Patients Without Traditional Cancer Screening Find Allies In Maccabi And Medial EarlySign

By Benjamin Ross

September 8, 2017 | 2017 Best Practices Awards | Maccabi Healthcare System, Israel’s second largest Integrated Healthcare Delivery Network, is looking to find colon cancer patients among individuals who didn’t go through traditional channels for screening. ColonFlag, a clinical decision support tool developed by Medial EarlySign, an Israel-based developer of machine learning tools for data-driven medicine, is used by Maccabi to indicate individuals in the Israeli healthcare system who have a high risk of having colorectal cancer.

ColonFlag analyzes routine blood tests of people who have not participated in screening programs and flags those at high risk of having colorectal cancer.  This enables Maccabi to focus their colonoscopy resources on high risk individuals, to help discover CRC cancers early, which may have otherwise gone undetected until later stages.

ColonFlag was implemented at Maccabi in October 2015. The goal of the implementation was to identify individuals who, for whatever reason, didn’t participate in the traditional guaiac fecal immunochemical test (FIT)-based screening program for colorectal cancers, but may still be harboring disease.

“People are sometimes afraid that something will be found, or are just afraid of colonoscopies in general,” Ori Geva, CEO and co-founder of Medial EarlySign, told Bio-IT World. “What we’re doing is providing information to the healthcare organization so that it can decide the most efficient way to get a patient engaged.”

Medial EarlySign and Maccabi used ColonFlag to analyze 80,000 complete blood count (CBC) test results over the course of a year, factoring in age and gender, to identify individuals at a high risk of colon cancer. General practitioners were then sent alerts in the Electronic Medical Records (EMR) system for those patients who scored above the threshold.  GPs reviewed this information and then determined if patients were suitable for further evaluation or colonoscopy. In practice, 85% of the individuals who saw their GP after being flagged by the ColonFlag system were referred either to a colonoscopy or a GI specialist.

“Everyone is trying to focus on the population at risk, of course trying to prevent disease before it occurs or before it gets worse,” said Medial EarlySign’s VP of Corporate Marketing, Tomer Amit, when speaking with Bio-IT World. “What we’re saying and what we’ve proven with Maccabi is that, with our technology, we can do it better.”

Amit says that the typical statistical technologies can only get you so far. Combining the machine learning and AI technology of Medial EarlySign with the experience of general practitioners in Maccabi’s healthcare system reveals better results.

690 individuals were flagged by ColonFlag as high risk. Of those, 220 colonoscopies were performed and 42% had findings: 20 with cancers (10%), 6 at high risk of developing some form of cancers (3%), 17 at intermediate risk (8%), 27 at low risk (13%), and 18 with no current risk (9%).

According to Medial EarlySign, “ColonFlag’s underlying core technology allows investigators to easily define a medical outcome and iteratively create a predictive algorithm, training and validating it numerous times against large volumes of medical data. The data is stored in condensed format and swiftly scanned and indexed, allowing for very fast access by the companies’ analytical algorithms.”

These algorithms consist of dedicated machine learning tools, as well as conventional algorithms, which are broken apart and rewritten to perform optimally with the data repository. The unique algorithms and storage format enables performance at considerably higher speeds than standard databases, according to Medial EarlySign

“Our core technology —an AI and machine learning framework that we developed over the years—is tailored to work with massive amounts of clinical data,” Geva said. “We’re able to deal with a gamut of data elements that originate from health records. This framework uses the most advanced machine learning techniques and medical algorithms out there, including our proprietary data repository. We’re able to tap into existing health records data and, by using very large databases to create these models, we can connect the little pieces of information that identify those individuals with a specific risk for a medical outcome. We can then create a model that will look at an individual person’s data and indicate or flag if that individual belonged to a group that has a much higher risk for that specific diagnosis.”

Medial EarlySign and Maccabi’s implementation of AI and machine learning technologies in the clinical space earned them a Bio-IT World Best Practices Award for Clinical IT & Precision Medicine this past May in Boston.

“We’ve seen that it works.  A healthcare delivery network cannot afford to have their physicians spend their precious resources on all people that have some care gaps,” Geva said. “In most cases, an informed discussion between the physician and the patient increases the probability that the individual will eventually get screened.”

For now, Maccabi is looking to use ColonFlag as it was intended: for colorectal cancers. However, Medial EarlySign sees ways for both organizations to gain something from the experience of their collaboration.

“I think [Maccabi]’s looking at other creative ways of using this product to get more people into the scope without increasing the use of a lot of potential resources,” Amit said. “One of the advantages they have for their advanced use of data is that they see we can create additional collaborations around this, and I think they’re doing that. They’re also working with other pharma companies, getting more for their data. I think this experience has shown them that investment in an advanced medical data strategy is useful.”

Editor's Note: This story also ran on Clinical Informatics News.