The recent preprint demonstrates results from Phase 2 of the COVID-19 wearable study. If you are interested in learning more about both Phase 1 and Phase 2 studies, we have covered them in better detail in a previous newsletter:

• Phase 1 study was published in Nature Biomedical Engineering in December 2020. The research team used retrospective wearable data to develop algorithms that could detect COVID-19 infection before symptoms emerge. The study shows 63% of COVID-19 positive cases could have been detected in real-time, with a median of 4 days prior to symptom onset. 

• In Phase 2, the research team aimed to validate algorithms to detect presymptomatic and asymptomatic COVID-19 cases in real-time.
   

What are our findings from the Phase 2 study?
 

1. Our system demonstrates a 78% accuracy in alerting study participants who are pre-symptomatic or asymptomatic: 

   • Among 68 COVID-19 positive study participants who have wearable data, our algorithm (NightSignal) sent alerts to 53 of them at or before symptom onset/ diagnosis or diagnosis date in asymptomatic cases.

   • The alert is generated at a median of 3 days prior to symptom onset, to enable effective early self-isolation and testing
      

2. The algorithms also identify signals resulting from vaccination:

   • Red alert is triggered after both vaccination doses or only one dose for Pfizer-BioNTech and Moderna vaccines.

   • For the first vaccine dose, the maximum resting heart rate overnight occurs on the first night after receiving Pfizer-BioNTech vaccine. 

   • For the second vaccine dose, the maximum resting heart rate overnight occurs on the first night or the second night after receiving Pfizer-BioNTech or Moderna vaccine.

   • In terms of reported symptoms, participants who received Pfizer-BioNTech vaccine reported fever after either dose; however, participants who received Moderna vaccine reported no fever after the first dose, but close to 60% of them reported fever after the second dose.
      

3. An association between symptoms and signals was investigated:

   • Fatigue and poor sleep are common symptoms reported by study participants who received red alerts. However, aches and pains, headaches, coughs, and feeling ill were reported more frequently by COVID-19 positive study participants.

   • Stress, intense exercise, and alcohol intake are symptoms most reported by COVID-19 negative participants. Interestingly, there are more red alerts triggered among COVID-19 negative study participants during the winter holiday season. This is a consistent phenomenon observed in our previous study before the pandemic. This observation implies people may experience increased stress, alcohol intake, or travel during the holiday season.
      

We invited Prof. Eric Topol, the Director and Founder of Scripps Research Translational Institute, as a guest speaker to a new course in the Department of Genetics, Cloud Computing for Biology and Healthcare (GENE222/CS273C/BMI222). 

“I don’t like the term precision medicine because if you keep making the same mistakes over and over again, it is very precise. We need accurate medicine.”, says Prof. Eric Topol. In the U.S., there are more than 12 million diagnosis errors each year, which means most of us will encounter one medical error once a year. 

One contributor to this largely overwhelming number is human factors among healthcare practitioners, such as errors in interpreting medical images, fatigue after working intense and long hours, and cognitive bias.
 

This is where Artificial Intelligence (AI) can step in. Improving the accuracy and effectiveness of human efforts is what artificial intelligence is good at. Prof. Eric Topol went through the advancements in using deep learning models to reduce human errors, and to improve patient experience:
 

1. Using AI on the human retina to diagnose kidney disease, control diabetes, predicting neural degenerative diseases, and predict the heart calcium score.

2. Using machine vision to detect abnormal tissue in real time, and to decide whether the tissue needs biopsy.

3. Paring AI with sensors to monitor the intensity of nursing support in ICU rooms to prevent falls, and to promote better hand washing in surgical settings.

4. Using AI diagnostic tool to sequence critically ill neonates’ whole genome with interpretation in as fast as 13.5 hours

Despite the enormous transformation AI is creating across all healthcare professions, it also provides patients with autonomy in managing their own health. The implication of patient autonomy enabled by AI is going to significantly improve healthcare in areas where trained healthcare professionals are scarce. Here are some early examples:
 

1. Comparing incoming heart rate data with personalized baseline established over time to notify smartwatch users of irregular heart rate patterns.

2. An At-home UTI testing kit utilizing a smartphone app to analyze urine samples on a dipstick and deliver immediate results. 

3. An average user without any training can create a high-resolution scan of organs within minutes within the convenience of their own home.

4. AI health tool launched by Google can diagnose skin conditions using a smartphone app.

AI health tools give more precious time back to healthcare practitioners to engage in thorough thinking and making better judgments. It also provides patients with better access to high-quality healthcare at convenience. Prof. Eric Topol ended the talk with a profound message: human needs are at the core of healthcare, so healthcare practitioners’ empathy for the patients plays a unique role that is irreplaceable by AI. 

Podcast: Dr. Michael Snyder on Continuous Glucose Monitoring and Deep Profiling for Personalized Medicine

“We all have habits that have unhealthy things in our routines, and glucose monitoring helps people pay attention to it.” As one of the most extensively monitored scientists, Dr. Michael Snyder exchanges fascinating stories with Dr. Rhonda Patrick about the personalized insights people learned about their health from continuous glucose monitoring. 

For example, the vinaigrette dressing in a perfectly healthy salmon salad may be the cause for a person’s glucose level to spike. Exercises, even a walk for as short as 15 minutes, could suppress a person’s glucose level after consuming a specific type of food that induces the glucose level spike. Continuous glucose monitoring allows people to gain personalized insights about how lifestyle impacts their health.

These personalized insights are not just anecdotal. Dr. Snyder is a pioneer of a biomedicine research approach, called longitudinal multi-omic baseline profiling. By monitoring a person’s molecular changes across a variety of omic domains over time, Dr. Snyder’s research team is able to gain a holistic and granular picture of the person’s health profile. What’s surprising is that each person’s health profile is vastly different. In fact, one person’s health profile is more similar to themselves when they are sick than to another person. 

This finding can change how diseases s are diagnosed. In conventional medicine, we are told we are sick when our health profile diverts from the population average of what is considered healthy. Dr. Snyder’s research shows, due to the wide individual differences, it makes more sense to compare a person’s current health profile to their own baseline healthy state than to the average healthy state of a population. In a longitudinal study with 109 study participants, Dr. Snyder’s research team is able to detect diseases, such as cancer, pre-cancer, and diabetes, before study participants receive diagnosis. This kind of health insight, if gained early enough, would empower people to change their lifestyles to shift the disease trajectory. The same research team is currently recruiting participants to understand the long-term health effect of dietary fiber supplementation.

Dr. Snyder asked during the podcast, “Guess what the puzzle looks like when you only have 5 pieces of the puzzle?” This could be what our health looks like through the lens of conventional medicine - limited health metrics are taken into consideration a few times a year. We will understand so much more about our health, and how it is changing as we make more diverse measurements over time. Dr. Snyder’s team is currently recruiting study participants for a series of studies, ranging from individual’s different responses to COVID-19 vaccines to environmental exposures’ impact on Crohn’s disease.


Other related podcasts and webinars:

• Sign up to access webinar recording: The Shrinking Distance between Human and Machine: Computing Where We End and the Technologies Begin. - by Science/AAAS and Tianqiao and Chrissy Chen Institute

• Michael Snyder on Using Data to Keep People Healthy - by Harry Glorikian at Moneyball Medicine Podcast

• Big Data and Our Health: Precision Health - by Bradford Cooper of Catalyst Coaching Institute

We are approaching the launch date of our virtual summer program: Personalized Medicine, Big Data, & A.I. Explorers Workshop.

• Time: July 19 - July 30, 2021 @ 9:00 am - 11:00 am PT, Monday-Friday
• Price: $1,500 per student. Scholarships are available.
• Prerequisite: No background experience is necessary, just a willingness to join the healthcare innovation revolution.

On the news & media

There are more we would like to share with you. We compiled a list of articles that featured our work on the news and media during this month for you:
 

• A Global Early Warning System for Pandemics: Mobilizing Surveillance for Emerging Pathogens - The Milken Institute

• More Than Just a Step-Tracker, Smartwatches Can Predict Blood Test Results and Infections, Study Finds - Fierce Biotech

• Can Your Fitness Tracker Predict Oncoming Viruses in Your Sleep? - Sleep.com

• Wearable Devices Can Predict Clinical Lab Measurements - Endocrinology Advisor

• The Invisible Future of Health Monitoring - Stanford Medicine | Center for Digital Health