Presentation
Self-Reported Privacy Concerns and Behavior Change While Using LittleBeats: A Child Multimodal Wearable Device
SessionDH8: Connected Care Systems: Integrating Digital Health, Clinical Workflows, and Patient Safety
DescriptionLittleBeats is a multimodal wearable device designed to be worn by infants outside of a clinical setting. This device passively measures indicators of biobehavioral development within an infant’s ecological home environment (Islam et al., 2024). The LittleBeats platform includes an electrocardiogram (ECG), inertial measurement unit (IMU), and audio sensors, and is worn inside a specially designed infant shirt. The audio sensors capture infant vocalizations, such as crying and babbling, as well as ambient speech, such as caregiver speech in the home. To protect participant privacy and simplify data processing, LittleBeats utilizes machine learning algorithms to categorize infant and caregiver speech, rather than research team members listening to raw audio from the home environment. To date, LittleBeats has only been employed in a research setting. However, the information measured by the device has the potential to provide clinicians with insight into the naturalistic behavior of their infant patients and help clinicians to identify developmental concerns. These concerns may include delayed gross motor development, as measured using the IMU, delayed speech development, as identified using the audio algorithms, and dysfunction of the autonomic nervous system and physiological stress response, as measured using the ECG. The latter is particularly important as early life dysfunction of the autonomic nervous system predicts multiple physical and mental health concerns throughout the lifespan (Kolacz et al., 2019; Porges, 2022). Assessing caregiver speech also allows clinicians insights into naturalistic caregiver-infant interactions, which are foundational to early-life development (Bornstein & Tamis-LeMonda, 2010). Taken together, the information measured by LittleBeats may help clinicians identify the need for early intervention, as well as monitor the effectiveness of these interventions.
Of course, the information provided by LittleBeats is only as useful as it is accurate to the naturalistic experiences of infants and their caregivers. One of the advantages of using wearable devices in a healthcare context is the potential reduction of the Hawthorne effect (or “observation effect”)—the phenomena wherein participants and patients alter their behavior when they know they are being observed (McCambridge et al., 2014)—as participants do not actively interface with the study team while data is collected passively. In this presentation, we report on participant concerns about being recorded, and changes in maternal and infant behavior while using LittleBeats in the home, as self-reported by mothers.
83 families participated in the current study wherein infants (Mage = 6.4 months, SD = 2.5) wore the LittleBeats device during their standard routine for eight hours per day, for three days. After completing in-home data collection, primary caregivers (exclusively mothers in this sample) completed a questionnaire during which they reported 1) their concerns about being recorded using the device, 2) the concerns of other family members about being recorded, 3) how reassured participants felt by information regarding the machine learning algorithms used to analyze their data 4) the degree to which their behavior changed while using the device, and 5) the degree to which the behavior of their infant changed while using the device. Participants responded to these items using a Likert scale of 1 (strongly disagree) to 5 (strongly agree), as well as optionally provided additional details via written response. Linear regression models indicated that concerns about being recorded and changed behavior were not significantly predicted by sociodemographic variables.
Mothers reported that they (M = 2.1, SD = 0.9; 6% agree or strongly agree) and other family members (M = 2.2, SD = 1.1; 14.4% agree or strongly agree) were relatively unconcerned about being recorded by the LittleBeats device. Participants who elaborated further (N = 13 & N = 5, respectively) described that they had a “fear of judgement”, or did not want to be recorded talking to other adults, such as when working from home or having “private conversations with [their] partner”. Most participants reported that receiving information about the machine learning algorithms alleviated their concerns (M = 3.9, SD = 1.0; 66.2% agree or strongly agree), although the participants who elaborated further (N = 8) described that they would have appreciated a more detailed explanation regarding the algorithms, or that they did not recall being told about the algorithms by study team members.
Mothers reported that their own behavior changed slightly while infants were wearing the device (M = 2.2, SD = 1.0; 31.3% agree). Open-ended responses (N = 22) described multiple themes. Participants described that they changed their behavior while interacting with their child. Examples include being more engaged with their child throughout the day, and being careful when holding their child, to avoid moving or “put[ting] pressure” on the device. Parents also described changing their behavior in ways not specifically related to their child, such as adjusting their schedule so that they used the device only on days during which they did not need to leave the house. Multiple participants described being initially aware that the device was recording, but that this awareness “faded with time”. In contrast with changed maternal behavior, participants reported that the behavior of their infant changed very little while wearing the device (M = 1.7, SD = 0.9; 12% agree or strongly agree) and further detailed (N = 7) that infants seemed physically uncomfortable while wearing the device, or were “more fussy” than usual.
This study offers preliminary evidence that most participants have relatively low privacy concerns regarding LittleBeats and self-report little behavior change while using the device. However, a few primary caregivers adjust their behavior slightly when they know the LittleBeats device is recording, and that some infants appear physically uncomfortable while wearing the device. Concerns about being recorded may be alleviated by providing participants with more detail regarding the algorithms which are used to analyze the raw audio files. Other designers of wearable health technologies may consider implementing a similar data processing system, wherein raw data is not shared with clinicians, but machine learning algorithms are used to present data to clinicians in aggregate. An additional suggestion for pediatric health researchers is to consider onboarding partners alongside the primary caregiver so that partners receive information directly from the study team. Getting buy-in from all family members may reduce recording concerns and improve the ecological validity of the data. The findings described herein inform a forthcoming HFE assessment of LittleBeats, and how this technology may support the workflow of clinicians.
Of course, the information provided by LittleBeats is only as useful as it is accurate to the naturalistic experiences of infants and their caregivers. One of the advantages of using wearable devices in a healthcare context is the potential reduction of the Hawthorne effect (or “observation effect”)—the phenomena wherein participants and patients alter their behavior when they know they are being observed (McCambridge et al., 2014)—as participants do not actively interface with the study team while data is collected passively. In this presentation, we report on participant concerns about being recorded, and changes in maternal and infant behavior while using LittleBeats in the home, as self-reported by mothers.
83 families participated in the current study wherein infants (Mage = 6.4 months, SD = 2.5) wore the LittleBeats device during their standard routine for eight hours per day, for three days. After completing in-home data collection, primary caregivers (exclusively mothers in this sample) completed a questionnaire during which they reported 1) their concerns about being recorded using the device, 2) the concerns of other family members about being recorded, 3) how reassured participants felt by information regarding the machine learning algorithms used to analyze their data 4) the degree to which their behavior changed while using the device, and 5) the degree to which the behavior of their infant changed while using the device. Participants responded to these items using a Likert scale of 1 (strongly disagree) to 5 (strongly agree), as well as optionally provided additional details via written response. Linear regression models indicated that concerns about being recorded and changed behavior were not significantly predicted by sociodemographic variables.
Mothers reported that they (M = 2.1, SD = 0.9; 6% agree or strongly agree) and other family members (M = 2.2, SD = 1.1; 14.4% agree or strongly agree) were relatively unconcerned about being recorded by the LittleBeats device. Participants who elaborated further (N = 13 & N = 5, respectively) described that they had a “fear of judgement”, or did not want to be recorded talking to other adults, such as when working from home or having “private conversations with [their] partner”. Most participants reported that receiving information about the machine learning algorithms alleviated their concerns (M = 3.9, SD = 1.0; 66.2% agree or strongly agree), although the participants who elaborated further (N = 8) described that they would have appreciated a more detailed explanation regarding the algorithms, or that they did not recall being told about the algorithms by study team members.
Mothers reported that their own behavior changed slightly while infants were wearing the device (M = 2.2, SD = 1.0; 31.3% agree). Open-ended responses (N = 22) described multiple themes. Participants described that they changed their behavior while interacting with their child. Examples include being more engaged with their child throughout the day, and being careful when holding their child, to avoid moving or “put[ting] pressure” on the device. Parents also described changing their behavior in ways not specifically related to their child, such as adjusting their schedule so that they used the device only on days during which they did not need to leave the house. Multiple participants described being initially aware that the device was recording, but that this awareness “faded with time”. In contrast with changed maternal behavior, participants reported that the behavior of their infant changed very little while wearing the device (M = 1.7, SD = 0.9; 12% agree or strongly agree) and further detailed (N = 7) that infants seemed physically uncomfortable while wearing the device, or were “more fussy” than usual.
This study offers preliminary evidence that most participants have relatively low privacy concerns regarding LittleBeats and self-report little behavior change while using the device. However, a few primary caregivers adjust their behavior slightly when they know the LittleBeats device is recording, and that some infants appear physically uncomfortable while wearing the device. Concerns about being recorded may be alleviated by providing participants with more detail regarding the algorithms which are used to analyze the raw audio files. Other designers of wearable health technologies may consider implementing a similar data processing system, wherein raw data is not shared with clinicians, but machine learning algorithms are used to present data to clinicians in aggregate. An additional suggestion for pediatric health researchers is to consider onboarding partners alongside the primary caregiver so that partners receive information directly from the study team. Getting buy-in from all family members may reduce recording concerns and improve the ecological validity of the data. The findings described herein inform a forthcoming HFE assessment of LittleBeats, and how this technology may support the workflow of clinicians.
Event Type
Oral Presentations
TimeWednesday, March 259:37am - 10:00am EDT
LocationNassau
Digital Health