Amazon Comprehend Medical Makes Sense of Scattered Healthcare Data

Amazon on Tuesday announced Amazon Comprehend Medical, a language processing service that lets users gather information — such as a patient’s medical condition and medication dosage, strength and frequency — from sources including doctors’ notes, clinical trial reports, hospital admission notes and patient health records.

Most health and patient data — such as medical notes, prescriptions, audio interview transcripts, and pathology and radiology reports — currently is stored as unstructured medical text. Identifying this information either requires skilled medical coders to handle data entry, or teams of developers to write custom code and rules to extract it automatically.

Blessing and Curse

Healthcare providers and insurers currently have to write and maintain a set of customized medical coding classification rules for natural language software. A change to a single classification code name can have a ripple effect on dozens of hard-coded rules, all of which then must be updated, or data will be missed or classified incorrectly.

Amazon Comprehend Medical lets users create models that reliably understand this medical information, vastly improving medical coding.

“The amount of regulation and regulatory requirements in medicine is both a blessing and a curse,” noted Ray Wang, principal analyst at Constellation Research.

“The curse is managing the disparate sets of information that are not always complete or entered in the right way,” he told TechNewsWorld. “The blessing is, this information can lead to a lot of insights for organizations to plan the next best action. This includes identifying medical coding errors, potential harmful drug interactions, epidemiological patterns, opportunities to prevent new illnesses, and billing malpractice and quality assurance.”

Data interoperability long has been a problem for the healthcare industry. Amazon, Alphabet, IBM, Microsoft, Oracle and Salesforce this summer announced support for the Fast Healthcare Interoperability Foundation (FHIR), which aims to create a common set of standards for the exchange of healthcare information .

Automating the Process

Amazon Comprehend Medical lets developers identify the key common types of medical information automatically, with high accuracy, and without having to write lots of custom rules. It can identify medical conditions, anatomic terms, medications, and details of medical tests, treatments and procedures.

“The ability to use Comprehend to extract and categorize unstructured medical data is significant, because it also enables personal information to be removed,” said Rebecca Wettemann, VP of research at Nucleus Research.

“Researchers and physicians can share and analyze data sets that were simply not practical to analyze as unstructured patient-specific notes,” she told TechNewsWorld. “More data is better when understanding diagnoses, trying new drugs or treatments, or evaluating existing treatments.”

Amazon Comprehend Medical uses advanced machine learning models. It offers two APIs developers can integrate into existing workflows and applications with a few lines of code. The APIs are Medical Named Entity and Relationship Extraction (NERe) and Protected Health Information Data Extraction and Identification (PHId).

Partners include the following organizations:

  • The Fred Hutchinson Cancer Research Center
  • Roche Diagnostics
  • Pricewaterhouse Coopers
  • Deloitte (ConvergeHEALTH)

Data Privacy and Security

Amazon Comprehend Medical is HIPAA-eligible. It quickly can identify protected health information (PHI) such as name, age and medical record number, so it can be used to create applications that securely process, maintain and transmit PHI, according to Amazon. It adheres to General Data Protection Regulation (GDPR) standards.

The service is covered under Amazon Web Services’ HIPAA eligibility and BAA.

Developers can implement data privacy and security solutions by extracting and then identifying relevant patient identifiers as described in HIPAA’s Safe Harbor method of de-identification. Further, the service does not store or save any customer data.

However, data security is a primary concern, noted Michael Jude, program manager at Stratecast/Frost & Sullivan.

“Data in motion is far less secure than data at rest,” he told TechNewsWorld. “In order to use these Amazon capabiities, data must be moved. Amazon will need to take pains to ensure the security of any patient data they have access to.”

What the Service Will Cost

Amazon Comprehend Medical charges are levied per block of 100 characters of analyzed text.

NERe API users pay one (US) cent per block and PHId API users pay 14 cents per block. Users pay only for what they use, and there are no minimum fees or upfront commitments.

Amazon Comprehend Medical offers a free tier for 25K units of text, or 25,000 characters, for the first three months when customers use both APIs.

“Anything that improves the delivery of healthcare will decrease costs and improve outcomes,” Jude said. “This [service] could help do both.”

Richard Adhikari

Richard Adhikari has been an ECT News Network reporter since 2008. His areas of focus include cybersecurity, mobile technologies, CRM, databases, software development, mainframe and mid-range computing, and application development. He has written and edited for numerous publications, including Information Week and Computerworld. He is the author of two books on client/server technology. Email Richard.

Leave a Comment

Please sign in to post or reply to a comment. New users create a free account.

More by Richard Adhikari
More in Health

Technewsworld Channels