by Josh Mitnick

Israeli HMO Maccabi taps AI, big data for healthcare innovation

Aug 21, 20199 mins
Artificial IntelligenceBig DataElectronic Health Records

The HMO has managed innovation through its health tech arm, big data institute, and partnerships

health doc tablet team
Credit: Getty Images

The headache has been hanging around longer than usual. Aspirin only dulls the throbbing temporarily. So you check the internet. But the possible explanations range from depression, to a stroke, to a brain tumour.

Now, an interactive mobile app promises users a more precise initial diagnosis. Developed using the data from more than two decades of doctor diagnoses chronicled in the records of Maccabi Healthcare Services, an Israeli health maintenance organisation (HMO), the K app queries users about symptoms, medical history and basic personal information. Then, it suggests a possible ailment with a much higher degree of reliability than the results from an internet search.

The K app is the fruit of one of several partnerships – in this case, with New York-based startup K Health – spawned by Maccabi Healthcare Services’ eight-year old innovation centre, Maccabitech, and its big data institute, MK&M, launched in 2016.

Israel has a universal healthcare system with four HMOs financed through income taxes and monthly premiums. Maccabi is the second largest HMO by membership.

Managing big data for healthcare innovation

At a time when technology executives everywhere are striving to find innovative ways to analyse and manage huge volumes of data to generate new services and revenue streams, Maccabitech has figured out how to work with partners to mine and derive new value from its own vast trove of patient data. In doing so it is helping to create new healthcare services that have the potential of benefiting not only its own members, but patients far beyond Israel’s borders.

Several years ago, Maccabitech director and founder Varda Shalev decided to harness the data from the HMO’s two-decade archive of electronic medical records (EMRs) and make it available for research projects led by external companies and institutes. The EMRs are considered a data gold mine for medical innovation. 

“Our purpose was to enable more researchers, physicians, and entrepreneurs to access big medical data, and develop novel tools to advance personalised medicine,” Shalev said.

The collaboration so far has produced 10 technology startups as well as research projects with IBM and Israel’s Technion Institute, all of which are aimed at improving the health and services for the HMO’s 2.3 million members. 

Big data analysis aimed at saving lives

The goal of the partnerships is to advance precision medicine – which aims to customise healthcare by predicting the future of a patient’s health and providing prescriptive advice for treatment, said Shalev.

“We use our data to create value for the organisation,’’ she said. “Big data in medicine can save lives and has the potential to disrupt the way clinical treatment and research are conducted.”

The innovation centre, which operates independently from Maccabi’s healthcare services, received donations from Israeli philanthropists Morris Kahn and Sami Sagol. The individual projects are funded by companies and academic institutions.

Research proposals are first reviewed Maccabi medical and ethics experts. Then, researchers are given access to the Maccabi big data platform. The platform, developed by Israeli startup MDClone, is operated under the auspices of Maccabi’s big data institute.

To preserve patient privacy, information on the platform is anonymised. The anonymisation process includes feeding patient data into MDClone’s synthetic data engine, which then generates a new data set that is statistically similar to the original but containing no actual patient information.

The data is made available over a web interface that allows scientists to search according to criteria such as lab results, demographics, and disease diagnosis. The platform is designed to give any researcher throughout the world the ability to work with Maccabi data.

About the time that Maccabi was getting ready to roll out its big data platform, a group of Israeli entrepreneurs in New York with no background in medicine were trying to figure out how people could get personalised health information online.

Siloed data is a stumbling block for analysis

They realised that the electronic medical records in the U.S. would not provide a good source of data: there was data on billing, diagnosis and treatment, but they lacked information on the decision-making thought process. The data was also siloed among different health care organisations, making it difficult to understand a patient’s treatment journey.

“That led us back to Israel, where we knew that there was a much more significant presentation of a medical journey in the EMR system,’’ said Ran Shaul, K Health co-founder and chief product officer about his first outreach to Shalev. “I said, ‘We want to solve people’s ability to search online and better understand their health condition.’ She was used to people trying to solve doctors’ problems, not consumers’ problems.’’

Through the Maccabi data platform, K Health and the K mobile app has access to 400 million doctor-patient consultations and 1.4 billion lab results. The EMR data provides a rich description of the medical context and insight into doctors’ decision-making processes. “It’s a robust, fully integrated medical record,” Shaul said.

AI tech powers healthcare app

K Health built four layers of artificial intelligence (AI) technology that transforms the case histories of 10,000 Maccabi doctors so it can be used to advise people who download the app.

Using a natural language processing neural network architecture known as long short-term memory, K Health was able to analyse notes from doctor-patient meetings and transform them into metrics that the computer understands.

The next layer of neural networks organises and classifies that structured data so the K Health app engine can compare cases of patients with the same medical profile suffering from the same symptoms.

In the third layer, a computer algorithm creates a logical sequence of questions for the K Health app to ask a user. The fourth layer feeds back the answers to the K Health app, enabling it to learn in real time about the user’s condition.

Shaul said the success metrics they followed in developing the app were medical accuracy, completion rate of dialogues, the ability of users to understand questions, and whether the user could understand what course of action to take. “In the last three years, I’ve learned that medicine is not as binary as it seems,’’ he said. 

None of the four co-founders of K Health came to the project from the world of medicine. “When you come from outside, and understand data, you can see how it can change an industry. We didn’t have any” predispositions, said Shaul. He added that initial development of the AI technology took about one year.

K Health users in the US can consult with a doctor after getting the initial analysis from the app of how patients with similar profiles and symptoms were diagnosed and treated. The app can help many users to take care of their ailments without making a doctor’s appointment, reducing stress on a health system.

Shalev said that eventually, users would be able to get referrals for medical tests without the need for a doctor. She expects the K Health app, launched in the US in 2018, to be rolled out and integrated within Maccabi’s own system within a few months in Israel.

Meanwhile, K Health last month teamed up with Anthem in the US to develop an app for the insurer’s 40 million members that will allow them to chat with a doctor for a fee that’s less than a regular co-pay. K Health, which has already raised US$50 million in venture funding, will get revenue from Anthem’s as well as investment capital from the insurer.

Working with partners on healthcare services

In addition to K Health, Maccabi data has been used to train AI algorithms that are being used by different startups for, among other purposes, identify early warning signs of colon cancer from colonoscopy images, perform back-up reviews of prostate biopsy pathological slides to identify false negatives, and analyse voice samples to detect heart disease. In addition, Maccabi’s mammography images are being analysed by IBM to improve breast cancer diagnoses.   

“What we try to do is use our data to create research and companies, and then put it to work for our patients,” said Shalev.

She said that most partnerships with startups involve Maccabitech getting royalties in return for access to the data.  Maccabitech also advises startups on data study and potential applications.

Despite Israel’s rich electronic medical records, obtaining timely regulatory approval for research and information sharing remains a major obstacle for big data projects. Shalev said hospital and health ministry ethics committees lack the expertise to understand the nature of the proposals.

“They do not understand what the essence of large data set research is, and how to approach it,’’ she said.  “Regulation in Israel does not allow for the implementation of innovation in medicine.’’

As for K Health, Shaul insists that company has only scratched the surface of what can be done with Maccabi’s data. It has yet to come up with an engine that can make use of data for mental health, chronic disease and cancer. 

“The data’s value is for years and years to come. We are going to change medical protocol,’’ he said. “The way we are thinking about our health today is going to change.”