Exponential Medicine, Day 1 - Cool Stuff Summary (Nov 4 2018)

Exponential Medicine, Day 1 - Cool Stuff Summary (Nov 4 2018)

This is a collection of the cool ideas and slides that I’d like to reference again in the future.

I wasn’t able to attend Exponential Medicine this year, but enjoyed watching the live stream from Kenya.

You may be interested in the Notes from Day 2, Notes from Day 3, or Notes from Day 4.

This post is summarized in the video below, and also available on your podcast app under: Gregory Schmidt.

Day 1 - Presentations

Sunday, November 4 2018

Session 1
The Future of Health & Medicine: Where Can Technology Take Us

1.1 Welcome - Will Weisman, Executive Director, Summits, Singularity University

1.2 xMed Framing - Daniel Kraft MD, Faculty Chair for Medicine, Singularity University and Founder & Chair, Exponential Medicine

Session 2
Exponential Essentials Part 1

2.1 AI & Robotics - Neil Jacobstein Chair of Artificial Intelligence & Robotics, Singularity University

2.2 AI meets Medicine - Anthony Chang MD MPH Chief Intelligence and Innovation Officer, Children's Hospital of Orange County

2.3 Quantum Computing - Deepak Kaura MD MBA Chief Medical Officer 1QBit

2.4 BlockChain 101 - Nathana Sharma JD, MBA SU Principle Faculty, Blockhain Policy and Ethics

Session 3
Exponential Essentials Part 2

3.1 Digital Biology - Raymond McCauley Chair, Digital Biology & Founding Faculty, Singularity University

3.2 Regenerative Medicine - Nina Tandon PhD Co-Founder, EpiBone

3.3 Deep Learning in Medicine - Jeremy Howard Deep Learning Researcher, Founder, Fast.AI

3.4 Exponentially Human - Carmen Morcos MBA CEO, Impact Visionary

Session 4
Community, Patients, Countries Included

4.1 Drawing the Inside Out - Anita Ravi MD MPH FAACP Founder & Clinical Director, Institute for Family Health’s PurpLE Clinic

4.2 The Coming Explosion of Health Data Spigots - ePatient Dave DeBronkart ePatient

4.3 An xMed Adventure - Shafi Ahmed Professor of Surgery and Associate Dean, Barts Medical School

4.4 Moonshots & Exponential Entrepreneurship - Naveen Jain Founder & CEO, Viome

Session 5
Keynotes: Catalyzing the Future

5.1 Bringing Healthcare Systems into the Future - John Halamka MD MS Chief Information Officer & Dean for Technology at Harvard Medical School

5.2 Uniting the Toolset and Mindset of Innovation - Tom Wujec Founder, The Wujec Group

The Morning Show

Shawna Butler RN MBA: Nurse Economist, partnerships at Exponential Medicine. EntrepreNURSE-In-Residence at REshape Center at Radbound University Medical Center in the Netherlands.

Jessica Damassa: Executive Producer & Host WTF Health

Session 1: The Future of Health & Medicine: Where Can Technology Take Us

1.1 Welcome

Will Weisman, Executive Director, Summits, Singularity University

Soul Machines - building life-like human avatars, with voice recognition and speech synthesis, from New Zealand

Weisman comments on the need for us to address the ethical and unintended consequences of the technologies we are developing. [This is good - because when I was at Exponential Medicine a few years ago, nobody was talking about consequences]

1.2 xMed Framing

Daniel Kraft MD Faculty Chair for Medicine, Singularity University and Founder & Chair, Exponential Medicine

Daniel gave his trademarked 1000 slide overview of where exponential technologies in medicine are at.

A few of the slides:

Pyramid of human needs

SWOT analysis of big tech in healthcare

An Apple a day keeps the doctor away

Causes of death from peak to 2012 chart

In addition to using voice to determine emotional state, some evidence suggests it can also show cardiac health

Voice technology could be used to help coach patients.

Cool to see an update (from the 2015 slide) of the cost to sequence a human genome. Newest slide now includes 2017 data.

More details available at:
- NIH: DNA Sequencing Costs: Data (updated 2018)
- NIH: The Cost of Sequencing a Human Genome (updated 2016)

“NATURE GENETICS - 13 August 2018 - Use of genome-wide polygenic scores for common diseases identifying individuals with risk equivalent to monogenic mutations”

[Exciting to see research is finally showing how what we label as a single disease, is actually multiple disease subsets.] eg. Type 2 Diabetes is multiple disease subsets.

The digital doctors bag, 2018 image update.

VitalConnect’s patch does live stream of vital signs.

AI algorithms that have received FDA approval in last year. From @EricTopol

Health data cloud, 2018 edition.

[Digital healthcare in China is huge. Glad to see a slide included on this. I still think in the West we are very ignorant, myself included, to the scale of digital healthcare platforms growing in China. Its top of my list to research more on. Unfortunately, when I tried to to learn more on this it was a bit difficult to find many articles in English]
Ping An Good Doctor - over 190 million users
AliHealth (from AliBaba) -
Guahao ‘We Doctor’ (from Tencent) - over 100 million users

Daniel has been working on prototype to custom print your own daily medication pill. [This technology is actually a lot further ahead than I thought it was. His Oct 18 2018 TED talk linked below]

The crowd-sourced Waze map of Rome was very cool. [This is something I need to look more into. We have considered crowd-sourcing data for purpose of epidemic and infectious disease tracking…interesting to consider what other ways crowd-sourced data can influence healthcare].

“We can all not only be organ donars, but ‘data donars’ as well”

2.0 Session 2
Exponential Essentials Part 1

2.1 AI & Robotics

Neil Jacobstein Chair of Artificial Intelligence & Robotics, Singularity University

Deep learning system has shown (previously unknown to physicians) that the retina pattern varies with age, gender, and smoking. Work done by Google & Stanford.

Slides from DARPA / Waves of AI

1st Wave - Handcrafted Knowledge - rule based, could get OK outcomes, but hard to change with new data

2nd Wave - Statistical Learning - neural networks, decent classification and prediction, but no reasoning and abstraction

3rd Wave - Contextual Adaption - perceive and learn -> reason and abstract. We’re not here yet.

Biased data sets causing errors when training data.

Machine systems not good at explanation. Unable to explain why it thinks the way it thinks. This is a problem for adoption to medicine.

Jacobstein comments directly on fact that “some people will tell you AI is all about prediction”, but he proposes the end goal is prescriptive analytics (“How can we make it happen”), not just predictive analytics, (“what will happen”). [This seems like direct comment at new 2018 book, Prediction Machines: The Simple Economics of Artificial Intelligence]

[I found this slide of AI/ML competitive advantages applied to business jargon amusing]

Screen Shot 2018-11-05 at 1.26.04 PM.png

Updated 2018 CB Insights Maps:

The AI 100 - "ranking of 100 most promising private artificial intelligence companies in the world”. Showing AI in every industry.

Healthcare Companies: Freenome, Babylon, Recursion, Tempus, AiCure, flatiron, Arterys,
Life Science Company: zymergen

CB Insights 2017: 106 Startups transforming Healthcare with AI. Shows AI in almost every field within healthcare. Full press release on Feb 3 2017 CB Insights page.


Some convergence in the firing patterns generated in synthetic AI systems, and that in nature.

Example of a hexagonal firing pattern to solve an AI maze vs hexagonal firing pattern found in a rat.

AlphaGo Zero: trained based strictly on rules of AlphaGo, and playing against itself (eg. not based on watching humans play games).

Within 3 days = AlphaGo Zero better than the version of AlphaGo that beat Lee Sedol (4 out of 5 games) in 2015

Within 21 days = achieves AlphaGo level that beat Ke Jie in 3 of 3 games in 2017.

Then AlphaZero on learning chess using re-inforcement learning as well. It achieved “superhuman” skill within 24 hours of training. [Full Academic paper from DeepMind December 2017 here]

Hardware. NVIDIA Volta - low energy aI chip

Kaggle (data science competitions) - bought by Google Cloud.

Experfy - website where you can post your big data problems, and experts bid on-demand how much it would cost to solve your problems.

Rather funny montage of robot footage (“people actually like watching robots fail”) , moving from 2015 where many robots at the DARPA competition would fall over, through the work of Boston Dynamics in 2016, 2017, and 2018. Now the robots have moved from funny —> to terrifying with terminator grade capabilities in balance and locomotion.

Speed of transportation change in NYC: 1904 almost all horses, 1912 more cars than horses, 1917 almost all cars.

Auris Health, received FDA approval March 2018 for their Monarch Robot - a flexible surgical robot to diagnosis and treat peripheral lung nodules. Videos on their website.

Itai Cohen & Paul McEuen - Cornell - doing work with at micron level with graphene panels (Paper in PNSA, Jan 2018) to create nanorobotics. An interview here.

2.2 AI meets Medicine

Anthony Chang MD MPH Chief Intelligence and Innovation Officer, Children's Hospital of Orange County

AI Myth 1: AI will replace doctors

Slide suggesting machine learning is another step in process already underway for many years of automation in healthcare.

Move from ‘Evidence Based Medicine’ -> to -> ‘Intelligence Based Medicine

Intelligence Based Medicine / Medical Intelligence = machine intelligence + clinical intelligence

Datathon’ = data scientists + clinicians

AI vs Doctors - chart from IEEE Spectrum from 2006 to present.

See the original chart to click on any of the polygons to link to that full story.

Screen Shot 2018-11-05 at 2.27.15 PM.png

As machines take on more perception skills (image interpretation, data analytics); radiologists will need to shift what they spend time on during the day to more cognition tasks (decision making and problem solving). Shown with shift from red -> to -> blue map.

AI Myth 2: AI is about big data and deep learning

Chart of structured vs unstructured data in medicine. 89% of healthcare data is unstructured.

[I found this original graphic published in JAMA, June 24 2014. “Finding the Missing Link for Big Biomedical Data” by Griffin M. Weber, Kenneth D. Mandl, and Isaac S. Kohane.]

AI is a single lightbulb in a primitive hut” - AI is just a single piece in a much larger problem of digital healthcare data.”

“[in healthcare] we cannot even solve a one hour time change problem, whereas other industries have long solved this”

AI Myth 3: AI will dehumanize medicine (more)

Patient satisfaction doubles when you remove the computer in the room (eg. using human scribes), but hope is that natural language processing will be able to help in this domain - and at same time reduce physician burnout.

Rate that medical knowledge doubled in 1950 was every 50 years.

In 2020 that medical knowledge doubles is every 73 days.

2.3 Quantum Computing

Deepak Kaura MD MBA Chief Medical Officer 1QBit

Moore’s Law. Cool slide of the number of components on a chip doubling every two years, shown visually:

Moore’s predictions of computer performance are coming towards end (see plateau) despite transistor number continuing to rise.

Today using classical approaches we are producing components that are already very small - 22nm. As we shrink them down towards 7nm the price to do this rises dramatically. Note: a single strand of human DNA is 2.5 nm.

The current level of computing power is insufficient to compute today’s problems, such as mapping a simple molecule,

“If I try to map a caffeine molecule problem on a normal computer, that computer would have to be one-tenth the volume of this planet in size”

-Arvind Krishna, Head of Research at IBM

Therefore - a new approach is needed - hence - quantum computing.

Stages of computing:

  • classical -> quantum-inspired -> noisy intermediate stage quantum -> quantum.

  • Classical stages: electromechanical -> relay -> vacuum tube -> transistor -> integrated circuit -> quantum bits

  • Quantum bit jargon / paradigms being tried: superconducting loops, trapped ions, photonics, diamond vacancy, quantum dot, topological

Qubit = fundamental component. Like a switch.

Weird things at the quantum level: Quantum Superposition, Quantum entanglement, and many more on Wikipedia.

Current players

Very rapid progression in Quantum computing over last 1.5 years.

Benefits of Quantum Computing in healthcare:

  • Drug discovery (a wet-lab/bench lab in a machine)

  • Quantum computing to solve machine learning problems

New techniques always emerging that are better than the previous:

Eg. Reinforcement learning. This technique trains the algorithm by telling it what it should do and what it shouldn’t do. For instance - what are the rules of Go. Then the algorithm plays against itself to learn how to play the game and get better.

AlphaGo used this technique to train a new version of AlphaGo - called AlphaGo Zero. This version beat the previous version of their software trained on the ‘older’ technique of supervised learning (learning to play Go based on watching how humans play go).


Google DeepMid AI trained AI to ‘walk’ using reinforcement learning.

Q: How will reinforcement learning be used in robotics, factories, and healthcare?

Security & Encryption: passing comment (no time in this presentation) to get into details of the work being done in this area. [This is an area of particular interest / concern for me. Because my understanding is that quantum computing will essentially render all existing encryption protocols easily breakable and useless. Therefore new quantum-encrpytion techniques are required to secure data].

2.4 BlockChain 101

Nathana Sharma JD, MBA SU Principle Faculty, Blockchain Policy and Ethics

Many problems in healthcare are from (a) incomplete information and (b) fake information.

“WHO: Fake medicines kill 1 million people a year”.

Up to 1 in 10 medications in developing world are fake.

Mediledger Project (website)

  • Open and decentralized network for the pharmaceutical supply chain

  • Drug verification, change of ownership, contracting, chargebacks

  • Started in 2017, consortium of major pharmaceutical companies involved

  • [This project is super cool. I heard at Stanford MedicineX only a few years ago about how providence and validation of the pharmaceutical supply chain is one theoretical use of blockchain. It is cool to see movement in this direction. Lots of benefits to this in cutting out middle-men from pharmacy supply chains and getting drug costs cheaper and crowding out fakes.]

Very new technology: Blockchain first discovered - 2009. First time advanced computation done on blockchain - 2015

Some blockchain concepts

Peer to peer distributed information: helps solve the data deletion problem. [I also like the fact it is decentralized, which removes power of major companies or single institutions…theoretically…]

Cryptographically connected: ensures data remains in sync securely. Creates trust and cooperation between different groups. Good for complex systems.

Network data updating problem: solved with blockchain. Many different techniques to do so. And can track the information changed, who, and when.

Permission Blockchain: can control who sees which data. A project in UK is doing this with medical records. [more information on Medicalchain (on Medium post Nov 2017)].

Book suggestions: The Truth Machine: the blockchain and the future of everything (2018) - by Micahel J. Casey & Paul Vigna.

Break Webcast

Jessica Damassa

Shafi Ahmed, Professor of Surgery and Associate Dean, Barts Medical School

Medicine is exciting, because many of the new technologies all converge into medicine.


  • Intuitive Surgical - previous dominated surgical robot field for last 15 years (da Vinci Robot). Now an explosion of new companies. Cambridge Medical Robots CMR Surgical, Verb Surgical (Google with Johnson & Johnson). & 4-5 other companies.

  • In addition to adding robots to assist in surgery, addition of robots will bring better quality outcome data in surgery and the skill of the operator - which is definitely needed.



  • Training and education. Next step from paper/papyrus for translating information.

Medical Education

  • Why is curriculum 5 years? Why not less? [I agree 100%, the number is arbitrary]

  • We are not teaching new doctors and medical students, the skills they will require to use modern and incoming technologies. [agreed 100%, we are selecting entirely the wrong applicant for medical school, and teaching them the wrong things]

“If you accept what we do today as the best, you are in fact mediocre.”

Reason people return to xMED is because of the people they connect with and meet from around the world.

3.0 Session 3
Exponential Essentials Part 2

3.1 Digital Biology

Raymond McCauley Chair, Digital Biology & Founding Faculty, Singularity University

Direct to Consumer Genetic Tests: 250+ companies in USA, and 400+ companies worldwide. Can buy off shelf in Walgreens, on CVS or Amazon online.

eg. Origin3N - test for propensity for obesity and addiction. [They offer test to conference participants for free]

eg. basepaws: genetic testing for animals

Pharmacogenomics: your bodies response and interaction with drugs. The FDA has removed all previous limitations placed on 23andMe and they will be releasing this information again to customers. Company - Color - also doing this.

Epigenetics: changes to gene (DNA) expression that impact health. For instance smoking can change your epigenetics. eg. Chronomentary ($600-1000), Zymo’s myDNAge (~$300), Life Epigenetics YouSurance ($free, if YouSurance underwrites your life insurance).

Acquisitions: Illumina acquires former rival PACBIO for $1.2 billion.

Corporate rivals: BGI (Beijing Genomics Institute) was previously Illumina’s biggest customer of sequencers. Then BGI manufactured their own - BGI MGISEC-T7 sequencer. It can do 7 terabytes of data in single day (easily a single human genome). Illumina tried to block BGI’s sequencer, by going to the US gov’t suggesting that it will give China too much dominance in this space

Micropigs: BGI genetically engineered micropigs, and selling them as pets in China. In USA genetic engineering of pets has been banned. Difference in approach China vs USA.

“De-Extinction” Revival of ancient species. We can’t yet mail ordered an ancient animal, yet; but we are getting the DNA.

The 700 year old extinct giant bird - the moa - has had their ancient DNA recovered (news article).

Book: Wooly: The trust story of the quest to revive one of history’s most iconic extinct creatures by Ben Mezrich. (Amazon)

Cellular Agriculture: growing meat sells outside of an animal. (not really ‘genetic engineering’). Many companies working on this and already delivering on it: Memphis meats, Hamptom Creek, Modern Meadow, Beyond Meat, Clara Foods, Impossible [their burger is in over 1000+ restaurants], Perfect Day, Geltor, Ginkgo Bioworks, Integriculture, Shojinmeat Project, Finless foods, Sugarlogix, New Wave Foods, evolva, MosaMeat.

CAR-T Therapies: only two approved. Re-engineered cell from your own body.

Kymriah (by Novartis). FDA approved after Phase II. Treats Acute lymphoblastic leukemia (ALL). Half of patients in remission after 12 months.

Costs $0.5 million. It uses results-based pricing - meaning - if the drug doesn’t work you receive a refund. [This is very interesting, and I think this is a good strategy to start using elsewhere in healthcare devices and treatments].

RetiSpec: camera + AI examins the retina of eye to detect biomarker of cognitive decline - prior to clinical cognitive decline.

HUMM: a ‘small’ electrical current to the brain improves working memory, concentration, and visual attention.

Children’s science kit for 2018 - amino labs - can change bacteria color and smell.

3.2 Regenerative Medicine

Nina Tandon PhD, Co-Founder EpiBone

75% of people currently have some sort of artificial implant in them.

EpiBone: take CT scan to get structure of object to replicate, take adipose tissue from patient and extract from that stem cells that are attached to scafold in a ‘bioreactor’, that encourages differentiation and growth into bone cells. Within three weeks they have a new piece of bone or cartilage for implantation. Human trials starting in 2019.

Benefits: the piece fits perfectly, and it integrates well into body (because the tissue is alive).

Tissue engineering

First bladder grown from a patient’s own cells implanted in 1999. (Dr Anthony Atala, Wake Forest University) [news article].

Humacyte: regenerative vascular medicine is in phase III clinical trials

United Therapeutics - working on printing complex tissues, such as printing lungs.

Sangeeta Bhatia (MIT) - working to diagnoise colon cancer with lumninescent bacteria that can be detected in a urine test. [news article].

Question: will we end up with complex 3D printed organs vs xenotransplantation via pig organs? Tissue engineered approaches likely a decade away, vs xenotransplant trials soon underway.


A new field - starting to ask, “can we do _________, with cells?”. Uses biology to make thing.

United Therapeutics - produces a tobacco plants that can produce collagen. eg. “Using a plant, as a manufacturing plant”

eg. textiles: silk engineered using yeast

Ecovative: uses muhrooms to engineer packing materials, “mycelium biofabrication platform TM”

Plant-e harness - try to harness the electrons

BioMason - bioactive concrete with microbes that help heal the cracks in the concrete

What can’t be done?
What are the boundaries of the human bodies?

First industrial revolution = machines
Second industrial revolution = data
? Third industrial revolution = life ?

Provocative experiment & image from late 1990s

3.3 Deep Learning in Medicine

Jeremy Howard Deep Learning Researcher, Founder, Fast.AIDeep

Pivotal moment for Howard, was a few year ago when he saw this slide from World Economic Forum, “Imitating traditional development paths is impossible for merging economies. Nigeria would need over 700,000 additional doctors to reach OECD levels by 2030”.

  • A 12x difference in the number of physicians required if Nigeria wanted to match OECD levels.

  • This would take 300 years to train enough doctors to meet current shortage

  • It would cost 20% of their GDP (or 10x current public health expenditures).

  • Howard see’s AI as solution to this disparity. [I entirely agree - and think this is the exciting part of scaling medicine and where ML is going to have its biggest impact in healthcare].

Several years ago, nobody in medicine cared about deep learning. [I agree 100% - I remember being shocked at how literally nobody in mainstream medicine meetings or journals was discussing deep learning even a few years ago.] However, when Howard went to the 2018 Radiology conference, RSNA, deep learning was everywhere. Though, this has not caught on in medical fields outside of radiology.

Myth of skill loss: People say, if AI does medicine, people will forget how to be doctors. This is a myth. We have seen such line of thinking before, such as:

  • Traction control and ABS brakes - people will forget how to drive

  • Spell and grammar check - people will forget to write

  • Calculators and spreadsheets - people will forget to do math

  • Speech recognition - people will forget to type

Deep learning is not just a fad, it is “a fad 60 years in the making

Deep learning algorithm has filled in - and invented - what the picture should look like. eg, gap in photo below.

Baidu has speech learning algorithm that can recognize English and Mandarin speech recognition and translation better than most people.

Deep learning is not just for images, but also speech, and language and text too.

For instance, Deep learning using medical notes was able to predict heart failure better than other logistic regression systems. [Deep EHR: Chronic Disease Prediction Using Medical Notes, NYU. Jingshu Liu, Zachariah Zhang, Narges Razavian] - is making it easier to use machine learning without having to know as much of the math. Example presented comparing fastai vs1 vs PyTorch; and fastai uses less code, less time, and makes less errors.

Their goal is to reduce the amount of code required each year by half, and double the performance (accuracy and time). Goal = within next 2 to 3 years you will not require any code at all.

He presented at the end of the lecture a series of case studies of what people built who’s only experience was one year of coding experience, and after one week in his lecture. The results are stunning.

WAMRI - Wicklow AI in Medicine Research Initiative - group to connect USF Data researchers and your institution’s medical data challenges on academic projects.

3.4 Exponentially Human

Carmen Morcos MBA CEO, Impact Visionary

Word exponent, comes from word expound - which means to ‘set forth’ ‘to explain’.

Social determinants of health - we now have technologies to measure these.

Q: What is your highest value in life?

4.0 Session 4
Community, Patients, Countries Included

Drawing the Inside Out

Anita Ravi MD MPH FAACP Founder & Clinical Director, Institute for Family Health’s PurpLE Clinic

PurpLE Clinic - focusing on marginalized patients in New York City.

Example of barrier to ‘quality healthcare performance metrics’ like high cancer screening rates:

Ravi asked a patient who always comes on time how they are able to do this. The answer? The mother and child left for the subway three hours before appointment. They would then beg for a subway ticket. If they couldn’t get a ticket, they would jump the turnstile - which places them at risk of arrest. (Since hearing this patient story, the clinic now provides MetroCards).

Problem of scribes: may make patients less likely to share personal stories and details, in particular if there is a history of violence or sensitive topics.

Some patients, have such high anxiety about going to the doctor, that it results in delayed and late presentations.

How do we design for patients who do not have a home? How do we represent these patients in data if they do not have a ‘zip code;?

How do we think about healthcare ‘quality’ metrics? How do we take into considerations patients for whom they have very different priorities and barriers to care?

Ravi is worried about patients in particular when they do not show to clinic. [This is a complex problem]

When looking at data: always consider ‘what data may be missing’. Who has not been captured by the data because they are marginalized? What voices have not been heard?

The Coming Explosion of Health Data Spigots

ePatient Dave DeBronkart ePatient

His new book: Super Patients. He has given over 600 talks in 19 countries.

“The difference between analysts and activists is if you study a problem for five year, the analysts just sell more reports, whereas the activists get pissed off.” Dave is an activist - he wants change.

He used Google Health in 2009, and found out the data he got from his hospital was garbage.

Gimme My Damn Data” - term coined in 2009 at Medicine 2.0 conference.

#PatientsIncluded hashtag invented by Lucien Engelen - who said he would not speak at any conferences that did not include patients.

Amenable Mortality - deaths in a country they had the financial / technical ability to cure. [This data is from global burden of disease study]. The USA ranks 181 in the world.

“Let patients help”

“there are a number of signs [healthcare] is optimized around something other than consumer need” - if you are the patient, this is a problem.

The patient is the ultimate stakeholder. They have the most at stake!

Mixed incentives. Goldman Sachs: ‘Is curing patients a sustainable business model?’. [CNBC April 2018 Article]

Dave asks, have you ever been in an industry that got “the bottom of the steamroller of disruption”? Previously he worked in graphic arts (tech marketing) and this was overturned by desktop publishing.

Then the end-customer started using the tool themselves to make the product. At first they didn’t know how to use it well. “You were stupid with fonts”. But people got better over time.

We will see the same in healthcare as data becomes available.

#WeAreNotWaiting - patients starting to work on healthcare innovations themselves. The NightScout project to create artificial pancreas to help monitor diabetes. They have been able to create something the industry said is not possible or ready.

An example of patient’s taking their own health (and life) into their own hands, because healthcare doesn’t view their condition or disease a high enough priority to expedite a treatment and solution for it. And it is working.

Liberate an industry by freeing the data.

Example by @KatieMcCurdy (a graphic designer) on how she displays her health history in graphical form. [Founded company Pictal Health to help doctors and patients visually communicate]

Example of Kate Sheridan. Who had deliberating case of Lyme disease (from star student to unable to read a single page). Saw 30 doctors and received 15 diagnoses.

Her mother typed into excel over 1.5 feet worth of medical papers. Imagine how much easiser this would have been if they had ‘data spigots’.

Example Michael Morris, diagnosed with Stage IV Colorectal Cancer. As a software developer built an amazing tool to monitor his health between treatments, and pull his health data from all four of his hospitals using FHIR.

[See screenshots below. It may It may be one of the best displays of a medical record I have seen. I need to look into this in far more detail… Post on Health Intersections Oct 10 2018

Video of Michael showing his Cancer Dashboard at September 2018 HL7 FHIR conference]


When an industry that previously closely held data, has their data available - chaos will ensure.

Chicken and egg of healthcare data.

The CMIO of Dave’s hospital (Beth Israel Deaconess) blogged that “download [of healthcare data] makes little sense since at the moment there is nothing a patient can do with a download. Of the 2 million patients at BIDMC, not one has ever requested a download.”

…Perhaps there is nothing that one can do with it, because the data available isn’t actually all there or in a sharable format?

He looked at his hospital’s website, they in fact did not even have a button to download their data.

So he emailed the hospital, and was directed to medical records who could help him get a digital copy of his data. Medical records of course said, ‘they don’t have a way to do that’.

[This story is crazy. And infuriating]

“Don’t become a proud expert on an obsolete model” - Dave

“Enable radically disruptive data spigots: set yourself on FHIR”

An xMed Adventure

Shafi Ahmed Professor of Surgery and Associate Dean, Barts Medical School

A fun story of where connections from communities and conferences lead.

Shafi met the Bolivians at xMed last year, as well as Hector Valle (from Mexico).

This led to a new Conference Digital Health Forum in Mexico.

He then met Martin Dockweiler - who invited him to South America. Martin has funded four universities in Bolivia.

The itinerary Martin set up for him:

“They are waiting for you”

When walked out of plane, national news had been interrupted to stream Shafi landing and walking through airport.

Met with President and Vice-President of Bolivia. Received the Keys to the City. Was awarded with an honorary PhD.

Martin sat down with Shafi and told him of the issue of healthcare in Bolivia. How the country is missing a large hospital, and he wanted to create a digital hospital of the future. Hence, why he is talking to Shafi / why they need to work together.

Also, Martin suggested they name the hospital; “Shafi Ahmed Martin Dockweiler University Hosptial SAMD”

Ha. Great story. Also, talk about knowing how to flatter a prospective client.

Moonshots & Exponential Entrepreneurship

Naveen Jain Founder & CEO, Viome

It is so much easier to solve a really hard problem. Because then the best in the world come to want to come and help.

You have to continuously re-think the problem. This means an ‘expert’ in the field may be the wrong person to solve it, because potentially the current line of thinking in an industry may be the wrong next-level solution.

Can you take the genes from bacteria that have evolved to be resistant to radiation, and use CRISPR to insert that trait into humans?

“Because I know nothing about medicine, I am the most dangerous person to disrupt the system. I challenge every fundamental assumption”

5.0 Session 5
Keynotes: Catalyzing the Future

Bringing Healthcare Systems into the Future

John Halamka MD MS Chief Information Officer & Dean for Technology at Harvard Medical School

Problems the same around the world

  • aging societies

  • longer lives, more costs

  • increasing mental health burden

  • low birth rates, means harder to pay for this

  • not enough clinicians, and the distribution of services is problematic in rural areas

Some things John has seen this fall

China: all the medical record systems there are silo’d and do not talk to each other. Data portability is a problem. They are moving very fast to do pilot with universal ID.

Japan: cloud computing good option for a country with many physical threats to local data centres.

New Zealand: there is only one medical record for entire population of 5 million.

UK: there is a single emergency treatment record so you can get critical information.

Nordics: Finland passed a law that all someone’s lifetime aggregated de-identified medical data will be available for machine learning research.

Washington: Problem of meaningful use, is that each small group suggested mandatory fields, and now there are 140 required data elements for each visit - which is not possible. Now priorities are (1) managing medication safely (esp opioids), (2) sharing data with patients, (3) using data for public health program, (4) building application program interfaces for interoperability

India: Aadhaar Identifiers, identifies every person in the country. Gates Foundation TB Projects starting to wonder

Africa - $100 finger print patient identifying project using Raspberry Pi.

What should we do?

AI/Machine learning:

  • “if AI can replace your doctor. AI should replace your doctor”.

  • Reduce burden of documentation.

  • The operating room schedule Beth Israel Deaconess uses machine learning to calculate OR time slots. How much times does this thin guy with no co-morbidities need for an appendectomy? It freed up 30% of the OR schedule.

  • We need tools to help patients better navigate the system. AKA triage.

  • Machine learning reads all the faxes that come into Beth Israel

Mobile/Internet of Things: These tools are not because we are self obsessed. But helpful to patients. John used these mends to titrate his metoprolol to the dose with the least side effects and best physiologic dose for him.

Big Data/Interoperability: 26 EHRs in Boston area. APIs will allow proper data analytics. Harvard has 17 hospitals, and used it to join the data together to mine. He found increased side effects for breast cancer therapy with Taxol causing neuropathy, and so they dropped his wife’s treatment dose so she could reduce her risk of neuropathy and continue as an artist.

Telemedicine/Telecare: John does 900 telemedicine calls a year. Why? Because he is the national expert in poisonous mushrooms and plants. Policy is problem - as licences are still regional yet he fields calls from all across nation.

Blockchain: good use cases: (1) auditing/integrity to show that medical records have never been changed, (2) patient consent, (3) micropayments

Three ways to influence a clinician: (1) pay them more, (2) make their lives better (3) avoid public embarrassment. Do all three at once.

Now that BIDMC is paid for outcomes, instead of admitting the same patient every time the temperature rose too high for a $20,000 COPD exacerbation, they went and purchased that patient a $269 air conditioner and the patient is doing well at home.

John doesn’t think that the EHR companies will be the new innovative companies, but those companies that are built around the EHR.

Uniting the Toolset and Mindset of Innovation

Tom Wujec Founder, The Wujec Group

Innovation is solved with (1) Toolset, (2) Mindset:

Random fact: this year will be produce more transistors than we will harvest in grains of rice.

“Sensors: the age of guessing is officially over”

Generative Design: using software with constraints and inputs to design physical objects.

The Next Rembrandt: scanned in past Rembrandts, and asked computer to generate a new one with specific characteristics. [website]

Board game for people newly diagnosed with diabetes to play to learn about diabetes an develop a similar (and healthy) mental model about diabetes. Game: “Conversation Map”

You may be interested in the Notes from Day 2, Notes from Day 3, or Notes from Day 4.

All slides are from the Exponential Medicine conference live stream November 4, 2018. Available on YouTube. Watch past Singularity University videos on their channel.

To register for next year’s conference, please visit their website, It sells outs, so register early.

Exponential Medicine, Day 2 - Cool Stuff Summary (Nov 5 2018)

Exponential Medicine, Day 2 - Cool Stuff Summary (Nov 5 2018)

Apple Event October 30, 2018 - Major Themes

Apple Event October 30, 2018 - Major Themes