Smartphones, wearables and such other mobile connected devices can provide powerful real-time epidemiology tools at scale. Combined with on-demand testing, this promises to shift the COVID-19 response narrative to a smarter ‘predict and prevent’ theme, while allowing to safely accelerate economic activities. Dr Arindam Basu, strategy consultant, healthcare and life sciences reviews a few of such devices.
While COVID-19 has spiraled the adoption of a wide variety of mhealth tools to rapidly scale a dynamic and collaborative data driven response to this mammoth public health crisis, he cautions that while more research to demonstrate and quantify impact is underway, such ‘data for good’ initiatives must also sensitively address the collective versus individual conflict of interest in generating these high quality open datasets
The COVID-19 juggernaut continues unabated, almost. Better prepared health systems and progressively falling fatality rates are the only silver lining. US, Brazil and India continue to lead the pack in terms of overall disease burden. While the compulsion of unlocking of economic activity and mobility coupled with rising prevention and hygiene fatigue amongst people, has led to resurgences across several other countries following initial successes. The re-imposition of lockdowns and other restrictions to manage the disease burden vis-à-vis health system resources is serving little benefit, while continuing to hurt livelihoods and economies. Further, the prolonged disruption of non-COVID-19 treatments is now threatening to undo the progresses made over decades in managing the burden of communicable diseases like TB and HIV and also NCDs including cancer. The deep financial damage to the private health systems amidst higher costs and lower revenues further threatens the continuity of health systems’ capacities in days ahead.
The singular strategy of testing, tracing, isolating and treating is heavily time dependent and resource intensive. And it remains largely a reactive approach to containing the disease transmission and mitigating impact. The huge strides in capacities to do so by health systems globally over the last few months seem only relative and insufficient, even more so for large, populous and low resource countries like India in the face of continuing surges across several of its states. The historic under-investments in public health and universal healthcare, huge gaps in the last mile care delivery along with lack of reliable population level data severely limit the ability to mount a more targeted and sustainable response. The public fear and stigma is further posing a big hurdle to people reporting symptoms voluntarily and seeking care early, proving the biggest deterrent to all mitigation efforts.
In such a setting, smartphones, wearables and such other mobile connected devices can provide powerful real-time epidemiology tools at scale. Combined with on-demand testing, this promises to shift the COVID-19 response narrative to a smarter ‘predict and prevent’ theme, while allowing to safely accelerate economic activities.
A 2018 study by Dr Aaron Miller, a postdoctoral scholar from University of Iowa, USA published in the Journal of Clinical Infectious Diseases found that anonymised smart thermometer data captured real time Influenza Like Illness (ILI) activity at national and regional levels in the USA, and for different age groups. Scientists analysed data from commercially available smart thermometers and accompanying mobile app, that recorded users’ body temperature measurements over a study period from 2015 to 2017. There were over eight million temperature readings generated by almost 450,000 unique devices in the field during this time. Existing forecasts relied heavily on the CDC data, but the information lagged behind real-time flu activity by at least two weeks. The study showed that such smart thermometer data, which also captures clinically relevant common symptoms apart from body temperature, much before a person goes to the doctor, adds much firepower to simple forecasting models. This greatly improved predictions of ILI activity in real-time and upto three weeks in advance. The study concluded, smartphone based sensors could enable real-time surveillance of infectious diseases at population and household levels.
Influenzas tend to produce higher and more protracted fever patterns when compared to common colds. Geo-spatial analytics can thus differentiate the unusual fevers from those caused by milder cold viruses. Real-time data on ‘anomalous’ fever spikes reported from homes using smart thermometers, serves as a powerful tool to target testing, isolation and containment efforts at local communities level early on. Also, the connected app gives users general advice on when to seek medical help, boosts tele-monitoring and early care initiation, thus reducing morbidity and mortality too. Kinsa Health has sold and distributed through diverse channels close to two million smart thermometers in the US. Kinsa now shares the data and the real-time maps of US counties and regions, showing anomalous ILI activity on the website www.healthweather.us
For COVID-19, numerous studies have established that fever remains one of the firsts and most consistent presenting symptoms amongst all cases, which constitute nearly one-half to two-thirds of all infections.
Kinsa’s data earlier indicated an unusual rise in fevers in South Florida in March this year, even though it was then not known to be a COVID hotspot. Within days, testing revealed that South Florida had indeed become an epicenter. A medRxiv pre-print ‘Real-time detection of COVID-19 epicenters within the United States using a network of smart thermometers’ highlights the role of distributed networks of smart thermometers in tracking COVID-19 trajectory in real-time and predicting emerging epicenters. In an interview to The New York Times, Dr Nirav Shah, a former New York State health commissioner, also an advisor to Kinsa said, real-time fever data “could speed up public health the way Twitter sped up the news cycle.”
Launched in March 2020 in the UK and extended to the US and Sweden, a crowdsourcing smartphone application – COVID Symptom Study was developed in collaboration with scientists with expertise in big data research and epidemiology at King’s College of London and Massachusetts General Hospital. This is a unique prospective population based study that collects daily symptom reports from volunteers, the largest of its kind in the world. Participants are not required to provide their names and any other personal information, only their ZIP codes. The study had recruited over 2.5 million participants (including health care workers) across the UK and the US between March to May 2020.
The prevalence of combinations of symptoms (three or more) including fatigue and cough followed by diarrhoea, fever and/ or anosmia was found to be predictive of a positive test verification for SARS-CoV2. Simple mathematical modeling from Wales, UK data predicted geographical hotspots of incidence five to seven days in advance of official public health reports.
The study offers critical proof-of-concept for the repurposing of existing approaches to enable rapidly scalable epidemiological data collection and analysis using mhealth tools, critical for a data driven response to this public health challenge. These findings appear in the recent June 2020 Science Magazine report ‘Rapid implementation of mobile technology for real-time epidemiology’ by Drew et al.
The ongoing study has since expanded to include close to five million participants. Another research team used a machine learning algorithm on a training dataset of completed cases, to analyse the first five days (from the onset of illness) of symptom logging data for these cases. This yielded six specific groupings of symptoms representing six distinct ‘types’ of COVID-19, type 1 through to 6 with increasing severity levels. Further they observed that people with Type 6 are 10 times more likely to need breathing support than patients with Type 1. The researchers further combined information about age, sex, BMI and pre-existing conditions together with these symptoms data. This was able to predict which cluster/ type a patient falls into and their risk of requiring hospitalisation and breathing support with a higher likelihood of being correct than an existing risk model based purely on age, sex, BMI and pre-existing conditions alone. A longitudinal clustering of symptoms data could flag high-risk patients, prognosticate outcomes especially the need for breathing support or more intensive care and guide resource mobilisation, about eight days ahead of the current average of day 13 that these patients present at the hospitals, the study noted. The pre-print paper ‘Symptom clusters in COVID-19: A potential clinical prediction tool from the COVID Symptom Study app’ is now available on medRxiv.
At Johns Hopkins, a team of engineers, epidemiologists and physicians at the Whiting School of Engineering, Bloomberg School of Public Health and School of Medicine launched on April 30, 2020 a smartphone app ‘COVID Control’ to track and analyse users’ body temperatures, along with any COVID-19 associated common symptoms. Anyone above 13 years of age with access to a thermometer and a smartphone can register. A simple and user friendly information submission system is a hallmark. The participants would take fewer than 10 seconds to submit their data to the app onto a single page, after taking their temperature with a thermometer on a daily basis. With all data aggregated across regions and anonymised, a participant’s individual data would never be shared. Users can also view a dashboard of analytical maps updated daily on both the mobile app and the website, showing emerging system identified hot spots.
The algorithm compares historically available national body temperature records and trends during previous flu seasons, to flag atypical spikes in body temperatures across US regions.
“Data from this app will allow us to map and identify hot spots of fevers across the United States, potentially indicating emerging outbreaks of COVID-19 before health care or testing is sought. That information can be key in our efforts to control and mitigate the spread of the Virus”, says Dr Frank C. Curriero, a professor in the Department of Epidemiology and director of Spatial Science for Public Health Center at the Bloomberg School of Public Health and one of the project leads.
Dr Robert D. Stevens, the other lead and associate professor of anesthesiology and critical care and neurology at the Johns Hopkins School of Medicine notes, “As an intensive care physician, I believe that one of the most important contributions we can make in the fight against this pandemic is to go upstream and create measures that will prevent people from developing the critical illness in the first place. One way to achieve this is to leverage the power of high resolution data that are captured via sensors and portable devices such as smartphones.”
Wearable tech fastest growing consumer tech sector
Wearable tech particularly fitness trackers, smartwatches etc represent one of the fastest growing sectors in consumer technology. Nearly 20 per cent of Americans already use one, notes a 2019 Gallup report. They use sensors to track a range of physiological parameters like sleep patterns, heart rate, breathing rate and (skin) temperature. They work passively and do not require people to manually enter their symptoms into an app, or use other devices to check these parameters on a daily basis.
If a cluster of otherwise healthy people living in the same area experience similar changes in their resting heart rate or respiratory rate or temperature, this could potentially herald an early outbreak warning at both individual and community levels, even before these people develop any symptoms. The WHOOP bands distributed by the PGA Tour recently have helped flag potential COVID infections in their golfers even before they showed any symptoms. Several of these later tested positive for SARS-CoV2. While researchers are excited about their potential, they warn that these are still early days and more research/ evidence is needed to turn these into reliable and actionable clinical tools.
West Virginia University researchers working with Oura Health have announced early results from a study of 600 frontline healthcare professionals, noting that such wearable ring data could drive AI models and help predict COVID-19 symptoms three days in advance of their onset, with over 90 percent accuracy. They have since scaled up to include more than 10,000 participants partnering with top academic institutions across US, in the next phase of the study. Benjamin Smarr, a data science and bioengineering professor at the University of California at San Francisco has been studying wearable tech data for the past decade and currently leads a major study on Oura Ring data and COVID-19.
Dr Eric Topol headed Scripps Research Translational Institute is leading another major ongoing fitness tracking study in the COVID-19 context. Scripp’s work in this space predates the current pandemic. In January this year, Dr Topol and his colleagues published a study that found that anonymised heart rate data from Fitbit users were able to significantly improve prediction models for Influenza-Like Illnesses when compared to using CDC data.
In an interview with GQ, Dr Topol said that such fitness trackers would pair well with high quality self-administered COVID-19 home tests, as and when they become available. This would allow infected people to self-isolate and seek medical attention much earlier. This could potentially and completely transform the overall epidemic response.
Fueled by novel data streams typically outside of public health systems like search engines, website access data, social media, mobile phones and wearable sensors alongside advancements in computing powers and data analytics, digital epidemiology has come a long way. Google Flu Trends is perhaps the best known starter of digital epidemiology, leveraging symptomatic search queries for the purpose of syndromic tracking of Influenza Like Illnesses.
Epidemiologist and chief innovation officer at Boston Children’s Hospital, Dr John Brownstein has been working on crowd-sourced symptom data for tracking influenza, through the ‘Flu Near You’ smartphone app for real-time disease tracking over the last decade. COVID-19 has spiraled the adoption of a wide variety of mhealth tools by individuals, communities, tech giants and policy makers across the world to rapidly scale a dynamic and collaborative data driven response to this mammoth public health crisis.
While more research to demonstrate and quantify impact is underway, such ‘data for good’ initiatives must also sensitively address the collective versus individual conflict of interest in generating these high quality open datasets. The not-so-great uptake of novel smartphone based contact-tracing applications rolled out by several countries in the wake of COVID-19 is a case in point. Clearly transparency and trust lead to greater participation and sustained engagement and are foundational for such connected health solutions to bring about a paradigm shift to preventive health, much beyond COVID-19. The ones that get validated would tremendously boost mobile technology enabled re-engineering of primary care.
The rising threat to humans from novel pathogens including zoonotic viruses calls for a smarter ‘frontline’ response to outbreaks. High quality real-time biometric and symptom data would flag and track the spread of contagions well before patients hit the hospitals, guide early and targeted interventions and essentially prevent the next pandemic!