Data-driven diagnosis: Barriers hinder AI-powered care models and MHealth apps

June 27, 2017

In August 2016, doctors in Japan reported the first ever instance in their country of AI-powered medicine saving a woman’s life. After doctors discovered their treatment was ineffective on a leukemia patient, they turned to IBM’s Watson, a cloud-based AI-powered software system that has learned from tens of millions of oncology papers and an enormous amount of data. The AI-powered software analyzed the test results and genetic data of the patient to look for patterns consistent with various forms of leukemia. Doctors were surprised to discover that Watson determined the original diagnosis was incorrect. The patient turned out to be suffering from a rare form of leukemia, which required a different treatment regime. Watson identified the correct diagnosis and recommended an effective course of treatment in under ten minutes, a tiny fraction of the time doctors would normally take to sift through volumes of research literature and genetic data to arrive at such a conclusion.

Although data-driven applications such as these may have profound implications for the future of healthcare, a litany of obstacles preclude these technologies from going to market. Stringent regulations, fragmented technical standards and the generally painstakingly slow nature of changing institutional norms, are some of the more prominent obstacles cornering innovative applications of information technology in healthcare into the experimental niche they occupy today. Thanks to accelerating innovation in communications technology, devices can be linked together in ways that were unthinkable just a decade ago. With Internet of Things (IoT) technology becoming common today’s medical devices, the amount of medical data being produced, aggregated and analyzed is growing exponentially. These innovative systems promise more efficient and accurate care than that of today. But still, substantial progress must be made before these products can successfully go to market.

One reason why promising applications are not yet widely known is the long regulatory approval process. The regulatory bar is set much higher for chronic care products than for wellness products, such as the Apple Watch, making accuracy and reliability of paramount importance. While consumer-facing wellness products have proliferated in recent years thanks to light regulation, products which allow for efficient and high quality self-care of chronic ailments through at-home analysis have yet to achieve mass market success in the U.S., partly due to the longer approval process required by government regulators. Within the emerging MHealth space, there exists a careful balance of cooperation and competition between medical device manufacturers and information technology firms, both of which rely on each other to develop products which are both technically innovative and that can pass the stringent regulatory hurdles by demonstrating sufficient accuracy and reliability. Large device manufacturers with experience running multi-year trials required under current regulations are well positioned to work within the system and technology firms certainly realize the benefits of working that process with an experienced partner.

Another concern for product developers is user uptake. Making sure products will be used by patients is critical concern for the developers of MHealth systems, since relying on them to provide care for chronic ailments, such as diabetes or heart disease, only makes sense if self-care models are just as effective as traditional ones. This puts the spotlight on the patients themselves, and more importantly, their ability to use the systems properly. Creating a user-friendly, foolproof system presents design challenges and inaccurate data or a clunky user interface could do more harm than good for patients. Accuracy, reliability and user uptake are the critical concerns for the developers of these home-based systems, since relying on them to provide care for chronic ailments, such as diabetes or heart disease, only makes sense if a self-care model can generate equally beneficial outcomes to those of traditional care models. Creating a user-friendly, foolproof system presents design challenges and the regulatory bar is set much higher in the chronic care space than for wellness products. Inaccurate data or clunky user interfaces could do more harm than good. Add that to the fact that many consumers today expect top-notch user experience from their apps and devices, and the challenge only gets tougher.

As information technology becomes further entrenched within healthcare delivery models, patient outcomes will only be as good as the information that these systems can aggregate and analyze. Although medical devices are churning out more data than ever before, the information bottleneck in healthcare today manifests itself at the aggregation stage. Complex systems must collate various information from different devices and medical records to provide useful insight to patients and their caregivers. Different standards and protocols among devices often hamper the aggregation and analysis capabilities of these systems today. Nowhere is this truer than in the healthcare space, where the transition to electronic health records (EHR) is occurring at a snail’s pace. On the bright side, experimentation in data-driven healthcare is quite active today, and partnerships between medical device manufacturers, healthcare providers and technology companies promise to accelerate the formation of industry standards in the same way cooperation has helped develop industry standards in the information technology space itself. Powerful new technologies such as artificial intelligence will not realize their full potential until the aggregation problem is solved.

Using data effectively in the healthcare space with new technologies, including AI and others, can deliver better healthcare outcomes to more people, thanks to a potential reduction in the need for human labor needed to make decisions and provide high quality care. And it is true that there is potential for many efficiencies in healthcare delivery models from these technologies, but change is coming more gradually in healthcare.

This is the second article in our series on the market dynamics of data-driven healthcare products in the United States. You can read more about the proliferation of MHealth delivery using advanced information technology here.

oter(); ?>