AI must overcome data challenges to reach healthcare potential
- Rapid digitization of health information in EHRs and other repositories is creating new opportunities for AI in healthcare, but challenges in data accessibility, privacy and security persist, according to a new ONC report.
- Frustration with legacy medical systems, the omnipresence of networked smart devices and consumer comfort with at-home services offered by Amazon and other tech vendors is driving interest in AI's potential.
- Smartphone, social and environmental data can all be potential sources to fuel AI's use in healthcare. However, the report concludes such data must be high quality and reliable. Otherwise, AI's promise will not be realized in healthcare.
AI is a hot healthcare topic but still needs to be translated into reality, especially in an industry as complex as healthcare.
During the second quarter of 2017, CB Insights counted 29 investment deals in the healthcare AI space — a record number — and predicted 2017 would set a six-year high.
Enthusiasm is expected to stay heated into 2018, with demand for tools that go beyond noting social determinants of health to using that data to inform patient care plans.
While investors will continue to fund wearables and biosensors, what grabs their attention are specific clinical use cases these technologies can support, Megan Zweig, director of research at Rock Health, told Healthcare Dive recently.
Tech giants including IBM Watson, Microsoft, Google and Apple are staking a claim in the space, too. Last month, Google launched Deep Variant, an open-source tool that uses AI to create a picture of a person’s genetic blueprint using sequencing data. The goal is to pinpoint specific genes or gene mutations that can help providers better manage disease states.
But challenges to widespread use of AI in health remain, as the ONC study shows. Among these are the acceptance of AI applications in clinical practice, difficulty leveraging divergent personal networked devices and AI solutions, access to quality training data on AI applications in health and gaps in data streams.
The report belies a large obstacle for rampant AI use. White noting the importance of high quality and reliable data, the industry has a data standards problem at the moment which needs to be ironed out.
Currently, different vendors and clinicians send unstructured data in medical records back and forth across EHR systems through continuity-of-care documents, which are format flexible. If the promise of AI relies on reliable data, standards will have to be well-defined to ensure the data are high quality.
On the bright side, the industry seems aware that healthcare is close to a breaking point at interoperability. The growing Internet of Things and consumerism in healthcare naturally demands a more networked, connected industry approach.
CMS Administrator Seema Verma in a town hall webcast on Wednesday with American Hospital Association CEO and President Rick Pollack said interoperability will be a topic of interest for the agency. She told listeners they will hear more from CMS in the future.
Follow Jeff Byers on Twitter