At AWS re:Invent, Amazon Web Services announced the launch of their new machine learning solution, Amazon Comprehend Medical. This new set of natural language processing APIs is powered by Amazon Comprehend, which Amazon recently shared as being supported within the Business Associates Agreement (BAA). Comprehend can extract data insights from unstructured data using natural language processing to find, extract and correlate insights around entities, key phrases, sentiment, language, syntax and topic analysis. What’s most exciting to me about Amazon Comprehend Medical is this solution is built on Comprehend with a series of data models trained specifically around PHI/PII for healthcare. This is big news for healthcare innovation.
This deep-learning natural language processing (NLP) service Amazon Comprehend Medical has many use cases in healthcare, but of primary interest is its ability to take unstructured data from applications including electronic medical records, and correlate sensitive data such as PHI/PII into meaningful insights. Initially Amazon Comprehend Medical will likely be used to find misdiagnosis of patients stemming from incorrect ICD-10 codes (the medical classification list of diseases and diagnoses kept by the World Health Organization) that can lead to sickness or death because of errors in medical records. In addition to reducing errors, Amazon Comprehend Medical will help improve patient care by enabling providers to have relevant data at their fingertips, quickly, and at a fraction of the cost it would take to do this without machine learning.
While Amazon Comprehend Medical will be used throughout the healthcare vertical, the real winner here is the patient. The patient’s affiliated researchers and care teams can correlate common data from multiple visits, or prescription history, without having to read through all of the unstructured data and notes, much of which is never read during a typical visit with your provider. The NLP does the heavy lifting, and with an error rate that is anticipated to drop significantly from the percentage humans have been performing at. Additionally, Amazon Comprehend Medical’s ability to use predictive modeling means it can display probabilities of prescription success based on its ability to collect data and draw insights from other patients with common symptoms and data features.
By being able to extract and organize the volumes of unstructured data living in doctors’ notes and EMRs, Amazon Comprehend Medical can provide percentages and probabilities that inform and improve diagnosis.
ClearDATA is poised and ready to build on Amazon Comprehend Medical to improve patient outcomes as never before. The plethora of sensitive data from EMRs and other patient apps that typically live inside AWS S3, Glacier and other data services are ready and waiting for us to innovate with Amazon Comprehend Medical. AWS can index and search at scale, and with ClearDATA, maintain compliance and privacy providing healthcare organizations better visibility into what data they have, and where it’s flowing within an ecosystem – visibility that anyone working with containers or Kubernetes will be interested in.
This article was originally published on ClearDATA and is republished here with permission.
Matt recently talked to Aaron Friedman from AWS Partner Network on his show CTO Talk. Take a listen.