Amazon Starts Selling Software to Mine Patient Health Records

quote:

---

Amazon.com Inc.is starting to sell software to mine patient medical records for information that doctors and hospitals could use to improve treatment and cut costs, the latest move by a big technology company into the health care industry.

The software can read digitized patient records and other clinical notes, analyze them and pluck out key data points, Amazon says. The company is expected to announce the launch Tuesday.



Amazon Web Services, the company’s cloud-computing division, has been selling such text-analysis software to companies outside medicine for use in areas such as travel booking, customer support and supply-chain management. The technology’s health-care application is the newest effort by Amazon to tap into the lucrative market.


This year Amazon paid $1 billion for an online pharmacy called PillPack Inc. to acquire the capability to ship prescription drugs. The retailer also has been trying to increase its sales of medical supplies by working with hospitals.


In addition, Amazon is eyeing greater sales of medical supplies through an app, embedded in electronic medical records, that doctors can use to send links to products that patients would buy, according to people who developed the app and doctors who have used it.

Other tech giants have also been trying to elbow into health care. The Wall Street Journal reported that Apple Inc. is talking with the Department of Veterans Affairs about software to allow veterans to transfer their health records to iPhones, and that Alphabet Inc.’s Google had hired a prominent hospital-system chief to oversee the company’s various health-care efforts.

The $3.2 trillion health-care market is a natural target for big-tech companies seeking new sources of growth, especially as patient medical records get digitized and pressure rises to provide better care of patients at lower cost.

Yet Silicon Valley’s forays have hit technical and regulatory roadblocks. Verily Life Sciences, Alphabet’s medical-device business, said this month it was putting on hold a collaboration to develop contact lenses that could measure diabetics’ blood-sugar levels because of insufficiently consistent measurements.


Analyzing patient medical records, an obvious application of the natural-language processing capabilities developed by tech companies, has also confronted technical hurdles.


Algorithms currently in use have encountered trouble identifying key data points due to misspellings, abbreviations and doctors’ idiosyncratic descriptions, according to Dr. Julia Adler-Milstein, director of the Center for Clinical Informatics and Improvement Research at the University of California, San Francisco.

Amazon officials say the company’s software developers trained the system using a process known as deep learning to recognize all the ways a doctor might record notes.

“We’re able to completely, automatically look inside medical language and identify patient details,” including diagnoses, treatments, dosage and strengths, “with incredibly high accuracy,” said Matt Wood, general manager of artificial intelligence at Amazon Web Services.

During testing, the software performed on par or better than other published efforts, and can extract data on patients’ diseases, prescriptions, lab orders and procedures, said Taha Kass-Hout, a senior leader with Amazon’s health-care and artificial intelligence efforts.

Demand for such health-data analytics capability is strong as hospitals and doctors offices have sought data-mining services to take advantage of the shift to electronic medical records.

The market for storing and analyzing health information is worth more than $7 billion, according to research firm Grand View Research. International Business Machines Corp.’s Watson Health and UnitedHealth Group Inc.’s Optum already compete.


The software will help the Fred Hutchinson Cancer Research Center, which assisted Amazon in testing and training its algorithm, to identify patients eligible for studies of experimental drugs, according to Matthew Trunnell, the center’s chief information officer.

It could also be an economic benefit to the Seattle-based center, Mr. Trunnell said. The center has employed about 60 people to scan and pull essential data from records on about 500,000 cancer patients. As automation does more of the work, some employees could do other tasks.

Users upload health records to Amazon’s cloud service, where they can run the text-processing software. Amazon’s algorithms analyze text for specific types of data and return the results in an organized format, similar to a spreadsheet.

Amazon Web Services won’t see the data

---