Unlocking patient data from medical reports

May 22, 2017 Article Janssen Pharmaceutica

The demand for real-world evidence in both clinical practice and pharmaceutical research is not new by any means. However, the majority of such data remains hidden in patient records in formats that are not analyzable. Belgian start-up Lynxcare has created an automated text-mining technology that can extract all relevant data points from these reports. At this year’s FlandersBio–Janssen Pharmaceutica Partner Day, Lynxcare CEO Georges De Feu explained how they can make important data available by working together with patients, hospitals and pharmaceutical companies like Janssen Pharmaceutica.

Treating instead of coding

Much of the information that’s hidden inside physicians’ reports can have big impacts on patient outcomes.

In a typical patient record, only a small part of the data (such as the main diagnosis) is stored in a coded format. Extensive information on the diagnostic process or subsequent treatment protocols is entered by the physician as unstructured data. “Much of the information that’s hidden inside physicians’ reports can have big impacts on patient outcomes,” says De Feu. “Physicians want to treat patients; they don’t want to invest too much time in coding,” he continues. “Studies have shown that physicians who make an effort to enter structured information invest 5–10% more of their time on this than those who don’t. They could do a lot of surgeries in this time.”

Artificial intelligence “understands” physicians

Other companies, like US-based Flatiron in oncology, stimulate the collection of structured data by providing hospitals with free software. In return, they sell the anonymized data to the pharmaceutical industry. This is a great business model, but it relies on convincing hospital management to change their Electronic Medical Record (EMR) system. “Our solution uses the raw data of any hospital EMR. This incentivizes the hospital management as they don’t have to do these huge implementations,” De Feu explains.

Lynxcare’s algorithm extracts relevant data points from the original reports using automated text-mining technology. The core of its system is a flexible, natural-language processing pipeline that is able to adjust the output of the system depending on several contexts. This allows it to handle different languages and the significant inter-hospital and even inter-department variability that exists in the way reports are written. “Due to local customs, physicians in different hospitals use different abbreviations,” De Feu clarifies. “On the other hand, the same abbreviation can mean something completely different in orthopedics than it does in cardiology. The artificial intelligence driving our system can handle this type of contextual variability, allowing an accuracy of 95% or more on most of the parameters we extract.”

Creating value for the patient

So, we provide data back to the patient and the physicians, and we give our insights to the department that’s using that data to ameliorate results and procedures.

Naturally, the patient needs to give his consent to collect and analyze his data. It was feared that Europe’s General Data Protection Regulation might be a hurdle for innovation in the pharmaceutical industry. However, Lynxcare views it as an opportunity to work with the patient and the hospitals more closely on real-world evidence data. “When the patient needs to give consent, you need to provide added value and give something back,” De Feu says. “That is what Lynxcare does. For example, we collaborate with the East-Limburg Hospital (ZOL), where we mine all the reports of the orthopedic unit and subsequently provide data insights for the physicians and the patients. We turn the information into improvements for outcome measurements,” he continues. “Such collaboration can also benefit pharmaceutical companies like Johnson & Johnson. The data gathered by hospitals about their products allows them to follow up, improving the quality of their products over time.”

“It is also of value when you talk about brain diseases,” De Feu adds. “These are hard to diagnose because you need a lot of real-world evidence data over a long period of time. So, we provide data back to the patient and the physicians, and we give our insights to the department that’s using that data to ameliorate results and procedures.”

Avatar photo
Janssen Pharmaceutica

At Janssen, we’re creating a future where disease is a thing of the past. We’re the Pharmaceutical Companies of Johnson & Johnson, working tirelessly to make that future a reality for patients everywhere by fighting sickness with science, improving access with ingenuity, and healing hopelessness with heart. We focus on areas of medicine where we can make the biggest difference: Cardiovascular & Metabolism, Immunology, Infectious Diseases & Vaccines, Neuroscience, Oncology, and Pulmonary Hypertension.

All posts

Subscribe to the BioVox newsletter