Feasibility and Acceptability of Wearable Sensor Placement for Measuring Screen Time of Children

Abstract

Introduction/Purpose:

Wearables that include a color light sensor are a promising measure of electronic screen use in adults. However, to extend this approach to children, we need to understand feasibility of wear placement. The purpose of this study was to examine parent perceptions of children’s acceptability of different sensor placements and feasibility of free-living 3- to 7-day wear protocols.

Methods:

This study was conducted in three phases. In phase 1, caregivers (n=161) of 3- to 8-year-old children completed an online survey to rate aspects of fitting and likelihood of wear for seven methods (headband, eyeglasses, skin adhesive patch, shirt clip/badge, mask, necklace, and vest). In phase 2, children (n=31) were recruited to wear one of the top five prototypes for three days (n=6 per method). In phase 3, children (n=23) were recruited to wear prototypes of the top three prototypes from phase 2 (n=8 per method) for 7 days. In phases 2 and 3, parents completed wear logs and surveys about their experiences. Parents scored each wearable on three domains (ease of use, likelihood of wear, and child enjoyment). Scores were averaged to compute an everyday “usability” score (0, worst, to 200, best).

Results:

Phase 1 results suggested that the headband, eyeglasses, patch, clip/badge, and vest had the best potential for long-term wear. In phase 2, time spent wearing prototypes and usability scores were highest for the eyeglasses (10.4 hours/day, score=155.4), clip/badge (9.8 hours/day, score=145.8), and vest (7.1 hours/day, score=141.7). In phase 3, wearing time and usability scores were higher for the clip/badge (9.4 hours/day, score=169.6) and eyeglasses (6.5 hours/day, score=145.3) compared to the vest (4.8 hours/day, score=112.5).

Conclusion:

Results indicate that wearable sensors clipped to a child’s shirt or embedded into eyeglasses are feasible and acceptable wear methods in free-living settings. The next step is to asses the quality, validity, and reliability of data captured using these wear methods.

INTRODUCTION

Electronic screens (e.g., TV, computers, tablets, smartphones) are now ubiquitous in the lives of children. Research suggests that electronic screen use may offer both benefits (e.g., early learning, social contact and support) and risks (e.g., low physical activity/fitness, poor sleep quality, obesity) to the overall well-being of children (1-10). However, the ability to assess the relationship of screen exposure to behavioral and health outcomes is plagued by multiple measurement issues (11,12). To date, studies of electronic screen use in children have relied largely on parental or caregiver self-report, which is susceptible to high levels of error and is limited in its ability to capture the short bouts of screen use that are characteristic of newer media (e.g., smartphones) (13-21). Some groups have developed smartphone apps or TV/computer allowance devices that measure screen usage more objectively (22-24), but these systems often require instrumenting every screen in a person’s environment, thus severely limiting real world application and generalizability. A more precise, scalable, and cost-effective measure of screen use is needed to understand and evaluate screen time effects on child behaviors and health outcomes.

Wearable sensors have become the standard for measuring lifestyle behaviors (e.g., physical activity, sedentary, sleep) due to their unobtrusive size and vast data collection capabilities. However, the development and calibration of wearable sensors for the measurement of electronic screen use has been slow to evolve. Recently, research in adults has shown >80% accuracy in measuring electronic screen exposure with wearable color light sensors under controlled conditions (25-28). Extension of these findings to real world application with children requires understanding several knowledge gaps. First and foremost, we need to understand what method or “housing” is feasible for extended wear while maintaining proper sensor placement. Second, to distinguish mere presence of a screen from looking at a screen, light measurements need to be taken as close as possible to the subject’s eyes (28,29). Head-placed wearables have been proposed to robustly detect light received at eye level. However, the acceptability of these head-placed wearables for children in free-living settings is unknown. For example, eyeglasses appear to be a practical, everyday accessory that can house a light sensor without changing their main function, to enhance vision, or substantially modifying the eyeglasses’ appearance. Other potential candidates are headband, wearable adhesive patch (3M Tegaderm™ adhesive), badge pinned/clipped to shirt collar, mask, necklace, or vest. The aim of this pilot study was to iteratively examine the acceptability of different wear methods for a light sensor and the feasibility of a free-living 3- or 7-day wear protocol and logging routine with children.

METHODS

Study Sample

A convenience sample of parents, children, and/or childcare providers were recruited in three phases through social media (i.e., LinkedIn, Facebook, and Twitter) or e-mail. Recruitment was stratified by child’s age (3-5 and 6-8 yrs old). For phase 1, parents and childcare providers were recruited in January-February 2021 to complete an online survey to examine perceptions of different wear methods for the wearable sensor. Participants were eligible if they had at least one child between the ages of 3 and 8 years old or if they were employed as a child care provider serving at least one child in this age range. Participants completing the online survey were entered into a raffle for one of three $25 gift cards. For phases 2 and 3, parent/child dyads were recruited (phase 2: April-May 2021, phase 3: May-June 2021) to examine the feasibility of free-living 3- or 7-day wear protocols. Each phase of data collection lasted ~3 weeks. Participants were eligible if they had at least one child between the ages of 3 and 8 years old and were living within a 30-mile radius of Chapel Hill, North Carolina. Families received monetary compensation for each day they participated in the study. Participating parents received $5 for completing day 1 assessments (demographic questionnaire and wear log diary), $5 for each additional day the wear log diary was completed, and $15 for completing the follow-up survey. Total possible compensation for phase 2 was $30 and for phase 3 was $50. Children received an age-appropriate gift (e.g., sidewalk chalk) for participating. Web-based informed consent was obtained from all participants prior to participation and the Institutional Review Board at the University of North Carolina at Chapel Hill approved all study activities (IRB number: 20-3207).

Phase 1

Based on previous light sensor placements (i.e., eyeglasses, headband) in adults and expert opinion (26,27), investigators identified seven possible wear methods to house a new screen detection sensor: 1) headband, 2) shirt clip/badge, 3) glasses with no lenses, 4) vest, 5) neckband, 6) mask, and 7) adhesive bandage. Parents of 3- to 8-year-old children and childcare providers were presented with a Qualtrics (Qualtrics LLC, Provo, UT) link to an online survey. The survey consisted of pictures (Supplemental Content 1, figure; 1) headband, 2) vest, 3) necklace, 4) mask, 5) eyeglasses, 6) adhesive patch, 7) shirt clip/badge) and explanations of each wear method and participants were asked to rate aspects of fitting and the likelihood of wear for each method (Supplemental Content 2, table). Parent- and childcare-provider-perceived acceptability of each item was quantified using a net promoter score approach by child age group and child sex (30). Participant responses to each question were first categorized as promoters (high ranking), passives (mid ranking), or detractors (low ranking). Supplemental Content 2 (table) shows cut off scores for each category by question. Second, a net promoter score was calculated by subtracting the percentage of detractors from the percentage of promoters, which ranged from a low of −100 (if every participant is a detractor) to a high of 100 (if every participant is a promoter). To simplify comparison, all scores were converted to positive values by adding a constant value of 100, for a range of 0 to 200. Third, an average net promotor score was calculated across all questions for each wear method. Wear methods were ranked according to average net promoter score by participant type (parent/provider) and child age × sex groups (3 to 5 yr. old: boys/girls; 6 to 8 yr. old: boys/girls). The top two wear methods from each stratification were retained for phase 2 testing.

Phases 2 and 3

In phase 2, parent/child dyads were stratified by child age group (3 to 5 yrs. old, 6 to 8 yrs. old) and randomly assigned to one of the wear methods identified in phase 1. These wear methods were implemented using commercially available items with no active sensors attached (Supplemental Content 3, figure; 1) adhesive patch, 2) shirt clip/badge, 3) vest, 4) eyeglasses, 5) headband). Therefore, no data were collected directly from these mock devices. In phase 2, parents were asked to fit their child with the mock device and have them wear the item during waking hours for 3 consecutive days. Written instructions on fitting the mock device were provided to the parent. Parents completed daily wear logs and a usability survey that asked about the feasibility and acceptability of the assigned wear method. Based on parent ratings and logs, three methods were selected for further testing. In phase 3, a new group of parent/child dyads were recruited and randomly assigned to test one of the top three wear methods for 7 consecutive days. Similar to phase 2, parents completed daily wear logs and a usability survey. Supplemental Content 4 (figure) shows the consort diagram for phases 2 and 3.

Phase 2 and 3 measures

Sociodemographic questionnaires

Parents completed a self-report demographic questionnaire via Qualtrics. The questionnaire captured information about their personal demographics (e.g., sex, age, household income, employment status, education) and the demographic characteristics of the children (e.g., age, sex, race/ethnicity).

Wear log diary

Parents were asked to complete a wear log during the 3- or 7-day assessment period. The times of waking and going to bed, when the device was put on and taken off, and screen use were recorded daily on a standardized, preprinted recording sheet. Parents completed the log diary in real time. Wear time was calculated for each day by summing the duration of wear noted on the log diary.

Usability survey

After the assessment period, each parent completed a usability survey via Qualtrics that asked about the feasibility and acceptability of the wear method including ease and satisfaction with the mock device, the degree to which children were willing to wear the device, challenges encountered during implementation, and positive aspects of the wear protocol (Supplemental Content 2, table). The questionnaire was scored using the net promoter score approach as previously described.

Statistical Analyses

Sample demographics and all outcome measures were summarized by descriptive statistics — means and standard deviations for continuous variables and frequencies and percentages for categorical variables. Participants were only included in the analysis if they completed all the assessment protocols. In total, 8 participants did not respond to contact attempts to schedule assessments prior to wear-method assignment (phase 2, n = 2; phase 3, n = 6). Reliability of wear log data across wear day and summary metric were assessed by computing within-subject variance values and respective 95% confidence intervals (CI). The within-subject variance reflected how much individuals in the sample tended to change their reporting of summary metrics across wear days; smaller values indicate less variation in measurements on the same subject by different days. All analyses were performed by using SAS software, version 9.4 (SAS Institute Inc., Cary, NC). Within-subject variance and respective CI values were computed using a freely available macro (%icc9) that uses the Proc Mixed procedure in SAS (31).

RESULTS

Participant characteristics are presented in Table 1. In phase 1, 280 adult participants (175 parents, 105 childcare providers) consented and initiated the survey. Final analyses included only responses with <10% missing data (115 parents, 62 childcare providers). The majority of parents completing the survey were between 30 and 40 years old (71.3 %) and providers were 50+ years old (56.5%). Adult participants were predominantly female (parents 85.2%, providers 100%) and non-Hispanic white (parents 71.3%, providers 61.3%). In phases 2 and 3, 62 parent/child dyads were consented and 54 dyads (87%) completed assessments.

Table 1.

Characteristics of participating children, parents, and childcare providers.

	Phase 1 (n=177)		Phase 2 (n = 31)		Phase 3 (n = 23)
	n	%	n	%	n	%
PARENTS
Age
≤30 years	14	12.2	1	3.2	20	87.0
30 to 40 years	82	71.3	24	77.4	3	13.0
40 to 50 years	17	14.8	5	16.1	0	0.0
50+ years	2	1.7	0	0.0	0	0.0
Missing	0	0.0	1	3.2	0	0.0
Female	98	85.2	30	96.8	23	100.0
Race/ethnicity
Non-Hispanic White	82	71.3	22	71.0	20	87.0
Non-Hispanic Black	11	9.6	6	19.4	1	4.4
Hispanic/Latinx	9	7.8	1	3.2	0	0.0
Other	8	7.0	1	3.2	2	8.7
Missing	5	4.4	1	3.2	0	0.0
Income Level
<$50,000	18	15.7	0	0.0	0	0.0
$50,000-100,000	32	27.8	12	38.7	4	17.4
$100,000+	65	56.5	18	58.1	19	82.6
Missing	0	0.0	1	3.2	0	0.0
Education Level
HS/GED/Some college	26	22.6	2	6.5	2	8.7
College degree	36	31.3	6	19.4	3	13.0
Graduate degree	49	42.6	22	71.0	18	78.3
Missing	4	3.5	1	3.2	0	0.0
Age of their child
3 to 5 years	63	54.8	14	45.2	9	39.1
6 to 8 years	52	45.2	17	54.8	14	60.9
Sex of their child
Male	59	51.3	18	58.1	14	60.9
Female	56	48.7	13	41.9	9	39.1
CHILD CARE PROVIDERS
Age
≤30 years	1	1.6	-	-	-	-
30 to 40 years	4	6.5	-	-	-	-
40 to 50 years	22	35.8	-	-	-	-
50+ years	35	56.5	-	-	-	-
Missing	0	0.0	-	-	-	-
Female	62	100.0	-	-	-	-
Race/ethnicity
Non-Hispanic White	38	61.3	-	-	-	-
Non-Hispanic Black	19	30.7	-	-	-	-
Hispanic/Latinx	2	3.2	-	-	-	-
Other	2	3.2	-	-	-	-
Missing	1	1.6	-	-	-	-
Education Level
HS/GED/Some College	25	40.3	-	-	-	-
College degree	30	48.4	-	-	-	-
Graduate degree	6	9.7	-	-	-	-
Missing	1	1.6	-	-	-	-
Quality Rating of their program
1 star	7	11.3	-	-	-	-
2 star	5	8.1	-	-	-	-
3 star	11	17.7	-	-	-	-
4 star	12	19.4	-	-	-	-
5 star	14	22.6	-	-	-	-
Missing	13	21.0	-	-	-	-
Total 3-5 yr. old enrolled, mean (SD)	60	4.1 (1.8)	-	-	-	-

Quality rating: Providers earn higher ratings as they meet more quality standards. CACFP, Child and Adult Care Food Program; GED, Tests of General Education Development; HS, high school.

Phase 1

Figure 1 shows parent reported average net promoter scores by child age and sex. The shirt clip/badge wear method had the highest average score across all age/sex categories (all scores >124 points). The second highest ranked wear method varied by stratification. In parents with a 3–5 year old, bandage was the second ranked wear method for both boys (76.8 points) and girls (83.8 points). In parents with a 6-8 year old, the second ranked wear method was glasses for boys (89.3 points) and headband for girls (78.8 points).

Figure 1.

Phase 1 parent (n=115) mean (standard error) net promoter scores by child sex (column 1 = boys, column 2 = girls) and age (row A = 3- to 5-year-olds, row B = 6- to 8-year-olds).

For providers, wear method scores were ranked as the following: shirt clip/badge (124.3 points), vest (76.8 points), mask (75.5 points), bandage (73.3 points), headband (71.8 points), glasses (70.8 points), necklace (55.5 points). The top two wear methods from each stratification (shirt clip/badge, bandage, glasses, headband, and vest) were selected for phase 2 testing.

Phase 2

During phase 2, 31 participants were given one of five mock devices identified from phase 1 to wear for 3 days (n = ~6 children per item). On average, wear logs were completed over 8.5 (vest group) to 12.4 (bandage group) hours per day. Proportion of observation time with wear was highest for the shirt clip/badge (90.0%), glasses (84.0%), and vest (84.4%), with the badge and glasses averaging 9.8 and 10.4 hours of wear per day, respectively. Additionally, average net promoter scores were higher for the glasses (155.4 points), shirt clip/badge (145.8 points), and vest (141.7 points) compared to the headband (112.5 points) and bandage (93.7 points). Wear time and net promoter score ranks were similar across age group and self-reported screen time stratification (data not shown). Supplemental Content 5 (table) shows detailed wear log observations and wear time summaries by wear method.

Figure 2 shows the percent of participants during phase 2 that wore each mock device for a given day and hour (10-, 8-, or 6-hour) wear criterion. Using a 3-day/10-hour criterion, glasses had the highest compliance with 42.9% of participants meeting this criterion. Only 33.0% of participants met this criterion for the shirt clip/badge, headband, and vest. No participant met the 3-day/10-hour criterion in the bandage group. By shortening the minimum required wear time to 2-days with at least 8 hours of wear the percent of children meeting the criteria increased substantially for the glasses (85.7%), shirt clip/badge (66.7%), and headband (50.0%), but did not change for the bandage (33.3%) or vest (33.3%). Wear time was consistent across the 3 days for the headband, shirt clip/badge, glasses, and vest (within-subject variance <0.24) compared to bandage (within-subject variance = 0.51). Together, data from the wear logs and survey suggest the glasses, shirt clip/badge, and vest had the highest potential and were further tested in phase 3.

Figure 2.

Percent of children meeting wear criterion (days/hours). Column 1 = phase 2 wear logs (n=31). Column 2 = phase 3 wear logs (n=23). Row A = 10-hour criterion. Row B = 8-hour criterion. Row C = 6-hour criterion.

Phase 3

During phase 3, 23 participants were given one of three mock devices (n = ~8 per device) identified from phase 2 to wear for 7 days. On average, wear logs were completed over 7.9 (vest group) to 12.4 (shirt clip/badge group) hours per day (Supplemental Content 5, table). The proportion of observation time with wear was highest for the shirt clip/badge (75.3%) compared to the vest (57.6%) and glasses (56.3%). Within-subject variation showed wear time varied the least for the shirt clip/badge and the most for the vest placement across the 7 days (Supplemental Content 5, table). Additionally, average net promoter scores were higher for the shirt clip/badge (169.6 points) and glasses (145.3 points) compared to the vest (112.5 points). Wear time and net promoter score ranks were similar across age group and self-reported screen time stratification (data not shown).

Figure 2 shows the percent of participants during phase 3 that wore each mock device for a given day and hour wear criterion (10, 8, or 6 hours). Using a 4-day/10-hour criterion, shirt clip/badge showed the highest compliance with 57.1% of participants meeting this criterion. Only 37.5% of participants met this criterion for the glasses. No participant met the 4-day/10-hour criterion in the vest group. By shortening the minimum required wear time to 3 days with at least 8 hours of wear, 85.7%, 50.0%, and 25.0% met the criterion for the shirt clip/badge, glasses, and vest, respectively.

DISCUSSION

Not knowing precisely how much screen time children receive limits our ability to link exposure to behaviors and health outcomes. As technology to assess screen exposure becomes available (i.e., sensor and processing), researchers need information on potential structure and monitoring methods that are acceptable for young children and feasible for parents. In three investigative phases, we systematically examined the acceptability and feasibility of several wear methods for a potential sensor designed to detect child’s screen exposure. Using a minimum wear time criteria and parent ratings, we found that a shirt-clip/badge or glasses were superior to a vest, bandage, necklace, headband, or mask for housing sensor technology.

The move from subjective questionnaires to wearable-device-based methods of screen time detection will greatly improve the accuracy of measurement and our ability to investigate exposure in a more nuanced way. Previous research in this area has been primarily tested in adults in highly controlled settings with small samples (25-28). While one group has developed a method for estimating screen time using a wearable wrist band for children (28), this work is limited largely because of wrist placement, resulting in a limited viewing angle, making computer screens, tablets, and phones particularly difficult to detect. It is critical to identify optimal sensor placement that is also acceptable to young children for long term wear. Our results have narrowed and identified two wear methods that are acceptable, unobtrusive, and allow measurement at or near eye-level. Ideally the two methods would be interchangeable, but further investigation is needed to determine if data collected from glasses and a shirt clip/badge are of similar quality under controlled and free-living situations. For example, quality of data collected from sensors will need to be evaluated in various situations, such as when blue light blocking settings are enabled on electronic devices or when the sensor is at varying distances and postures relative to the electronic device (e.g., laying on the floor watching TV vs. sitting at a desk using a laptop) or in environments with bright ambient light settings (e.g., outside, near windows, riding in a car). Additional considerations may be necessary when using the shirt clip/badge placement of the sensor in situations where clothing (e.g., coats, scarves) may block the light signal. Future research identifying measurement error and other limitations will clarify applicable settings and situations for each. If both placement methods produce comparable data under a variety of conditions, using participant preference to determine method of wear could increase compliance and overall data quality in both children and adults.

In the field of physical activity and sedentary behavior measurement, adequate stability (i.e., intraclass correlations coefficients ≥0.80) of device-based measurement is generally believed to be achieved if an assessment administration includes at least 3-4 days of at least 10 hours of wear time over a 7-consecutive-day wear protocol (32-36). However, research has also shown adequate estimates in device-measured physical activity with as little as one day with at least 10 hours of wear (37). While the feasibility and acceptability of the glasses and badge were good with at least 1-3 days of at least 10 hours of wear, the percent of children with at least 4 or more days was lower than ideal. Wear criterion required for light sensors to produce valid data for young children have yet to be established. Our results indicate that wear compliance drops substantially after 3 days of wear, with very few participants getting 6-7 full days of wear. Further research will be needed to determine how much device-based data are required to get a “good” estimate of screen exposure in various groups of adults and children. If more than 3 days are needed, additional strategies to ensure adequate wear may be necessary. For example, some families said they presented wearing the device as a “big kid research job” to their children. The few families that took this approach reported their children had a higher desire to wear the device each day. Another strategy, used in behavioral change research but not assessed in the current study (38-40), is testing different incentive structures or allowing children to choose their own incentives. Understanding what type of incentive most motivates an individual child or the timing the incentive is introduced (e.g., presenting the incentive before or after wear period) may increase the child’s motivation to complete study protocols.

Limitations

This study benefits from a multi-phase process to identify, test, and refine wear protocols for a wearable light sensor for young children. However, it is limited because all participants were from a convenience sample of educated, predominantly non-Hispanic white families and, thus, the results are not generalizable to other populations. Furthermore, this study only assessed the feasibility to wear the device and did not include a sensor to validate plausible data capture. However, to date, no suitable off-the-shelf wearable light sensor exists. Identifying and refining wear protocols is a necessary first step to reduce the cost and time burden of future validation studies. Although the wear methods were tested without active sensors, the size (diameter: 2 cm, height: 0.8 cm) and weight (19 g) of the sensor is negligible and should not have altered the results. Another limitation is that no data was collected regarding the numbers and types of reminders that participants needed to wear the devices. Additional email or text-message reminders to parents throughout the week may ensure adequate wear beyond the ~4 days observed in this pilot study. Lastly, net promoter score is traditionally based on a single survey question. While we believe the average net promoter score calculation is a stronger method compared to simple average Likert score for determining acceptability, validity evidence for this method in this context is unavailable.

Conclusion

To effectively link children’s screen time to health outcomes, we must be able to accurately measure their exposure. As screen exposure measurement devices become available, it is important that researchers have information on acceptable and feasible monitoring methods. This study sets the stage for tests of the utility of potential wear placements for use with young children and creates minimal expectation for adherence to wear protocols. Future studies need to determine the validity of wearable light sensors in young children and determine wear time criterion needed to produce valid data.