reprint

Two-thirds of attendees polled at a recent innovation summit by The Economist agree that healthcare is the sector that will benefit the most from artificial intelligence. However, questions loom on exactly how it will help the industry, and perhaps more importantly, if there is the possibility of it accomplishing what it promises.

The latter concern was recently experienced by IBM when its Watson cognitive recognition system was used as part of the “moon shot” project launched by the MD Anderson Cancer Center to diagnose and recommend treatment plans for certain forms of cancer. The project costs spiraled past $62 million while the system had yet to be used on an actual patient. The extremely bold and ambitious initiative failed to deliver.

At the same time, the IT group at the Cancer Center was experimenting with more humble and simple applications of AI. These included making hotel and restaurant recommendations for patients’ families, determining which patients needed help paying bills and addressing staff IT problems. These smaller efforts paid off in time savings, patient satisfaction and financial performance. Perhaps there is a lesson here.

It is a lesson that many seem to be taking to heart. While there is widespread acceptance of the transformative power of AI, a recent HBR study of 152 projects revealed that highly ambitious moon shots are less likely to be as successful than “low-hanging fruit” projects that enhance business processes.

These practical applications of AI in healthcare, are providing process optimization in many areas. For example, automated call centers or health centers using AI-based technologies, such as chatbots (virtual agents that can provide around-the-clock patient support) are reducing cost of care delivery and healthcare operational costs.

And, according to Gartner, virtual care can reduce costly emergency room visits in rural areas. By 2023, U.S. emergency department visits will be reduced by 20 million because of enrollment of chronically ill patients in AI-enhanced virtual care facilities. It is also a reason Gartner is making AI-enhanced virtual care one of its top 10 predictions for 2019 and beyond.

Health systems are making other aggressive deployments of AI. These deployments are delivering tangible improvements, such as estimating possible health outcomes to correct the course of current care plans and provide improvements to overall care pathways.

AI is on the path to helping to identify disease markers and comparing the data to the patient’s medical history. The Internet of Things (IoT) can add data to the mix through a whole new scope of offerings in wearable technologies that capture patient health data in real time. These applications can deliver significant new insights into health risk and enable very customized treatment profiles for patients—from lifestyle coaching to targeted medication therapies—or predictive indications for stronger interventions such as surgery.

Health systems, however, are struggling to make strides in AI initiatives, largely because of funding issues and the need to allocate budget to other higher priority areas.

In the near future, a more promising approach will be found with product companies which have the funds and marketing incentives to develop beneficial AI applications, or with clinical research projects that are targeting AI to tackle specific diseases.

While there are many new exciting and pragmatic uses of AI already underway in healthcare, data will continue to be the biggest challenge. As data continue to be the essential fuel for the effective functioning of any AI project, healthcare institutions should invest significantly in improving and maintaining the quality of their data.

When asked, an AI-based security start-up flatly declared that improvements in data sets drove all major advances in AI. The CTO of the company viewed collecting, classifying and labeling data sets as “the grunt work” that is fundamentally necessary to make algorithms work.

Another area where healthcare may struggle is in addressing the scalability and dissemination of the knowledge needed to develop and drive AI. AI skill sets are often locked up in the minds of those that develop the technology. Gartner predicts that „80 percent of AI projects will remain ‚alchemy,’ since talent working on AI development does not scale well through the rest of an organization.”

Finally, there is the issue of simple adoption. Other industries have been creating omni-channel data capture, aggregation and analysis capabilities for some time, but healthcare still has data locked up behind proprietary firewalls. And, after a decision is made to move forward on AI, healthcare organizations must develop intervention models and care plans to make use of AI-driven insight. The challenge will be to align these models with what is realistic regarding the restrictions currently throttling AI.

In 2019, while AI will begin seeing major progress in healthcare, it is not going to be for moon-shot projects or as a quick remedy to all that ails the industry. Pragmatic AI, being applied through chatbots in support centers or doctor’s offices; or predictive analytics to help visualize possible outcomes, will gradually make significant progress. With realistic expectations, AI will slowly build momentum as the healthcare champion it is ultimately destined to become.


reprint