This article was part of a series on data science in the lead up to Science for health 2021, where Dr. Gabriel Chodick took part as a keynote speaker.
Two days before his 50th birthday, Dr. Gabriel Chodick felt a pain in his abdomen. He turned to an app on his phone, entering his symptoms for a self-diagnosis. The result: appendicitis. Skeptical, as his age and symptoms didn’t fit the usual profile, Chodick went to his family doctor for a physical examination, but the doctor wasn’t sure either. He asked Chodick to return the next day – then, he was finally able to confirm the initial diagnosis, and Chodick was sent straight to hospital for an emergency operation.
The app that correctly diagnosed Chodick’s appendicitis – well ahead of his family doctor – is an AI platform called K Health. In one of life’s quirks, the app owes its existence to Maccabitech, the research institute where Chodick is Head of the Epidemiology and Database Research department. The app is one of many success stories emerging from Maccabitech, where data science is being used to drive healthcare innovation.
A gold mine of data
Maccabitech is the research arm of Maccabi Healthcare Services, one of Israel’s four health maintenance organizations (HMOs). With 2.4 million members and 9000 independent physicians, Maccabi is the second largest of these non-government, non-profit HMOs, and has a huge database of electronic medical records.
“We have a wealth and depth of data, going back 30 years, which allows us to support advanced research to improve healthcare.” – Gabriel Chodick, Maccabitech
“In the 1980s,” Chodick explains, “Maccabi wanted a better way to keep track of the care being provided to its members. The HMO decided to computerize all physician’s clinics, a goal which was achieved by 1993. Everything is covered by these electronic medical records: symptoms, diagnoses, tests, prescriptions, hospital procedures and more… We have a wealth and depth of data, going back 30 years, which allows us to support advanced research to improve healthcare.”
The harmonization of Maccabi’s data makes them much easier to work with than records in more fragmented healthcare systems. Members can opt out at any time, but very few choose to do so. In fact, 10 percent of Maccabi’s members have opted to additionally take part in a Biobank, where a small part of their test samples are stored for future research.
“The biobank already contains over 500,000 biological samples,” Chodick says, “It shows you the great trust our members have in Maccabi. The combination of medical records and samples is a great recipe for 21st century personalized medicine.”
Protecting privacy while promoting innovation
Maccabitech was established to facilitate safe and ethical access to Maccabi’s rich database. Research projects are usually initiated by external collaborators, which include a range of academic and industry partners.
“We’re very careful that the data never leaves Maccabi,” Chodick explains. “We’ve set up a system where researchers can log in to the Maccabitech servers and work on the anonymized data remotely, but they don’t have direct access to the raw data and can’t download them. We also support some partners by conducting the analysis in house and simply providing them with the results of the study.”
“The future lies in having many researchers using big datasets without jeopardizing the privacy of patients, and there are very interesting technical solutions to overcome those problems.” – Gabriel Chodick, Maccabitech
Maccabitech uses the MDClone platform to manage data anonymization, a platform which also enables the generation of synthetic datasets.
“The MDClone platform is gaining popularity in the US where all the big players in the HMO market are involved in data access and science. The future lies in having many researchers using big datasets without jeopardizing the privacy of patients, and there are very interesting technical solutions to overcome those problems, including synthetic datasets. I think most people understand that data are extremely valuable and should be shared for the benefit of all.”
Training AI to help doctors
One of Maccabitech’s greatest success stories is the collaboration with K Health. Used by 5 million people and worth $1.5 billion dollars, this digital health app was made possible thanks to Maccabi’s data.
“The founders of K Health approached us asking to analyze notes taken during doctors’ visits. After an intensive de-identification process, we were able to extract patient complaints proceeding diagnosis in several hundred million records. K Health used these data in combination with deep learning methods to create a self-diagnosis app. Most health apps use rule-based pre-programming; K Health was the first to create an app that just gets better the more it is used. It shows the power of combining AI with big datasets to create new tools that can be used to supplement the services of physicians.”
Read this article to learn more about why keeping up with data science matters for our health!
Like K Health, Israel’s digital pathology company Ibex also got started with Maccabi. Using medical images from Maccabi’s database, the start-up trained a machine-learning algorithm to recognize cancer in biopsy slides. Within four months, the machine was 98% accurate; Maccabi started using it as a second opinion to physicians’ diagnoses. On its first day of operation, the machine caught several instances of cancer that the doctors had missed. Maccabitech now receives royalties from the use of the Ibex technology, and Maccabi is free to use the screening tool for their members at no cost.
Using data to improve healthcare
In addition to supporting external innovation, Maccabi has used insights gained from its dataset to improve its own services. One of these examples was in treatment non-adherence, a huge issue in healthcare around the world:
“We did a study on statins – medications used to prevent cardiovascular diseases. We found that a third of Maccabi patients who were prescribed statins discontinued the therapy within a few months. We’re fortunate at Maccabitech that – unlike most academics – we’re embedded in a healthcare organization and can take practical steps towards patient impact. We implemented an alert in the electronic medical records so that a notification is sent to a patient’s physician if they discontinue their therapy. The physician can then follow up with their patient. We’ve already been able to see the real-time results of this simple solution, which has increased treatment adherence dramatically. We were able to bring this study full circle, from research finding to improved healthcare.”
The HMO’s records have also been used to glean invaluable insights throughout the COVID-19 pandemic. Chodick tells of one particularly impactful example of academic research that took place in 2020:
“In the early days of the pandemic, some researchers published speculation that ACE inhibitors – a type of hypertension medication – might encourage viral infection. Now, we have over 300,000 hypertension patients in Israel; if even a small percentage of them stopped taking their medication, we would have had an enormous rise in heart attacks and strokes. It was vital to establish if there was any cause to the claim as quickly as possible.”
“In the hands of the right people, big data can benefit the world.” – Gabriel Chodick, Maccabitech
“We analyzed the data, working 24 hours a day to compare COVID-19 cases and controls in the Maccabi dataset – in the end, we didn’t find any evidence for a detrimental effect of ACE inhibitors. Thanks to our well-organized database we were able to show – rapidly and using real-world data – that there was no cause for physicians to advise their patients to discontinue their medication. It’s a simple but critical example of how researchers, given the right tools, can help physicians in their struggle during the pandemic. In the hands of the right people, big data can benefit the world.”
Don’t let data go to waste!
Despite success stories like Maccabitech, progress in the use of healthcare data is sluggish in most countries. Often it’s politics, not science, that presents the biggest hurdle. Exacerbated by a lack of up-to-date regulations, conservatism surrounding issues of data privacy and ownership means that data which could be mined for insights – to improve the health of billions – is languishing unused. It is up to all of us to make sure that this treasure trove of potential can be safely unlocked.