A New Kind of Clinical Trial

The kickoff

A couple of weeks ago at HQ, I was lucky enough to join an inspiring team that came together to finalize plans to launch the first of a new kind of clinical trial. As well as our’s internal team, other advisers and collaborators in attendance included:

  • Prof Chirag Patel, PhD (Harvard Medical School biomedical informatics)

  • Prof Arjun Manrai, PhD (Harvard Medical School biomedical informatics)

  • Udi Manber, PhD ( advisor, previously Head of Search at Google Chief Scientist at Yahoo and Chief AI at Amazon)

  • Alan Greene, MD ( co-founder , creator of, which has advised over 100 million unique visitors over the last 20 years)

  • Chethan Sarabu, MD (Director of Clinical Informatics at, Stanford Children’s Hospital)

  • Neeraj Kashyap, PhD, Principal Investigator “Can AI Predict your risks for allergies?”

What was it that had such a group of people coming together at And what do I mean by a “new kind of clinical trial”? First, some background…

Background: AI and medicine

When I started Enlitic, the world’s first organization focused on bringing deep learning to medicine, the words “AI” and “medicine” were rarely heard together in the same sentence. But I knew that the opportunities were extraordinary — by combining the expertise and experience of doctors with the unparalleled ability of neural networks to crunch vast amounts of medical data, we could give every doctor superpowers. The most immediate opportunities were in radiology, since the use of deep learning to analyze images was already well understood, and every modern hospital already had an extensive archive of digital medical images. Today, the power of deep learning to help radiology is widely accepted, with the National Institutes of Health (NIH) releasing papers and hosting workshops about AI in radiology, the American College of Radiology (ACR) and MICCAI collaborating on a new AI partnership, an annual Conference on Machine Intelligence in Medical Imaging (I’ll be key-noting this year’s conference, so come along and say hi!), and many more examples.

But radiology is just the tip of the iceberg — the opportunity is just as great in every part of diagnostics and treatment planning. The problem is that we don’t have the access to the data to allow us to take advantage of that opportunity. The data is spread over many places — your pharmacy has your medication records, your lab test provider has your labs records, your hospital records may be spread across multiple institutions and departments, your personal health data (such as from your wearables, meal logging, etc.) is on your phone, and so forth. And all this data is in different, incompatible formats, and much of it is not available to researchers due to regulation and privacy concerns. has come up with a really clever solution: create data import pipelines for each of the major healthcare data sources (e.g. for the major pharmacies, labs providers, hospital networks, etc.) that can run on your phone. Allow you to run an app on your phone that can grab all of your medical data and consolidate it in a single place, in a single format. You can use this app to get “at a glance” information about your health, based on all this information. Then, if a researcher thinks that they might be able to use your data to help diagnose and treat other patients, they can ask your permission to share it with them, and compensate you for it.’s first data trial, with our advisors from Harvard and Anthem

After much hard work, this week we are ready to try out this idea. The mobile app is ready to download, and the importers are written, so anyone can now, for the first time, truly take ownership of their own medical data. And also for the first time, people can now choose to contribute this medical data to helping advance medical research, by enrolling in the first data trial: “Can AI predict your risks for allergies?’

The trial is being run with our advisors at Harvard Medical and our team, with the support of Anthem. The goal is to better understand what triggers allergies, to help allergy sufferers (like me!) avoid the irritating (and sometimes exhausting) symptoms.

Data scientists (like me!) will be able to analyze the data to try to identify previously unknown patterns and relationships, which medical researchers can then use to develop new plans and treatments for allergy sufferers.

It’s such a simple idea, and if you’re outside of the medical mainstream you might be surprised at how revelatory this is! Previously, patient data (although, in theory, owned by the patients themselves) has in practice been locked up in silos. It was reported that “a recent Stanford University study found that 93 percent of medical trial participants in the United States are willing to share their medical data with university scientists and 82 percent are willing to share with scientists at for-profit companies”. There have been some attempts to give patients the ability to share their data with researchers, such as the UK Biobank, which have been extremely well received by patients. However, they’ve been limited in scope and geography.’s approach is the first to really let any patient take ownership of their health data in such a comprehensive and straightforward way.

For tabular data of the kind we’ll be using in this trial, data scientists generally use an ensemble of decision trees, such as a gradient boosting machine, or a random forest. If you’re interested in learning about these powerful techniques, I’ve made an online lesson available to introduce random forests — which I expect to be amongst the most useful tools here. Random forests are both extremely accurate (I have used to win international machine learning competitions in the past) and highly interpretable; they can show (amongst other things) which features are most important, how they are related, and how individual decisions are made.

Snapshot from’s Intro to Machine Learning course

I’m very excited about this new kind of clinical trial. If it takes off, it could dramatically change the technology curve in medicine.


As you may know allergy symptoms may arise when exposed to changing weather conditions, pollen counts, and even air pollution. Can these events predict triggers — in advance — of the onset of an allergy? With you we want to learn if AI (artificial intelligence) may be used to predict when people will get allergies or allergy patterns.

  1. DOWNLOAD THE APP on the APPSTORE or access VIA WEB (Chrome only)