Mobile Health-Collected Biophysical Markers in Children with Serious Illness-Related Pain

Abstract

Context: There is an ongoing established need to develop engaging pain assessment strategies to provide more effective individualized care to pediatric patients with serious illnesses. This study explores the acceptability of wireless devices as one option.

Objective: To evaluate the ability of wrist-wearable technology to collect physiological data from children with serious illnesses.

Methods: Single-site prospective observational study conducted between September 2017 and September 2018 at Rady Children's Hospital, San Diego, California, inpatient wards. Pediatric patients with diagnoses of cancer and sickle cell disease admitted to the hospital for acute-on-chronic pain and taking opioid pain medications were asked to complete two 24-hour continuous monitoring periods with the Empatica E4 wristband.

Results: Data collected from the device correlated with manually obtained vital signs. Children responded favorably to wearing the device. Participants with reported subjective pain versus no pain had average heart rate increased by 16.4 bpm, skin temperature decreased by 3.5°C, and electrodermal activity decreased by 0.27.

Conclusions: This study shows the possibility of collecting continuous biophysical data in a nonobtrusive manner in seriously ill children experiencing acute-on-chronic pain using wearable devices. It provides the framework for larger studies to explore the utility of such data in relation to metrics of pain and suffering in this patient population.

Background

Over 12,000 children are diagnosed with cancer every year in the United States,¹ whereas sickle cell disease (SCD) is the most common serious genetic childhood disease, affecting ∼1 in 2500 births and 100,000 individuals in the United States.² For both of these populations, pain related to their illnesses has a substantial negative impact on their quality of life.^1,3,4 Limited training in pain assessment and management, overestimation of providers' own skill to treat pain, and failure to refer patients to pain/palliative care specialists can result in suboptimal pain management with devastating effects on quality of life, physical functioning, and increased psychological distress.^5–8

The growing popularity of mobile technology among children, and the resulting ease with which clinical data can be collected, has resulted in a vast amount of literature centered around the feasibility of using mobile health technology (mHealth) as a pain assessment and management tool for patients.^9–16 Most of the current pain research involves using smart phones, tablets, and associated applications (apps) as electronic pain diaries.^17–19 To date, there have been no published data that combine personalized biometric data gathered from wearable mHealth with the self-reported symptoms of pediatric patients who have cancer or SCD.

In light of this, there is a need to develop innovative and engaging pain assessment strategies to provide more effective individualized care to pediatric patients with serious illnesses.^7,20–23 To address this need, we explored if it was possible to collect continuous biophysical data recorded from a noninvasive, validated, and wearable device in children with cancer or SCD hospitalized for a pain crisis, and evaluated how these data changed with subjective reports of pain and objective opioid use.

Methodology

Study design and setting

A single-site observational study of 12 children with acute-on-chronic pain between September 7, 2017, and September 24, 2018. This study was approved by the University of California, San Diego, and Rady Children's Hospital, San Diego, California, Human Research Protection Program. Written consent/assent was obtained from all participants and their parents.

Patient selection

Pediatric patients admitted to the hospital for opioid-dependent acute-on-chronic pain were screened for participation in this study. Patient inclusion and exclusion criteria are outlined in Table 1 and include either a diagnosis of SCD or cancer of any type, and disease-related pain requiring opioid analgesia. Recruitment of participants began at five years of age, as this is the earliest age when self-reporting of pain symptoms has shown to be reliable.²⁴

Table 1.

Eligibility Criteria and Demographics

Inclusion criteria • Male or female between the age of 5 and 21 at time of screening • Documented diagnosis of sickle cell disease or cancer experiencing cancer/chemotherapy-related pain as determined by a primary hematologist oncologist and/or palliative medicine specialist • Developmentally able to wear a wireless sensor and can follow study instructions with or without parental guidance • Subject or subject's legal guardian speaks or reads English fluently • Subject or subject's legal guardian able to understand and grant informed assent/consent Exclusion criteria • Subject has an implanted electronic device in the body • Inability to complete subjective data as required by study (e.g., on wireless sensors and on questionnaires) • Has infrequently scheduled clinic visits	Age
	Age, mean (SD)	15.3 (3.5)
	Age, range (years)	7–20
	Sex (%)
	Male	6 (50.0)
	Female	6 (50.0)
	Race/ethnicity (%)a
	Hispanic/Latino	6 (54.5)
	Black/African American	5 (45.5)
	Cancer diagnosis (%)
	Pre B-cell ALL	3 (25.0)
	T-cell ALL	1 (8.3)
	Osteosarcoma	1 (8.3)
	Acute myeloid leukemia	1 (8.3)
	Sickle cell disease type (%)
	Sickle cell anemia	4 (33.3)
	Sickle hemoglobin-C	1 (8.3)
	Sickle beta thalassemia	1 (8.3)
	Location of pain (%)
	Mouth	4 (33.3)
	Shoulder/arm	4 (33.3)
	Hips/legs	3 (25.0)
	Chest	3 (25.0)
	Lower back	2 (16.7)
	Stomach	2 (16.7)
	Throat	1 (8.3)

^aThe race/ethnicity of one participant was not recorded during the screening process at request.

ALL, acute lymphocytic leukemia; SD, standard deviation.

Data collection

Each study participant was enrolled for one week to limit study burden.

Biophysical data collection

Following informed consent/assent, each participant was fitted with an Empatica E4, a wireless wristband sensor, and instructed on its use. Although the E4 is validated for ease of use and acceptability by an adult patient population, its use in pediatrics is very limited.²⁵ It provides direct measurements of heart rate (HR); electrodermal activity (EDA) to measure sympathetic nervous system arousal; motion-based activity; and skin temperature. It also comes with an event tag button that can be used to mark events (Fig. 1).

FIG. 1.

Image of the Empatica E4 sensor detailing the components of the wireless wearable sensor.

Participants were instructed to wear the sensor all day, aside from bathing and showering, on two separate 24-hour periods during their admission. By wearing the wristband continuously for at least two separate occasions, the sensor captured data at baseline, during periods of acute pain, and moments of relief between opioid doses.

Manually recorded vital signs

Within 30 minutes of applying the E4, the patient's nurse would take and record vital signs and enter them into the hospital's electronic health record (EHR). Nurse-recorded vital signs included HR (peripheral pulse), respiration rate, and temperature taken with a temporal thermometer. The time stamps for when the vital signs were recorded into the EHR varied by ∼20 minutes from when they were obtained as per the Rady Children's Hospital nursing protocol.

Usability data collection

Upon completion of the study, participants were asked to complete a ten-item usability questionnaire (Appendix A1) assessing the ease of wearing the E4 wristband as a part of daily living, and their opinion of whether if it could be used to help assess pain. For study participants age 6 and younger, a proxy survey was provided for the parents to fill out.

Patient-reported data collection

At the end of the two 24-hour periods, participants completed the PQ Memorial Symptom Assessment Scale (PQ-MSAS) or PQ-MSAS proxy survey, depending on their age. The PQ-MSAS^26,27 is a validated self-reporting tool for patients seven years of age and older and provides a proxy form for use by parents/guardians in younger patients. It evaluates physical and emotional symptoms as well as assessing other external factors such as school and friendships.

Opioid use data collection

When experiencing pain-related distress that required an opioid, the patients were instructed to tap the event tag button on the E4 upon receiving the opioid, providing a time-stamped marked event on their biophysical recording (Fig. 3). In addition, the time, doses, and names of the opioid given to the participants were obtained from the EHR.

FIG. 3.

Study participant A016F.518446 recording of heart rate, temperature, and EDA over a 24-hour period with medication administration super imposed. EDA, electrodermal activity.

Study objectives

The primary outcome of this study was to evaluate the ability of collecting multiple continuously monitored biometrics—HR, EDA, and skin temperature—in an unobtrusive manner in children admitted for a pain crisis, and the accuracy of those data. Qualitative and quantitative data regarding the children's perception of the wristband were collected from the usability survey. We analyzed the change in baseline sensor recordings compared with the recordings surrounding opioid use and patient-reported outcomes of pain-related distress from the PQ-MSAS.

Data analysis

Pearson's correlation was used to determine the relationship between manually measured vital signs and Empatica sensor data, averaged within the first 30 minutes when vital signs were recorded. Manually measured vital signs were recorded in the hospital's EHR and could range in inaccuracy by as much as 20 minutes from the time that they were actually obtained. Therefore, when comparing these manually measured vital signs to the continuous vital signs time measurements, trimmed mean averages of the continuous vital signs were calculated over a time window of ±20 minutes from the time stamp of the manually reported vital signs. This explorative study was not powered to show a statistically significant result of opioid medication consumption and changes in biometric data, but rather highlight the unique aspect of continuous biometric tracking in this patient population. Documentation of each participant's self-reported symptoms through the PQ-MSAS surveys was combined with the biophysical data acquired by the E4 wristband, to track and assess patient-related distress due to disease-related pain. Yes/no responses were then compared with the mean and variance (standard deviation [SD]) of data recorded from the Empatica. A Wilcoxon test was applied to compare the group difference.

Results

Recruitment enrollment

Out of the 30 individuals screened for participation in this study, 12 provided informed consent and/or assent and were enrolled in this study (Fig. 2). Of the 18 who did not participate, 7 declined to participate, informed consent was unable to be obtained in 6, 3 did not meet the eligibility criteria, and 2 were not being treated with opioids.

FIG. 2.

Consortium flow diagram of study participant.

Demographics

Participants ranged in age from 7 to 20 years (mean 15.3, SD 3.5). Six (50.0%) were male and six (50.0%) were female (Table 1). Five patients (45.4%) identified as black or African American and six (54.5%) identified as Hispanic or Latinx. The race/ethnicity of one participant was not identified as per their request.

Usability survey

Complete results of the usability survey are shown in Table 2. When asked if they liked wearing the wristband, seven participants (58.3%) responded “yes” and five (41.7%) responded “no.” Two of the participants who responded “yes” to this question made the following remarks on why they liked wearing the wristband:

Table 2.

Usability Results

Survey questions	% Yes	% No	0 times	1–3 times	4–6 times	More than 7 times
Wristband easy to wear	91.7	8.3
Wristband tight	25.0	75.0
Wristband hard to use	8.3	91.7
People outside of family ask about wristband	58.3	41.7
Did not like people noticing wristband	0.0	100.0
Only look at wristband to push button	90.9	9.1
Hard to only push button to take pain medication	0.0	100.0
Extra times that button was pushed			66.7	16.7	8.3	8.3
Liked wearing the wristband	58.3	41.7
Desire for doctors to use wristband recordings to treat pain better	83.3	16.7

Because I like to know what is going on with my body. A012A8-1

It was very comfortable to wear, I liked how it looked normal, and it has a nice simple style. A01828-1

One patient who responded “no” to the same question expressed concerns about the comfort of the wristband.

It was either too loose or tight, there was no way to make it just right. A01E4-1

Another patient who also responded “no” to this question associated wearing the device with the status of his/her condition:

Because it means I'm sick if I am wearing it. A016F4-1

When participants were asked if they would like doctors to use the wristband to help them treat their pain better and faster, 10 individuals (83.3%) responded “yes” and only 2 (16.7%) responded “no.” Those who responded no did not leave a comment. Three of the people who responded “yes” to this question made the following remarks on why they would like doctors use the wristband:

Because it would help a lot of kids, especially younger ones, to get treated faster. A01E4

They would be able to have more info about how I am doing and what they can do to help. A01E4-1

Instead of hooking me up to a million and one wires, they could just use a watch; it's comfortable and convenient. A016F4-1

Comparison of E4 data with nurse-collected vital signs

The median values from the E4 occurring during the first 30 minutes of monitoring were used for comparison with nurse-collected vital signs obtained at the start of monitoring (Table 3). Body temperature measures were correlated with one another (p < 0.01), although device measures were on average 4.8° lower (p < 0.01) with larger variance (p < 0.01), likely due to the anatomical difference in location of where the temperature was taken. Likewise, HR measures were correlated (p < 0.01), and there were no differences in the mean or variance between the sources.

Table 3.

Comparison of Empatica-Collected Data and Nurse-Recorded Data

Subject no.	Mean Empatica heart rate	Nurse-recorded heart rate	Mean Empatica temperature	Nurse-recorded temperature
A01888.350568	87	95	27.0	36.2
A014E4.485698	89	91	34.5	36.8
A01828.448866	118	127	35.6	37.0
A01155.440412	107	117	28.4	36.5
A016F4.448854	83	76	32.9	36.2
A0106C.498638	91	92	28.3	36.1
A010FF.498602	102	141	34.3	37.1
A016F4.518446	87	60	32.7	36.0
A012A8.547764	84	91	30.8	36.2
A01828.555026	85	81	33.7	37.0
A014E4.555014	92	85	31.1	36.7

Heart rate measures were correlated (p = 3.01 × 10⁻³), but there were no differences in the mean or variance between the sources. Body temperature measures were correlated with one another (p = 0.036), although device measures were on average 4.8° lower (p = 1.26 × 10⁻⁴) with larger variance (p = 2.02 × 10⁻³).

Comparison by pain-related distress questions and mean of median EDA, temperature, and HR device data

Ten of the 12 participants' data were included in this analysis, one was not included because recordings from the E4 were corrupt, and we did not include the one whose parent completed the proxy survey (Table 4). There are notable clinically relevant differences among study participants who identified pain-related distress in key areas. The difference for those who identified shortness of breath had average differences of 0.9 in EDA, 6.5° in temperature, and 23.7 bpm average differences compared with those who did not. Those who identified as feeling sad had average differences of 1.04 in EDA, 4.6° in temperature, and 19 bpm compared with those who did not. In addition, those who subjectively indicated experiencing pain had an average difference of 16.4 bpm compared with the one participant who did not identify has having pain.

Table 4.

Comparison by Pain Related Distress Questions and Mean of Median ED, Temperature, and HR Device Data

Questions	Count		EDA			Temperature			Heart rate
	Y	N	Mean (SD)		p	Mean (SD)		p	Mean (SD)		p
			Y	N		Y	N		Y	N
Difficulty concentrating	3	7	0.21 (0.03)	1.03 (1.17)	0.38	33.4 (1.3)	31.5 (6.7)	0.83	93.8 (13.4)	90.8 (19.9)	0.83
Experienced pain	9	1	0.75 (1.09)	1.02 (—)	0.60	31.6 (5.8)	36.1 (—)	0.20	93.3 (17.7)	76.9 (—)	0.40
Lack of energy	7	3	0.85 (1.23)	0.61 (0.48)	1.00	31.5 (6.6)	33.3 (2.9)	1.00	94.3 (20.2)	85.7 (8.2)	0.83
Have a cough	3	7	0.21 (0.03)	1.03 (1.17)	0.38	33.4 (1.3)	31.5 (6.7)	0.83	93.8 (13.4)	90.8 (19.9)	0.83
Feeling nervous	3	7	0.21 (0.03)	1.03 (1.17)	0.38	33.4 (1.3)	31.5 (6.7)	0.83	93.8 (13.4)	90.8 (19.9)	0.83
Have dry mouth	6	4	0.97 (1.30)	0.50 (0.46)	1.00	33.3 (1.8)	30.2 (9.0)	0.91	92.2 (9.1)	91.0 (27.9)	0.48
Feel like vomiting	8	2	0.84 (1.14)	0.56 (0.66)	0.89	31.8 (6.1)	33.2 (4.2)	0.89	93.4 (18.9)	85.0 (11.4)	0.89
Feel numbness	4	6	0.61 (0.81)	0.89 (1.22)	0.91	33.9 (1.4)	30.8 (7.1)	0.76	93.6 (11.0)	90.5 (21.7)	0.61
Difficulty sleeping	7	3	0.93 (1.20)	0.43 (0.51)	0.52	31.3 (6.5)	33.7 (3.1)	0.52	94.2 (20.3)	85.9 (8.2)	1.00
Problems with urination	2	8	1.73 (2.16)	0.54 (0.62)	0.53	32.9 (1.2)	31.8 (6.3)	0.71	97.4 (16.6)	90.3 (18.5)	0.71
Vomiting or throwing up	5	5	0.60 (0.74)	0.96 (1.34)	0.84	31.0 (8.0)	33.1 (2.2)	0.69	93.5 (23.0)	89.9 (12.2)	0.84
Shortness of breath	3	7	0.15 (0.13)	1.05 (1.15)	0.18	27.5 (9.3)	34.0 (1.9)	0.07	108.3 (23.4)	84.6 (8.9)	0.27
Diarrhea	2	8	0.10 (0.14)	0.95 (1.10)	0.18	24.4 (10.8)	34.0 (1.7)	0.09	120.2 (15.6)	84.6 (8.2)	0.04
Feelings of sadness	4	6	0.16 (0.11)	1.20 (1.18)	0.07	29.3 (8.4)	33.9 (2.0)	0.26	103.1 (21.7)	84.1 (9.6)	0.26
Sweats	4	6	0.57 (0.85)	0.92 (1.20)	0.61	29.4 (8.5)	33.8 (2.0)	0.35	104.4 (20.6)	83.2 (8.8)	0.17
Worrying	3	7	0.21 (0.03)	1.03 (1.17)	0.38	33.4 (1.3)	31.5 (6.7)	0.83	93.8 (13.4)	90.8 (19.9)	0.83

Opioid medications

The majority of participants were prescribed oxycodone or a combination opioid/acetaminophen on an outpatient basis and continued on those home medications (Table 5).

Table 5.

Opioid Medications Provided to the Study Participants

Opioid type	Dosage range	Route
Oxycodone (n = 9)	• 2.5–10 mg	• Oral
Oxycodone/acetaminophen (n = 1)	• 5–7.5 mg	• Oral
Morphine sulfate immediate release (n = 4)	• 1–4 mg	• Intravenous
Morphine sulfate (n = 1)	• 1–2 mg	• Patient-controlled analgesic
Hydromorphone (n = 4)	• 200–500 mcg	• Patient-controlled analgesic
Hydrocodone/acetaminophen (n = 1)	• 5 mg every 4 hours (n = 1)	• Oral

The majority of the subjects enrolled in the study were prescribed oxycodone or a combination of opioid acetaminophen, on an outpatient basis, and continued on their home medications, with three of them having IV morphine added for breakthrough pain and three of them transitioning to a hydromorphone PCA during their time on the study. One subject was controlled with nonsteroidal anti-inflammatory drugs (NSAIDs) at home, but upon admission needed intravenous morphine for breakthrough pain. One patient was only controlled with a morphine PCA and a second only on a hydromorphone PCA upon admission with no oral opioids added during the study period.

PCA, patient-controlled analgesic.

Comparison of E4 data before and after medication

EDA, in addition to skin temperature and HR, collected from the E4 before and after the administration of analgesic medication was compared to determine the effect of medication on device measures. See Appendix A2 for full results.

Discussion

The objective of this study was to evaluate the ability of a wrist-worn wireless device to collect biometric data from children with serious illnesses and explore changes during pain crisis. This study is the first of its kind to evaluate a data set of continuous biophysical data from children with serious illness as it relates to their pain-related distress. It focuses on pediatric patients with a diagnosis of cancer and SCD as opposed to any other chronic illness because these patients are known to experience disease-related pain.

Pain assessment in children is challenging and often burdensome to the child as it is complicated by the age, developmental stage, and cognitive ability of the child to articulate the location, duration, or intensity of pain.²⁸ Exploring methods that can reduce this burden and help address other known challenges in assessing pain in the pediatric population is urgently needed.^29–31 The results of the usability survey show that it is feasible to collect these data from children admitted to the hospital for a pain crisis and that they are willing to wear devices, particularly if the information can help manage their symptoms. Moreover, survey comments demonstrate a desire to find noninvasive and less obtrusive modes of evaluating and assessing pain. These findings are in line with the Kleiman study which showed that, even in distressing situations, adolescents were willing to wear the E4 for prolonged periods of time motivated by a desire to contribute to scientific understanding, and found particular value in helping people similar to them in the future.²⁵

The results of this study provide some insight into the pain and suffering in children with serious illnesses that bears further investigation. The E4 allowed us to measure electrodermal activity, which by itself has been studied in automated pain detection in children and adults and can provide information about pain,^32–34 and is a measure of sympathetic arousal,³⁵ in addition to HR and temperature. It is notable that children who subjectively reported feeling sad and short of breath had associated changes across sympathetic arousal, temperature, and HR, and that these differences were greater than in those children who subjectively reported only pain, even though all were admitted for a pain crisis.

Although this study was not powered to show a difference in the data collected before and after an opioid was taken, digital technologies provide unique tools to explore behavioral medication use and physiological parameters that are not influenced by observational bias.^36–38 It can create a space to explore the influence on medication use on physiological changes in-depth and in real time.

Limitation of this study included the overall small sample size, which limited our ability to draw certain conclusions. A primary takeaway from this study is that to conduct a study aimed at evaluating the difference in data collected before and after an opioid administration, a substantial amount of data are needed to be able to establish relationships. During analysis, substantial quantities of noise had to be removed from the data before analysis of paired differences. In addition, when looking at opioid analgesic use, the study design would need to take into consideration both the 4-hour half-life and the 8-minute-to-1-hour time to peak effect depending on the route of opioid administration.³⁹ Second, we did not include an analysis of heart rate variability (HRV) as previous research shows that photoplethysmography (PPG)-based HRV is reliable only under ideal resting conditions, which was not possible in this study.⁴⁰ A key lesson learned in conducting this study was the importance of passively collecting the biometric data from the wireless devices, rather than trying to have the participants push the event tag button. The instances of the participants pushing the button very rarely correlated with when the nurses recorded administering the medications. The button pushing events created more noise and did not provide meaningful data. Thus, we only used time stamps from medications as recorded in the EHR.

This study creates a framework for evaluating correlations between patient-reported outcomes and biophysical measurements as objective markers of pain-related distress. We showed that it is possible to continuously collect biometric data from a wrist-worn device, and that children are willing to wear these devices. We provided preliminary data that need further investigation in exploring measures of pain and suffering in children with serious illnesses. These data further add to existing literature by combining previous research on emerging mHealth, investigating pediatric cancer and SCD pain through unbiased objective data, and understanding pediatric cancer and SCD pain through subjective clinical data to address a gap that is yet to be explored. Results from this study will open the gateway for further research into the utility of implementing different mHealth as pain assessment tools for children with serious illnesses.

Appendix A1. End-of-Study Questionnaire for Age 7 or Older

Study ID:______________________________

• 1.The wristband is easy to wear

• Yes No

• 2.The wristband was tight

• Yes No

• 3.The wristband was hard to use

• Yes No

• 4.People who are not my family asked me about the wristband

• Yes No

• 5.I did not like people noticing my wristband

• Yes No

• 6.When I had the wristband on, I only looked at it when I needed to push the button

• Yes No

• 7.It was hard to only push the button when I had to take pain medicine, sometimes I pushed it when it wasn't time to take my pain medicine

• Yes No

• 8.Aside from when you were supposed to push the button, how many extra times did you push the button?
• 0time 1–3 times 4–6 times More than 7 times
• 9.I liked wearing this wristband

• Yes No

• Why?

• 10.Would you like your doctors to use the wristband to help them treat your pain better and faster?

• Yes No

• Why?

Funding Information

This work was supported by the National Institutes of Health National Center for Advancing Translational Sciences Clinical and Translational Science Award (Grant No. 5 UL1 TR001114-05) CTSA KL Pilot Award 2018-2KL and The Rady Children's Translational Cancer Research Award.

Author Disclosure Statement

The authors declare that they have no conflicts of interest. The research group that designed and developed this study is financed by The Scripps Research Institute, Scripps Translational Research Institute, and Scripps Health.