Contact tracing is a key element of the public health response against the spread of infectious diseases [1-3]. Contact tracing implementation depends on a target pathogen’s characteristics and available treatments, vaccines, and other preventive services. It also frequently involves identifying, screening, and quarantining persons at risk of infection based on exposure to known cases, isolating and treating infected individuals, and later detecting other infected individuals. By tracing contacts, health professionals are able to document an individual’s risk and facilitate quarantine when necessary [3-7]. Its use has been highly successful in controlling the spread of tuberculosis, syphilis, Ebola virus, and SARS [1-3]. There is also evidence that contact tracing and accompanying isolation and quarantine likely blunted the emergence of the SARS-CoV-2 outbreak, containing the spread of virus in <4 months in the city of origin, Wuhan, China [3].

Traditional contact tracing is the status quo in the field of public health, typically managed by state, regional, and local health authorities. A traditional contact tracing strategy involves interviewing recently diagnosed cases and identifying with whom they have come into contact within specified physical distances, time frames, and lengths of each interaction. For example, a close contact for COVID-19 is defined per Centers for Disease Control and Prevention guidelines as someone within 6 feet (2 meters) of any case for at least 15 minutes over a 24-hour period [8,9]. This traditional method of contact tracing, although successful in many contexts, is labor-intensive and time-consuming, making it difficult to scale with an increasing number of SARS-CoV-2 infections [9]. Owing to these challenges, public health agencies struggled to trace COVID-19 contacts efficiently, especially in settings like schools and workplaces with high rates of contacts [10-12].

Schools have special social dynamics of concern for COVID-19 transmission including crowding in institutional environments and extracurricular activities that may increase risk of transmission [13,14]. Children and adolescents also have daily close contact with their parents, who may be active in many sectors of society. Thus, controlling outbreaks among youth helps keep schools safe and protects the well-being of the community at large [15-17]. Children and adolescents may struggle to follow guidance on mask use and hand hygiene and may have difficulty recalling their close contacts compared with adults, reducing the effectiveness of traditional contact tracing [18,19]. Given the potential impact of school-based contact tracing on reducing community transmission, we sought to increase the effectiveness of tracing within school settings with digital solutions based on suitable technologies.

After the onset of the COVID-19 pandemic, several app-based digital contact tracing methods have been proposed and implemented, though much remains unknown about their overall impact. Most contact tracing apps use GPS data or Bluetooth received signal strength indicator measurements as a proxy for distance. However, GPS data are only accurate up to an error of 10 to 15 feet; this is both insufficient for accurately detecting contacts within 6 ft and a poor indicator of location both in urban and in indoor settings [12,20,21]. Therefore, Bluetooth may be the most appropriate modality for measuring close contact interactions in school settings.

To address the limitations of traditional contact tracing and app-based digital contact tracing methods, such as scalability, notification delays, recall errors, and contact identification in public spaces, our colleagues at the Yale School of Engineering developed a carriable and wearable Hardware-Assisted Bluetooth-based Infection Tracking (HABIT) device that we previously pilot-tested in a university campus [22]. Implementation of the HABIT contact tracing intervention requires four components: (1) a local health center or clinic; (2) a central server; (3) carriable or wearable Bluetooth devices, called dongles thereafter; and (4) relay devices (smart phones or tablet) that transfer information from the dongles to the central server. These components are further described in the Multimedia Appendix 1.

We previously compared this approach (without the encryption functionality active) with daily self-report interaction data in a university setting and found HABIT to have high sensitivity (94%) and specificity (95%) [23]. We also found this approach to be superior to a comparable app-based digital contact tracing tool developed by the university in terms of specificity, sensitivity, and usability. HABIT also addresses several privacy concerns related to other technology-assisted contact tracing interventions. As a local authority, the health center is permitted to know if any user has tested positive, but the system only allows the central server to gather limited interaction data about users who have interacted with COVID-19 cases. A user who never encounters a positive COVID-19 case remains anonymous to the central server. Furthermore, this approach does not require personal mobile devices to constantly emit and collect Bluetooth data, as this activity is restricted to the dongle. As such, HABIT is ideally suited for use in centralized organizations such as schools, universities, hospitals, or businesses in which every community member carries the dongle, and a local health authority is able to manage the system.

Following the encouraging results from pilot implementation of HABIT in a controlled university environment, we wished to further evaluate its implementation in a real-world setting. Therefore, we implemented HABIT in a high school to test its performance and implementation success. We will refer to the independent prep school as “the School” for the remainder of this paper. In the first part of the paper, we evaluate the performance metrics of HABIT and subsequently examine implementation outcomes by integrating both quantitative and qualitative data. Our goal was to provide guidance through the results from this study to schools and other institutions seeking to implement digital contact tracing programs.

We deployed an embedded mixed methods design (Figure 1) to evaluate the implementation of HABIT, combining externally valid insights from quantitative data with detailed contextual explanations from the qualitative data [24,25]. We first evaluated the performance metrics of HABIT as well as its implementation outcomes quantitatively through mock case simulations and survey analysis. Following the collection and analysis of the quantitative data, we conducted qualitative focus group discussions with 2 groups of participants to gain a deeper understanding of factors that affected the implementation of HABIT.

The complementary study design uses the qualitative data to clarify, illustrate, and provide more depth to the results from the quantitative log data to enhance interpretation and substantiate conclusions [26,27]. Results are synthesized through the aforementioned performance metrics and implementation outcomes using the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework (Table 1) [28-30].

This study was conducted between March 2021 to May 2021. At the time of our study, the SARS-CoV-2 wild type (alpha strain) was the predominant strain. The Food and Drug Administration declared all adults in the United States (≥18 years) eligible for COVID-19 vaccines 1 week after our study began. All adolescents (ie, people aged 11-17 years) became eligible for vaccination 1 week before the end of our study period.

The study was undertaken in collaboration with an independent school for girls in grades 6 through 12 in Northeast United States. There were approximately 300 students in the School, ranging in age from 11 to 19 years. The student-to-faculty ratio at the School is 7:1.

Participant eligibility criteria included English fluency, ability to provide consent or assent, member of the School (students, teachers, or school staff), and at least 11 years old. We excluded middle school (grades 6-8) students from the mock case simulations and postintervention focus group interviews at the School’s request.

In December 2020, the research team began planning for this study. HABIT was tested by researchers before the start of the study to ensure that it was free of technical difficulties.

In February 2021, the research team and the School held a webinar for all faculty, students, and parents. The presentation covered a variety of topics, including the purpose of the digital contact tracing device, the development of HABIT and the dongle, the security features of HABIT, and the daily responsibilities of each study participant. The research team and a representative from the School responded to all questions from attendees. Following the webinar, a 20-page user manual was sent to the School health center and participants. The dongles were delivered to the School on March 18, 2021. The School organized and attached the dongles to lanyards and developed a distribution plan.

The School additionally described the study to all students again during a student assembly. Faculty advisers answered all questions during small group advising sessions with students. The consented participants were scheduled to pick up their devices between March 25, 2021, and April 1, 2021. The faculty adviser guided the participants through the process of setting up the app and syncing their devices. The School sent weekly reminders to the students to sync their HABIT devices. The teachers also reminded students verbally to adhere to the study protocol.

This study was approved by the Institutional Review Board at Yale School of Public Health, Human Subjects Committee (#2000028753) and deemed to be minimal risk. We collected opt-out parental permission, student participants (regardless of age) assent, and faculty or staff participants consent. The consent and assent contained information about research study summary, benefits, risks, and freedom consent and withdrawal. We did not offer incentives for participation. Randomly selected mock case interview participants were given US $10 gift cards upon completion of the second interview. Their privacy were safeguarded through encryption of the Bluetooth-collected interaction data and secure storage of the web-app survey data. All data collected via interviews and surveys were anonymized at the earliest point possible to further protect their privacy.

Of the 304 school members, 284 (93.4%) members consented to participate in the study; 284 (93.4%) participants responded to the preintervention survey, and 112 (36.8%) responded to the postintervention survey. As presented in Table 3, in terms of gender, participants were mostly women (preintervention survey: 258/284, 90.8%; postintervention survey: 96/112, 85.7%) which is not surprising given that the School is an all-girls school. There were no statistically significant gender differences between pre- and postintervention surveys (P=.62). Racial and ethnic information were not collected from individuals because of the risk of exposing the identities of participants, given that most students at the School are White. For participants’ roles at the School, most (205/284, 72.2%) participants in the preintervention survey were students. There was an increasing percentage of faculty who filled out the postintervention survey (79/284, 27.8%) relative to the preintervention survey (51/112, 45.5%; P=.006). High school students (grades 9-12) represent most participants in both the pre- and postintervention surveys (182/205, 88.8% and 50/58, 86%, respectively).

In terms of living arrangements, approximately half (130/284, 45.8% and 45/112, 40.2%, pre- and postintervention survey, respectively) of the participants lived on campus and half (151/284, 53.2% and 66/112, 58.9%, pre- and postintervention survey, respectively) lived at home. Of the ones who lived on campus, most participants had 11 to 15 floormates (preintervention survey: 36/130, 27.7%; postintervention survey: 10/45, 22%). According to the preintervention survey, 38.5% (50/130) of participants had no roommate; 30.8% (40/130) had 1 roommate; 12.3% (16/130) had 2 roommates; 18.5% (24/130) had ≥3 roommates. According to the postintervention survey, 22% (10/45) of participants had no roommates; 18% (8/45) had 1 roommate; 27% (12/45) had 2 roommates; 33% (15/45) had ≥3 roommates. Participants had more roommates during the preintervention survey period compared with the postintervention period (P<.001). This observation can be explained by the continually evolving nature of the COVID-19 pandemic. Between the pre- and postinterventional periods, a number of breakthroughs were achieved regarding SARS-CoV-2 transmissions and COVID-19 prevention and treatment, which resulted in many changes to the safety protocol, such as the number of roommates. Of the ones who lived at home, most individuals have 3 or 4 family members (preintervention survey: 77/151, 51%; postintervention survey: 34/66, 52%; P=.24). In addition, 3.2% (9/151) and 15% (17/66) of individuals reported living with older adults (P=.26); and 20.1% (57/284) and 22.3% (25/112) reported living with people with chronic health conditions (P=.37) in pre- and postintervention surveys, respectively.

In total, 5.6% (16/284) of the preintervention survey respondents reported contracting SARS-CoV-2 before and during the study period. In the postintervention survey, 6.3% (7/112) of participants reported being infected with SARS-CoV-2. The extracurricular activities in order of participation were sports, music, academic club, and model United Nations. Overall, the characteristics of the postintervention survey were similar to those of the preintervention survey, with the exception of the percentage of faculty participation (P=.006) and the number of roommates (P<.001).

From May 3, 2021, to May 28, 2021, we randomly selected 60 participants from the School to invite for mock case interviews and subsequently excluded 10 (17%; 7/10, 70% never received nor initiated dongles; 3/10, 30% for other personal reasons). Of the remaining 83% (50/60) of mock cases, we interviewed 33 (66%; the remaining 17/50, 34% did not respond to the mock case interview requests). Out of 50 mock cases, 19 (38%) decrypted their Bluetooth data, 17 (89%) of whom were among those that had been interviewed. Approximately half were staff (9/17, 53%) and half were students (8/17, 47%) with roughly equal distribution across grades 9 to 12. Because the School is an all-women school, most (16/17, 94%) participants were women with only 1 man staff member.

Considering data from 17 mock cases that completed an interview and decrypted their HABIT data, we identified a total of 161 contact interactions involving 118 contacts. We found significant differences between the number of contacts reported by interview versus Bluetooth, and each method identified a significant number of unique contacts that were not identified by the opposing method (Table 2). Of the 161 interactions, we identified 127 (78.9%) by interview only, 26 (16.1%) by Bluetooth only, and 8 (5%) by both methods. Most (85/118, 72%) contacts were students with roughly equal distribution across grades 9-12 and only 7 (5.9%) contacts identified from grades 7 to 8.

Contact tracing is one of the most fundamental public health strategies for infectious disease control. Despite its utility and documented effectiveness in multiple disease settings [23], there is a paucity of research on the effectiveness and usability of digital contact tracing systems. Most studies that assess the effectiveness of digital contact tracing methods are modeling studies. Modeling studies reflect considerable uncertainty, thus it is difficult to objectively assess how realistic the assumptions (and therefore the results) of modeling studies are [32]. Among the few studies where digital contact tracing has been tested, implementation outcomes and user acceptability have not been addressed [33,34]. Our study is the first to examine the performance and implementation of a novel contact tracing method in a secondary school.

In this implementation study, participants found the technology to be appropriate for school-based contact tracing and were eager to participate and use the novel HABIT device. Syncing difficulties, however, significantly reduced the usability of the technology and led to decreased engagement over time. Despite these barriers, HABIT was still able to identify unique close contact interactions that were not reported during traditional interviews, pointing to the potential benefit of using HABIT or similar technology in school or institutional settings to enhance contact tracing. We also expect that the performance of this study device may improve once we address the syncing difficulties and add reminder features to the app for better adherence. Despite the setbacks, participants reported that HABIT provided a sense of security, which can be extremely beneficial to students’ mental well-being during a pandemic.

This study builds upon an earlier pilot study evaluating HABIT on a college campus and is one of the first mixed methods studies to evaluate such a tool for COVID-19 contact tracing in a school setting. Compared with our previous college campus study, which reported high sensitivity (94%) and specificity (95%), this study reported decreased performance when HABIT was compared with traditional interview methods in a less-controlled high school setting, partially owing to technical difficulties and poor adherence, which should be addressed in the future implementation studies and investigated further. School safety is an urgent topic currently because of the debates over school openings and continuing prevention measures in the context of new SARS-CoV-2 variants. Implementing interventions among teenagers within a school setting is one of the most challenging scenarios, as teenage adherence to many interventions is 40% lower than that of adults [35]. Hence, our study taps into these challenges and provides valuable insights into the implementation of digital contact tracing in a “real world” setting.

Overall, the faculty and students had high satisfaction with the interface, ease of use, coherence, and appropriateness. However, the shape and design of the device can be improved to increase user adherence and usability. In addition to addressing the technical difficulties, we recommend that the app provide syncing and charging reminders to increase usability. Many found that HABIT provides a sense of security, which can be extremely beneficial during a pandemic. Further research is needed to determine implementation strategies for improving adherence and therefore maximizing HABIT’s performance. The study points to the necessity to use digital contact tracing and provides valuable guidance on implementing contact tracing interventions in a school setting. Contact tracing can be incredibly useful to keep students, teachers, and other school members safe and healthy in the face of global pandemic or outbreaks.

There are several notable strengths of this study. First, this study was conducted in a real-world high school setting, allowing for broader generalizability of our findings. Second, our use of mixed methods allows for triangulation of data and increases the internal validity of our findings. Third, the study is timely in offering information about contact tracing in secondary schools. The importance and originality of the intervention advance the understanding of implementing digital contact tracing in secondary schools.

There are also several limitations to this study. First, the School was composed of mostly women and White participants, and we were not able to collect any racial and ethnic information owing to the risk of exposing participants’ identifies. This therefore limits the generalizability of our findings. However, other studies on adolescent medication adherence have identified similar types of implementation difficulties, for example, low adherence, especially around exam periods; stress; etc [36,37]. This implies that our findings likely remain relevant in other types of high schools in the United States. Second, the participation rate for our survey questionnaire dropped 40% in the postintervention survey. The low response rate to the postintervention survey may lead to response bias and may not represent the entire target population. Despite the low participation rate, we were able to conduct planned statistical analyses and identify significant findings despite the reduction in power. Third, our study was also limited by recall bias because participants were asked to recall events 2 months after they occurred with self-reported data collected in the focus group discussions and the postintervention survey. Last, it is uncertain if all close contacts reported by interview and Bluetooth truly met the criteria for close contact interactions (within 6 feet for at least 15 minutes). We initially intended to conduct follow-up interviews with each mock case to discuss discrepancies between the interview and Bluetooth contact lists but were ultimately unable to complete this activity owing to logistical limitations. Despite these limitations, this study serves as a starting point for the implementation of contact tracing devices in primary and secondary school settings.

Despite the increasing use of digital contact tracing in the mitigation of pandemics, there is a lack of empirical evidence of its effectiveness and usability. This study aims to address this gap in the literature by implementing a “HABIT” system in a challenging secondary school setting during the COVID-19 pandemic. The effectiveness of the HABIT in picking up contacts that were not reported through traditional interview-format contact tracing confirms that more efforts should be made to promote optimized development and implementation of HABIT during a respiratory viral pandemic. An embedded mixed methods design was used to collect and analyze data from students and staff to determine the feasibility and usefulness of such an intervention. Both faculty and students were very satisfied with the interface, ease of use, coherence, and appropriateness. We identified some important challenges in the implementation process. The study showed that the technical components and adherence were suboptimal, leading to suboptimal user experiences and reduced performance. Thus, more attention should be paid to improving technical usability and exploring strategies for improving adherence. Findings of this study will be useful in addressing implementation strategies among adolescents in a secondary school setting, enhancing implementation success. Use of HABIT and strategies to ensure adherence appear to be pivotal. The prototype design and technical difficulties during syncing can be improved to increase usability and adherence. Further studies addressing these factors are needed. The added degree of emotional reassurance and sense of security provided by digital contact tracing should also be explored further. During a respiratory viral pandemic, Bluetooth digital contact tracing can play a critical role in keeping school members safe.