2	x			x			x
	3	x	x	x	x			x (and spouse)
	4	x	x	x		x		x (and spouse)
	7	x	x		x						x (rocking only)
	8		x	x	x			x	x
	9	x					x
	10	x						x
	14	x	x	x	x			x			x
	15	x	x	x	x			x		x	x
	16	x	x	x	x			x		x	x
	17	x	x	x	x						x
Unusable EDA data	5	x					x	x (and spouse)
	6	x	x	x	x	x	x	x (and spouse)			x
	11	x						x
	12	x						x	x
	13	x	x	x	x			x (and spouse)

EDA: electrodermal activity.

Despite wearing the E4 wristband for 30 min prior to the intervention to build up moisture between the electrodes and skin, SCL was overwhelmingly low across participants (range: 0.097–3.781 μS), with only four participants having recordings with an average SCL > 1 μS either at rest or during the sensory intervention.

3.5. Change in EDA waveforms in context of potential stressors

3.5.1. Sensory interventions

The majority of the parents chose to implement multiple sensory interventions simultaneously, with interventions lasting a median of 10.0 min (interquartile range: 10.0–14.5). Fifty-three percent (n = 9) of parents provided tactile and auditory interventions; 29 % (n = 5) of parents provided tactile, auditory, and vestibular interventions; 12 % (n = 2) of participants provided tactile and vestibular interventions; and 6 % (n = 1) provided a tactile intervention only. Implementation of tactile interventions (alone or in conjunction with other sensory-based interventions) were most common, including skin-to-skin care, holding, gentle human touch (hand hug or containment hold), infant massage, and stroking. Auditory interventions included talking, reading, and singing. Vestibular interventions included rocking the infant in the parents’ arms. See Table 2 for sensory interventions engaged in by parent-infant dyads during the study.

3.5.2. Infant states

Most infants were predominantly in state one (deep sleep, n = 6), followed by state two (n = 5, light sleep), and then state four (quiet alert; n = 1). During the 10-min sensory intervention period, 17 % (n = 2) of infants had a change in state. Across the entire observation period (including when parents extended the time for the intervention beyond 10 min), 25 % (n = 3) of infants had two state changes, 17 % (n = 2) of infants had one state change, and 58 % (n = 7) experienced no state change.

3.5.3. Infant stress signs

During the 10-min sensory intervention period, 33 % (n = 4) of infants exhibited stress signs, with one infant demonstrating three stress signs and three others showing one stress sign. Throughout the entire observation, the number of stress signs ranged from 0 to 5. Fifty-eight percent (n = 7) of infants exhibited no stress signs, 25 % (n = 3) one stress sign, 8 % (n = 1) two stress signs, and 8 % (n = 1) five stress signs. The infant with five stress signs was the only infant in a quiet alert state and was the only infant who experienced a desaturation event during the observation; however, this parent-infant dyad engaged in the sensory intervention for 22 min, more than twice as long as the minimum 10-min intervention required. This infant exhibited three stress signs during the 10-minute intervention period, and a desaturation event (as well as another two stress signs) occurred after the initial 10 min.

3.5.4. Environmental factors

Interruption of the sensory intervention was common in the NICU. Across the sample with usable EDA data (n = 12 dyads), 58 % (n = 7) experienced an interruption during the first 10 min of the sensory intervention, while 83 % (n = 10) experienced an interruption at any point during the observation period. Staff interruption was the most frequent (67 %, n = 8), followed by alarms or monitors sounding (58 %; n = 7). During the first 10 min of intervention, 17 % (n = 2) experienced alarms sounding, and 25 % (n = 3) had staff interruptions.

3.5.5. Change in EDA waveforms in context of potential stressors

Overall, there was a slight, but non-significant, increase in sympathetic activation as measured by both SCL and NS-SCR frequency during the sensory intervention (see Table 3, Fig. 1a and b). Three main patterns of EDA emerged from baseline to intervention. The first pattern included a temporary increase in EDA at the beginning of the intervention, followed by a decrease and plateau [50 % of sample (n = 6)]; see Fig. 2a. The second pattern included a plateau in EDA waveforms from baseline to intervention [33 % of sample (n = 4)]; see Fig. 2b. The third pattern included a gradual rise in EDA throughout the intervention [17 % of sample (n = 2)]; see Fig. 2c.

Table 3.

Changes in mean electrodermal activity measures from baseline to sensory intervention.

EDA measures	Baseline (mean ± SD)	Intervention (mean ± SD)	Difference between baseline and intervention (mean ± SD)	One-sample t-test p-value**	ANOVA p-value***
SCL (μS)	0.64 ± 0.61	0.97 ± 1.28	−0.33 ± 1.18	0.35	0.42
NS-SCRs (per minutea)	0.89 ± 0.66	1.92 ± 2.40	−1.03 ± 2.25	0.14	0.25

SCL: Skin conductance level; NS-SCRs: Non-specific skin conductance responses.

^aFrequency of non-specific skin conductance responses represents the number of skin conductance response increases ≥0.05 μS per minute.

^**p-value is from investigating relationships of SCL at baseline and intervention using one sample t-test analysis.

^***p-value is from investigating relationships of NS-SCLs at baseline and intervention using repeated measures analysis of variance (ANOVA).

Fig. 1.

Mean skin conductance level (Fig. 1a) and frequency of non-specific skin conductance responses (Fig. 1b) at baseline and sensory intervention, by participant.

Fig. 2.

Examples of three patterns of electrodermal activity exhibited during the first 10 min of sensory intervention: (a) initial rise during sensory intervention, followed by a decrease in electrodermal activity and sustained plateau; (b) plateau from baseline to sensory intervention; and (c) gradual rise in electrodermal activity throughout sensory intervention. Note. Please note different y-axis values on figures 2a-2c.

3.5.6. Responsiveness of EDA to infant and environmental factors

Overall, infant behavioral stress signs or changes in states of arousal did not appear to consistently impact EDA output. EDA responses occurred following the observed infant stress in only 33 % (n = 5) of parents. Among 15 infant stress signs reported across the sample, 5 were more easily identified as stress signals or infant discontent by parents (e. g., facial grimace, grunt, fussy behavior), as opposed to 10 being subtle stress signs (e.g., hiccups, finger splay, yawn, sneeze). Among the 5 more recognizable stress signs, 80 % of the stress behaviors elicited a psychophysiological response from parents (4 of the 5 occurrences). The remaining 10 subtle infant stress signs elicited a psychophysiological response of stress in the parent only 10 % of the time (1 of the 10 occurrences).

Similarly, <40 % of environmental interruptions/distractors elicited psychophysiological responses from parents (e.g., talking on the phone, talking to a friend, alarm in room or unit). However, the majority of these physiological responses occurred when a nurse and/or other health care provider entered the room (86 %). SCL related to these occurrences were the highest amplitude and most sustained (see Fig. 3).

Fig. 3.

Example of electrodermal activity responsiveness to environmental interruptions/distractors.

4. Discussion

Despite being a small-sample feasibility study, these findings suggest that the use of wearable sensors in the NICU to measure parent stress appears feasible, with parent acceptance of the EDA technology, and evidence of EDA signals that can be used to assess concurrent stress responses in NICU parents. Preliminary data also indicate parent stress levels during parent-infant sensory interactions were low and not significantly different than levels at baseline, and environmental factors (e.g., identifiable signs of infant distress, presence of health care providers) appear to increase parent stress levels.

Parent stress and its impact on mental health are now globally recognized as an important focus of care in the NICU [4,20]. Current methods to assess stress include behavioral observations of parent behavior, interviews, and standardized questionnaires such as the Parental Stressor Scale NICU (PSS NICU) [21] and the Parenting Stress Index (PSI) [22]. Self-report measures can be subject to a range of reporting biases stemming from misunderstanding the questions, the nature of question content, social desirability, and response shift bias [23]. Further, they include general questions that are often administered after the parent has been removed from the stressful encounter. It has been suggested that measuring stress in a timely manner during or following the stressful event, rather than through parent-report measures used afterward, may be beneficial in decreasing bias related to recall and difference in mood during assessment, as well as to increase accuracy of the measure in context [24].

In addition to the valuable information obtained from subjective parent reports of experiences in the NICU, utilizing objective measures of parental psychophysiological states may enhance the measurement of stress, and the current study is a first step in defining the feasibility of EDA in the NICU. Assessing naturally-occurring and dynamic stress in context can potentially provide important insights into the magnitude of stress experienced by parents in the NICU. It may also highlight how different factors in the environment exacerbate or alleviate parental stress. Further, utilization of wearable sensors to collect sympathetic nervous system data may be more practical in the NICU setting, as they allow for ambulatory measurement of EDA while simultaneously enabling participants unimpeded use of their hands during data collection activities (traditional electrode placement is on the fingertips or palms of the hand, which restricts activity). Use of the E4 to record EDA as a measure of sympathetic activity, found in this study to be feasible and acceptable to parents of infants in the NICU, paves the way to measuring parental stress in real-time, either during or in the absence of sensory interventions in the NICU setting.

The need for reliable data collection sensors is critical when considering the integration of these stress measures into clinical care, where access to usable data will be crucial. Our findings of feasibility in EDA providing reliable signals of stress in the NICU are consistent with other studies that have supported the validity of wrist-based EDA measurement, despite the use of dry electrodes and the location containing fewer eccrine sweat glands as compared to other body sites such as the fingertips, palms, or forehead [17,18,25–27]. Our findings were also consistent with other research that has reported a meaningful percentage of unusable EDA data when collected from wrist-based wearable sensors (e.g., 15 % [28]; 73 % [29]) and an untraditionally low average SCL (e.g., 0.5μS [28]). In this study, our rate of unusable data was 29 %, with no appreciable differences between parental or infant factors or intervention choice for the usable vs. unusable data groups. Although this increased our confidence in the feasibility of collecting data, we were unable to pinpoint the reason for unusable data within our sample, and this will be an important area for future research.

While this was a feasibility study, markers of stress via EDA provide important preliminary data, paving the way for future study. Low EDA in both baseline and intervention phases suggests that parents did not demonstrate marked changes in stress from baseline to sensory intervention during this study. Participants received education regarding how to conduct the sensory exposure based on the SENSE program; such education and guided engagement could have potentially resulted in lower stress levels; timing of education should be assessed in a standardized manner in future studies. While high stress levels are well documented among NICU parents using self-report measures [4], measures of psychophysiological stress responses in real-time have not typically been examined. As such, this study differs from others we are aware of as it investigated stress in context of the parent engaging in a sensory intervention with the infant, rather than assessing self-reports of overall stress related to the NICU (e.g., PSS NICU, PSI). It will be important to explore the relationship of parent-report and psychophysiological measures of stress in future research, as these tools are complementary in efforts to obtain a holistic understanding of parental stress in the NICU. Measurement of stress in real-time also can give more insights into how cumulative stress may impact parent perception of overall stress reported on self-report measures.

The most commonly exhibited pattern of psychophysiological response in our study included a temporary increase in EDA at the beginning of the intervention, followed by a decrease and plateau, which may be indicative of adaptive habituation [16]. Within controlled laboratory settings, it would be possible to determine the number of trials or exposures that a person requires before habituation occurs [16]; however, within a real-life NICU setting these types of analyses are likely not feasible. Alternatively, it is possible that parent engagement in sensory interventions with their infant reduces parental stress. Studies utilizing self-report measures of parental stress have found that engagement in specific sensory interventions (e.g., skin-to-skin care) is related to lower levels of reported stress [25,26]. In addition, lower parental salivary cortisol has been associated with engagement in skin-to-skin sensory interventions with infants during their NICU hospitalization as well as with infants with complex cardiac conditions in the intensive care unit, further suggesting that parent engagement with their infants may reduce parent stress in the NICU [28,30].

Our findings of low levels of stress during the intervention phase may hint at potential beneficial effects of engaging in positive sensory interventions that include, but are not limited to, skin-to-skin care with infants. Defining how EDA changes in relation to mediating activities and NICU stressors may further elucidate its practicality and possible benefits in identifying parental stress. This potential is supported by our data demonstrating indications of psychophysiological responses to most of the conceivably stressful encounters [e.g., easily identifiable infant stress signs, entrance of health care providers into room]. Further research to refine methodology and identify stressful encounters can improve timely initiation of interventions to decrease stress in context.

4.1. Limitations

This study had several limitations including a small sample size. No structured interviews or anonymous surveys were administered to formally assess acceptability following the completion of data collection; in future studies these activities should be undertaken in a systematic manner to minimize the risk of bias as it relates to the acceptability of wearable sensors. Interruptions during the study procedures were common and could have interfered with stress signs and interpretation. Despite overall successful data acquisition, only 71 % of our EDA data was usable. Although this rate is within the range reported in other studies utilizing this technology, future researchers will need to consider this as a factor when determining sample size and also ensure that certain parental medications serve as exclusion criteria (e.g., anti-cholinergic medications). For this study, separate observations of infant stress cues/environmental factors for each minute of recording were used so relationships to changes in EDA could be preliminarily assessed. These are broad interpretations, and future studies would benefit from continuous video recording to allow for behavioral coding of infant and environmental observations. Replication in a larger sample with video recording would also allow for measurements of specific skin conductance responses as well as examination of EDA components such as amplitude of response. Further, the 30-min period prior to measurement, as well as the 10-min baseline period in which the parent was told to sit quietly at the bedside, could have been therapeutic and resulted in decreased stress that we would not have seen otherwise so findings will need to be interpreted accordingly. Finally, this was a study of feasibility of the use of EDA in the NICU, and data should be interpreted accordingly.

5. Conclusion

This is the first study that we are aware of to assess the feasibility of the use of wearable sensors to examine parental stress during a sensory intervention with their infant in the NICU. The use of the wearable sensors appears feasible during sensory interventions between parents and their infants in the NICU environment, paving the way for future studies to utilize these techniques to assess parental stress. While this study examined stress in parents, future innovations that enable concurrent measurement of stress in infants could also enhance care at the bedside. Such technologies have the potential to improve our understanding of the impact of changes in infant medical status, different health care professional interventions, parent engagement, and environmental factors on parental stress in the NICU. In addition, monitoring of parental stress in real-time can potentially result in the implementation of parent-mediated stress interventions and/or support provided in the moment, leading to better outcomes for parents of high-risk infants in the NICU.

Abbreviations:

NICU

neonatal intensive care unit

PMA

postmenstrual age

EGA

estimated gestational age

EDA

electrodermal activity

SCL

skin conductance level

NS-SCRs

non-specific skin conductance responses