The potential for both artificial intelligence and robotics in healthcare is considerable, and artificial intelligence and robotics are increasingly a part of our healthcare eco-system. An example is with Apixio, which focuses on healthcare data science and the implementation of artificial intelligence in healthcare. The company has been using artificial intelligence and augmented intelligence to solve the overwhelming data problem facing the industry today.
To discover more about the ‘state of the industry’ and the work that Apixio has been undertaking, Digital Journal spoke with Darren Schulte, Apixio’s CEO.
Digital Journal: What impact is technology having on healthcare?
Darren Schulte: Technology has had a tremendous impact to extend and enrich life. From biomedical advances, such as novel drug therapies for autoimmune disorders or different cancers, to diagnostic four-dimensional, non-invasive imaging to robotic surgery, innovations in healthcare are having a significant impact on both patients and healthcare providers. And, the biological revolution initiated by CRISPR-Cas9 genome editing technology will have a profound effect on ways in which we treat disease.
We are at the beginning of realizing the value from information-based technology advances. Now that we have achieved a high adoption of electronic records by care providers, and there are more connected devices and wearables, we can start to focus on making sense of this generated health information to improve access, affordability, and outcomes.
DJ: What’s driving this, in terms of the medical profession or medical technology companies?
Schulte: There appears to be a strong pipeline of technology development in this country, which includes lab research, translational research in academic medicine, and the life sciences industry. With greater amounts of private capital available than ever before, coupled with less federal government support, there is more innovation originating from the industry. Private and public companies have a great degree of interest in applying basic science breakthroughs from academia to create novel therapies and diagnostics. With respect to digital health, the primary drivers of innovation are coming from the industry, fueled primarily by large amounts of investment capital and perceived market inefficiencies.
DJ: How about patient expectations?
Schulte: For patients who suffer from diseases without good treatment options, there are high expectations that science can develop breakthroughs and that the FDA will permit more experimentation. For others, expectations remain that scientists will continue to push the frontiers of what is possible to cure disease.
DJ: How important is artificial intelligence becoming for healthcare?
Schulte:It is too soon to tell. We are at or near the top of the hype cycle now with AI. There are areas in which use of AI-based technology can help alleviate physician workflow challenges, improve diagnosis and treatment decisions, and augment key revenue cycles and care management workflows. There is real promise to accelerate phases of drug discovery and clinical evaluation. But we are just at the beginning of using AI. As was seen with the internet, it will take several decades to realize the benefits of breakthrough technologies such as AI.
DJ: What the advantages?
Schulte:AI technologies can address the variation across healthcare in terms of practice, patient population, documentation, electronic systems, etc., if in fact they are exposed to a large corpus of labeled training data sets. And these systems can be trained using many more cases than any one physician can experience throughout their career lifespan.
DJ: Are there any notable case studies?
Schulte:There are now FDA approved wearables available to the public, which can detect abnormal heart rhythms as well as, or better than, most cardiologists. These devices can be potentially life-saving.
There are other academic or industry studies describing computers able to aid in the detection of abnormalities in X-rays, retinal eye scans, or pathology slides. But these software programs are not yet in wide use. As well, there are industry claims that AI-powered chatbots can diagnose disease as well as internists. But these claims have not been put to clinical study rigor and large, varied populations.
DJ: What are disadvantages?
Schulte:An over-reliance on these systems might degrade physician skills over time. An analogy might be the pilot who flies a plane with sophisticated computers that can fly the plane on auto-pilot. What happens when something goes awry and the pilot needs to take over control of the aircraft to make an emergency landing? While AI could potentially help alleviate the workload and administrative burdens of physician practice, it is important for physician diagnostic and treatment skills to be maintained.
DJ: Are data analytics also becoming important?
Schulte:Data analytics are vital to the new world of healthcare practice. As treatments become more varied and expensive, as with biologic drugs and genetic engineering, it is critical that they are applied for the right reasons with both outcomes and costs considered. Analytics which can personalize treatment options using a “patients like me” approach acquired through the study of real world evidence are essential.
In addition, there is promise that with data analytics, researchers can conduct a wide variety of experiments “in silico” using very large datasets. Essentially, we can learn from national variation in medical practice to find novel uses for existing treatments; determine what works for what patient phenotype; and predict outcomes so that we can intervene to alter individual trajectory.
DJ: Where do you see health technology heading over the next five years?
Schulte:There will be a maturation in practice uses of new genetic technology (e.g. CRISPR) and analytic technology (e.g. AI). That said, I suspect that data analytics outside of image or voice recognition will experience a slower maturation curve because of continued health record fragmentation – although “quantified self” data will be more widely available for use.
In addition, unless payment methods evolve faster in either the direction of value-based contracts or greater patient responsibility, adoption of AI-based technologies will be slow. It is only with meaningful data sharing and “interoperability” that there will be a large corpus of training data available for faster progression of AI technology performance.
