Last week, Amazon announced the launch of Comprehend Medical, a tool to mine unstructured health data from text.
The press around the announcement jumped to the long-term implications of AI mining medical records, describing use cases where Amazon could support clinical decision-making (aiding in diagnosis, finding mis-diagnosed patients, etc.) and help recruit for clinical trials; it also discussed the promise of direct-to-patient applications that allow patients to view their medical records and perform tasks like auto-scheduling appointments based on their medical history.
What was actually launched
While it’s likely Amazon eventually does break into these areas, the brilliance of Amazon’s launch was its simplicity: all Amazon actually launched was a tool to classify unstructured health data as part of AWS.
As an example of the product, below is a screenshot of how Amazon classifies the text “Mr. Smith is a 63-year old gentleman with coronary artery disease and hypertension. Current Medications: taking a dose of lipitor 20 mg once daily”:
That’s all the product does: classifies text. It doesn’t cure cancer. It doesn’t replace doctors. And it doesn’t help patients take control of their care.
But underlying this functionality, there is a major need: the vast majority of health data is unstructured, sitting in PDFs, hand-written text, audio recordings, pathology reports, and doctor’s notes. The problem of unstructured data (and previously, the problem of non-digitized data) has been one of the biggest bottlenecks for the emergence of health analytics tools.
Why the strategy is brilliant for Amazon
Big Tech (and Silicon Valley broadly) has a large graveyard of failed healthtech launches. While there are many reasons for this, Amazon’s approach solves for two of the big causes of tech’s poor track-record:
- Avoiding Hype. Tech companies entering healthcare have repeatedly promised to solve problems that will take decades to solve, leading to skepticism by many and disappointment by early adopters (and often, shareholders and employees). Instead, Amazon has taken on a byte-sized problem; by looking at text classification problems, it is addressing a real need that it has the capacity to solve well. While not as exciting as a cure for cancer, this is an important step.
- Building an ecosystem. Many big tech companies have focused on their relationship with consumers as the core of their strategy (this has been most notable with repeated attempts at building patient-centric medical record systems by Google, Apple, and Microsoft). This has not been a successful strategy in healthtech: too much power resides with healthcare institutions (payers, providers, and pharma) to build a strategy that ignores their incentives. Amazon has addressed this by building a product aimed directly at institutions that hold health data, not at patients. With Comprehend Medical, not only is Amazon building an ecosystem of hospitals, insurers, and pharma companies who send Amazon their health data, they are building a developer ecosystem; the thousands of health analytics companies building applications on top of health data all struggle with how to manage unstructured text.
Healthcare is an underserved market by big tech, and all of the major players are beginning to pay attention and more actively invest in their healthcare initiatives. Expect Amazon’s launch to be followed by several others trying to match Amazon’s approach, unleashing the most meaningful entry of tech into healthcare to date with a series of incremental, enterprise-focused products. These incremental steps are what are needed first to unlock the bigger visions of technology revolutionizing healthcare.
.svg)
