The coronavirus disease (COVID-19) is a rapidly spreading infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1]. A total of 4.0 million cases and 143,000 COVID-19–associated fatalities have been reported in the United States as of July 25, 2020 [2]. Residents of nursing homes and long-term care (LTC) facilities represent only 0.7% of the total US population yet account for 8% of cases and 47% of all COVID-19 fatalities in the United States [2,3]. LTC residents also exhibit an infection fatality rate of 18.6%—a rate that is 13 times higher than for the total population [2-8].

The vulnerability of LTC facilities to respiratory disease outbreaks is well documented, and several factors have contributed to the recent COVID-19 outcomes: high-risk population (the majority of LTC residents are advanced in age and have one or more underlying conditions), high-risk setting (the frequency, type, and duration of close contact between the residents and staff), and epidemiological features and transmission dynamics (people infected with SARS-CoV-2 can be infectious before showing symptoms and 40% of new COVID-19 cases are transmitted by asymptomatic cases) [9,10]. Due to these factors, symptoms-based monitoring and slow manual contact tracing methods presently used by LTC facilities have proven inadequate, and new tools are needed to better control COVID-19 outbreaks [11-13].

Advanced age and underlying comorbidities are well-established risk factors for severe COVID-19–associated illness, hospitalization, and death [14,15]. Adults 85 years and older represent 2% of the US population but have contributed to 33% of all COVID-19 deaths (Multimedia Appendix 1) [2,3,16,17]. This death rate is 613.1 (per 100,000 population), 14 times higher than the overall population rate [2,18]. The average COVID-19–associated hospitalization rate for adults 85 years and older is 607.3 (per 100,000 population), roughly 6 times higher than for the overall population [2,18]. Older adults are also disproportionally affected by chronic conditions, where 60% have two or more conditions, and such persons are known to be at an elevated risk for severe COVID-19–associated illness [19,20]. Richardson et al [21] found that 94% of patients hospitalized with COVID-19 exhibited one comorbidity, and 88% of patients exhibited two or more.

In addition to housing vulnerable residents, LTC facilities exhibit several intrinsic characteristics that make them high-risk settings conducive for the rapid spread of SARS-CoV-2 [22]. First, in LTC facilities, residents live together in close quarters, eat communal meals, and participate in many group social activities. Second, caregiving staff frequently assist residents with their activities of daily living (ADL) such as bathing, dressing, and eating. ADL assistance requires intimate resident contact, which increases the probability for transmission from an infected staff member or resident. Third, during the course of a work day, facility staff move from room-to-room to provide care for many different residents. In addition, many staff members may work at multiple facilities or home care agencies; thus, if they become infected, they can serve as potential vectors between facilities [11,12,23]. Overall, the frequency, type, and duration of contact between residents and staff has contributed to increased SARS-CoV-2 transmission both within and between facilities.

The epidemiological features, infection progression characteristics, and transmission dynamics of SARS-CoV-2 and COVID-19 have also contributed to the difficulties faced by LTC facilities to contain outbreaks. Such parameters are also fundamental to the development of accurate mathematical models, control systems, and effective infection control policies [9,14,24-27]. The SARS-CoV-2 virus is known to spread primarily person-to-person through large respiratory droplets (>5 µm) expelled when an infected symptomatic or asymptomatic person coughs, sneezes, or breathes [9,10]. Airborne virus transmission is also possible in confined, poorly ventilated environments such as LTC facilities because when an infected person speaks they can expel aerosols, tiny virus containing droplet nuclei (≤5 µm), that can linger in the air for up to 14 minutes [28-32]. SARS-CoV-2 is also believed to be viable and infectious on surfaces for hours; therefore, transmission may occur indirectly via fomites, contamination of surfaces in the environment [33,34].

Isolation of confirmed and suspected cases, and identification of contacts via contact tracing are crucial to effective control efforts. These methods hinge on three key epidemiological parameters: (1) basic reproduction number (R₀), the average number of secondary infections generated by each infection; (2) serial interval, duration between successive infections and speed of viral spread; and (3) proportion of asymptomatic transmission. Best estimates indicate that the R₀ for SARS-CoV-2 causing COVID-19 is 2.5, which is significantly higher than the flu [35]. The serial interval, duration between symptom onset in a primary and secondary case, is estimated to be 3.96 days, which is almost twice as fast as SARS-CoV-1 [26,27]. The mean latent period, time from infection to onset of infectiousness, is estimated to be 3 days, which is shorter than the 5.1 day incubation period, time between infection and onset of symptoms (fever, cough, shortness of breath; Figure 1) [9,14,24-27]. Consequently, people infected with SARS-CoV-2 are most infectious 1-3 days before showing symptoms and up to 10 days after symptom onset [14,25]. SARS-CoV-2 is transmitted via symptomatic, asymptomatic, and presymptomatic routes, and current best estimates indicate the following: 25%-81% of cases are asymptomatic [36-38], symptomatic and asymptomatic cases are equally infectious [35], and 40%-44% of new COVID-19 cases are transmitted from presymptomatic individuals [14,35,36,39,40]. These features are consistent with early reports from LTC facilities, where 56%-73% of residents that tested positive for COVID-19 were asymptomatic at the time of testing [11,12,41] and that both presymptomatic and asymptomatic cases contributed to rapid facility spread [11-13]. Thus, symptom-based screening alone failed to detect asymptomatic infectious cases, and Arons et al [11] posited that conventional screening approaches in LTC facilities are inadequate because symptoms-based screening and polymerase chain reaction (PCR) tests are only being performed on symptomatic persons [12,13]. LTC facilities need contact tracing systems to rapidly identify, contain, and then broadly test asymptomatic infectious contacts [42].

Contact tracing, a core disease control measure used by public health authorities (PHAs) to prevent the spread of infectious diseases, is now being employed to identify and isolate individuals that came in close contact with a person infected with SARS-CoV-2 [43]. The manual contact tracing process is slow and has inherent time delays between confirming a case and finding a person’s contacts [9,44,45]. These time delays give secondary contacts more time to transmit the virus even further in the facility. Manual contact tracing also relies on humans both for data collection and data entry, which increases the potential for inaccurate or incomplete results due to human error. For the tracing process, a case needs to remember and report all contacts made over the past 14 days. In the LTC setting, an infected resident may have 10-30 close contacts, and older adults that may be experiencing memory impairment or dementia may forget their close contacts. Since more than 70% of contacts must be traced to control an outbreak [46], this may be difficult to achieve using manual contact tracing in a LTC facility.

Since SARS-CoV-2 spreads too fast to be contained by slow manual contact tracing, several digital contact tracing tools using smartphone-based apps have been developed [47,48]. If widely adopted, these apps show promise to effectively mitigate the spread of SARS-CoV-2 for the general population; however, smartphone-based contact tracing may have limited utility in LTC facilities for several reasons. First, LTC residents are typically older adults, and only 17% of adults 80 years and older own a smartphone [49]. Second, staff in many LTC facilities are not permitted to use a smartphone during the work day. Finally, smartphone-based approaches use Bluetooth technology, which transmits through thin walls in a facility and can result in false positives. Due to these limitations, there is benefit to having a digital contact tracing system built specifically for use in LTC facilities.

In this study we describe the development and implementation of a real-time digital contact tracing system designed specifically for LTC facilities to mitigate the spread of SARS-CoV-2 infections. Additionally, we developed a new susceptible-exposed-infectious-recovered (SEIR)–type infectious model that was adopted and parameterized specifically to describe propagation of COVID-19 in LTC facilities. The model was also used to simulate and assess the interventional performance of digital contact tracing compared to symptom-based mapping, manual contact tracing, and PCR testing.

Between March and July 2020, over 13,000 LTC facilities in the United States reported COVID-19 cases [56]. Many of these LTC facilities have experienced uncontrollable outbreaks resulting from the rapid and widespread transmission of SARS-CoV-2 [11,12,23,57]. As a result, residents of LTC facilities have been disproportionally impacted by SARS-CoV-2 and have accounted for over 40% of all COVID-19 fatalities worldwide [3,16]. Symptoms-based monitoring including temperature assessment [58] fails to identify asymptomatic infectious cases, and slow manual contact tracing methods have proven inadequate for controlling SARS-CoV-2 transmission in LTC facilities [11-13,23,42,59]. In this study, we describe the development of a new digital contact tracing system designed for use in LTC facilities. Our computer simulation results comparing different intervention approaches suggests that this system shows promise to be an effective tool to control COVID-19 outbreaks in LTC facilities.

In this study, we developed an epidemic compartmental model that was specifically parameterized to quantify SARS-CoV-2 transmission and control in LTC facilities. We used the model and considered various scenarios to assess the effectiveness of several intervention groups to control outbreaks: no intervention, symptom-based monitoring, PCR testing, manual contact tracing, and digital contact tracing. Under all conditions tested, the digital contact tracing system outperformed all intervention groups, achieving reduced SARS-CoV-2 spread, fewer total cases, and fewer fatalities. Most importantly, we show that the time delay is the most critical and sensitive parameter of the model. All conventional control methods (symptom-based monitoring, manual contact tracing, and PCR testing) except digital contact tracing have intrinsic time delays that cannot be compensated for with increases in efficiency. We conducted several simulations where we increased each interventional group’s probability of tracing a contact, and the results indicated that the control performance still could not reach the level achieved by digital contact tracing. Thus, the primary advantage of automated digital contact tracing methods is the speed at which potentially infectious contacts (both symptomatic and asymptomatic) can be instantly identified, classified, isolated, and tested. Given the high proportion of asymptomatic infections, the ability to quickly identify and test potentially infected persons before they show symptoms is key to preventing future transmission in LTC facilities.

Results from our simulations indicate that symptom-based screening alone was the least effective intervention group, resulting in 60%-71% greater cases and 10%-20% more deaths than the other methods. A limitation of symptoms-based monitoring methods such as temperature monitoring for a fever is that subclinical or presymptomatic secondary cases are missed. In LTC facilities, asymptomatic cases are equally prevalent and infectious as symptomatic cases and, thus, can be major contributors to COVID-19 outbreaks in LTC facilities [11,12,14,25,41]. Our data also suggests that symptom-based monitoring alone has intrinsic time delays due to the time required for people who are infected to both exhibit symptoms and then be identified by facility staff. To complicate matters, evidence is emerging that many older adults may not actually present the signature COVID-19 symptoms (ie, fever, cough, shortness of breath) [12,15,60]. Due to their blunted immune response systems or underlying chronic conditions, which may mask fever and acute illness, older adults may present atypical, nonspecific symptoms when ill with COVID-19, including increased falls, changes in activity and behavior (such as sleeping more and eating less), impaired mobility, malaise, fatigue, nausea, and even vomiting [12,15]. Thus, staff may require more time and use lower thresholds for suspicion to identify infected older adults that exhibit subtle symptoms. Such delays may translate into further spread of infection in the facility.

Manual contact tracing is a useful core disease control that is a key part of our country’s multipronged approach to mitigate COVID-19 transmission [43]. Estimates indicate that a large workforce of 300,000 tracers will be required for adequate tracing in the United States (nearly 1 tracer per 1000 people) [61]. The manual tracing process is error prone and slow because it requires a human tracer to interview new cases (~2 hours/interview) and then list, classify, and follow up with each contact (~1 hour/call/contact). Results from our simulations indicate that the time delays created by manual processes render the method less effective in LTC facilities than digital contact tracing methods. We found that digital contact tracing methods resulted in 12% fewer cases and 3% fewer deaths than manual contact tracing. As a result, manual contact tracing approaches will need to be supplemented with other rapid and efficient control measures. There are several additional challenges with using manual contact tracing alone in the LTC setting. First, an infected resident or staff member may have 10-30 close contacts, and estimates indicate that between 6 and 15 tracers require 12-24 hours to fully trace one case [44,45,62]. The delays created by this process give secondary contacts more time to transmit the virus even further in the facility. Second, manual contact tracing relies on humans both for data collection and data entry. This increases the potential for inaccurate or incomplete results due to human error. Accurate manual contact tracing requires the case to remember and report all contacts made over the past 14 days. In the LTC setting, many of the residents may have memory impairment or dementia, and thus, they may forget their contacts. The digital contact tracing system described in this study can automatically identify all of the contacts for a case and can be used to help augment manual contact tracing efforts performed by PHAs.

The most commonly used and reliable test for diagnosing SARS-CoV-2 infected cases is the reverse transcription-PCR (RT-PCR) test. PCR tests measure viral RNA and are performed using a nasopharyngeal, throat or saliva swabs, and take 1-2 days to process. PCR tests can effectively measure infection in people who are symptomatic with COVID-19 but are less likely to detect infection during the case’s latent period when they are presymptomatic [14,24]. The results from our simulation indicate that PCR testing can be an effective control method for rapidly identifying infection and minimizing transmission. However, for PCR testing to be effective, testing needs to be implemented on both symptomatic and asymptomatic exposed contacts on a universal and serial (weekly or daily) basis. In a recent study, Dora et al [41] investigated the benefit of serial RT-PCR testing of residents and staff at an LTC facility after an initial COVID-19 case was diagnosed. In this study, they found that after the first positive case was identified, 19 residents tested positive for SARS-CoV-2 and 73% were asymptomatic. All of the positive cases were rapidly transferred to an isolated ward to successfully break the chain of transmission [41]. One issue with daily universal testing at a LTC facility is the expense. PCR tests are expensive (US $150 per test), so daily testing at a 120 bed facility would cost US $18,000. Frequent PCR testing for all nursing home and LTC residents is reported to be unsustainable, where one-time tests would cost the industry US $672 million [63]. To address this challenge, many LTC facilities to date have performed PCR tests only on symptomatic COVID-19 cases. Given the high proportion of asymptomatic cases, we propose that digital contact tracing systems could be used to identify all high priority possibly infectious contacts that should be selected for PCR testing. This approach would be a cost-effective and effective method to control COVID-19 outbreaks.