Why Artificial Intelligence Hype In Health Care Isn't A Bad Thing

Last updated: 10-09-2019

Read original article here

Why Artificial Intelligence Hype In Health Care Isn't A Bad Thing

In the year we celebrate America’s moon landing from 50 years ago, we are reminded that aiming high and thinking big have led to feats that once seemed impossible. But there were many exploded rockets on the launchpad before our astronauts set foot on the moon.

That’s why both the hype and the disappointment around artificial intelligence (AI) in health care don’t bother me. Although IBM Watson Health did not meet expectations for its “moonshot” goal to tame cancer with AI, there’s reason to be optimistic. Incremental progress for health care AI may not astonish, but it is nonetheless very promising.

As a technology CEO serving the health care industry with care management contracted for more than 50 million Americans, I talk with people regularly who express both excitement and trepidation about AI. Despite people’s fears that AI will dehumanize us, I believe it will enhance our humanity by freeing up resources that can be devoted to higher-level interactions. The AI of any given time will never outpace the incredible capacity of the human mind to perform insanely complex tasks. In fact, AI will spur us to think and act in ways that are more sophisticated once we are relieved of certain lower-complexity tasks.

The health care industry is undergoing a dramatic transformation to value-based care. Providers and health care organizations are assuming more financial risk for improving patient health outcomes, with stronger links between outcomes and reimbursement. This is a huge shift, and it’s creating enormous pressures, incentives, disincentives and unintended consequences in an already vast, fragmented and unpredictable industry that makes up nearly 18% of the U.S. economy. Information technology in this sector is attempting to leverage huge datasets in ways that create greater efficiency and better health. AI can wake up that data and bring it to life.

Since no clinician can keep up with the explosion in medical literature, predictive modeling to anticipate health risk is highly worthwhile. A simple example is the use of machine learning (ML) to send an electronic alert to a physician when a diabetes diagnosis or series of data points the ML engine stitches together from claims and other data to identify a diabetic case. The alert tells the clinician they should screen the patient for depression since depression can worsen diabetes. In a more sophisticated use of AI, an Alabama hospital experienced a 53% reduction in sepsis deaths by blending real-time electronic surveillance, algorithms and sensitive AI clinical decision support into a mobile application.

AI works well in risk stratification for population health by teasing out the most medically complex and fragile sliver of the population accounting for the highest costs to the system. AI is being used to plot graphics of patient populations and focus intensive resources on the most vulnerable. The American Medical Association recently gave a boost to just such a project, which aims to improve the identification and clinical management of diabetic patients with rapidly advancing kidney disease. The process uses algorithms blending data from electronic patient records with predictive biomarkers from blood. To further deepen insights, health plans and health systems are seeking ways to fold in social factors like geographical location and environmental variables to pinpoint interventions and using AI models to suggest appropriate actions.

Hospital readmissions within 30 days of discharge are one of the costliest failures of the health care system. In the past, the industry has used uniform discharge practices for entire patient populations, but there’s a new focus on targeting the riskiest patients for special interventions.

A current AI tool at the Children’s Hospital of Pittsburgh has been able to predict with 79% accuracy which patients are most at risk for a hospital readmission. The University of Maryland Medical System used ML in a 2016 study to generate risk scores for 16,000-plus patient discharges that better predicted readmission risk than standard methods. And there's now the potential to blend predictive insights with programs like the one at Johns Hopkins Medicine that sends nurses to the homes of the medically fragile who live by themselves. The right AI tools will identify vulnerabilities that are not as obvious.

Another factor that will keep us from an AI dystopian future in health care is that the best way to influence patient behavior toward better health choices is to work with human emotion, which technology cannot do. I’ve been reading Sapiens: A Brief History of Humankind, by Yuval Noah Harari, which is about the uniquely human ability to create stories through imagination. I believe tying this capability to motivational interviewing could change patient behavior. Expressing empathy and creating stories around the benefits of weight loss, for example, will be more effective if a patient sees healthier habits as a means to something they already cherish, like being able to play with their grandchildren.

I respect IBM Watson Health for its ambitious approach, but it's also important to note those making incremental progress elsewhere in the industry. Whether progress is a small step or a giant leap, there’s tremendous opportunity to harness AI for a healthier world.

In the year we celebrate America’s moon landing from 50 years ago, we are reminded that aiming high and thinking big have led to feats that once seemed impossible. But there were many exploded rockets on the launchpad before our astronauts set foot on the moon.

That’s why both the hype and the disappointment around artificial intelligence (AI) in health care don’t bother me. Although IBM Watson Health did not meet expectations for its “moonshot” goal to tame cancer with AI, there’s reason to be optimistic. Incremental progress for health care AI may not astonish, but it is nonetheless very promising.

As a technology CEO serving the health care industry with care management contracted for more than 50 million Americans, I talk with people regularly who express both excitement and trepidation about AI. Despite people’s fears that AI will dehumanize us, I believe it will enhance our humanity by freeing up resources that can be devoted to higher-level interactions. The AI of any given time will never outpace the incredible capacity of the human mind to perform insanely complex tasks. In fact, AI will spur us to think and act in ways that are more sophisticated once we are relieved of certain lower-complexity tasks.

The health care industry is undergoing a dramatic transformation to value-based care. Providers and health care organizations are assuming more financial risk for improving patient health outcomes, with stronger links between outcomes and reimbursement. This is a huge shift, and it’s creating enormous pressures, incentives, disincentives and unintended consequences in an already vast, fragmented and unpredictable industry that makes up nearly 18% of the U.S. economy. Information technology in this sector is attempting to leverage huge datasets in ways that create greater efficiency and better health. AI can wake up that data and bring it to life.

Since no clinician can keep up with the explosion in medical literature, predictive modeling to anticipate health risk is highly worthwhile. A simple example is the use of machine learning (ML) to send an electronic alert to a physician when a diabetes diagnosis or series of data points the ML engine stitches together from claims and other data to identify a diabetic case. The alert tells the clinician they should screen the patient for depression since depression can worsen diabetes. In a more sophisticated use of AI, an Alabama hospital experienced a 53% reduction in sepsis deaths by blending real-time electronic surveillance, algorithms and sensitive AI clinical decision support into a mobile application.

AI works well in risk stratification for population health by teasing out the most medically complex and fragile sliver of the population accounting for the highest costs to the system. AI is being used to plot graphics of patient populations and focus intensive resources on the most vulnerable. The American Medical Association recently gave a boost to just such a project, which aims to improve the identification and clinical management of diabetic patients with rapidly advancing kidney disease. The process uses algorithms blending data from electronic patient records with predictive biomarkers from blood. To further deepen insights, health plans and health systems are seeking ways to fold in social factors like geographical location and environmental variables to pinpoint interventions and using AI models to suggest appropriate actions.

Hospital readmissions within 30 days of discharge are one of the costliest failures of the health care system. In the past, the industry has used uniform discharge practices for entire patient populations, but there’s a new focus on targeting the riskiest patients for special interventions.

A current AI tool at the Children’s Hospital of Pittsburgh has been able to predict with 79% accuracy which patients are most at risk for a hospital readmission. The University of Maryland Medical System used ML in a 2016 study to generate risk scores for 16,000-plus patient discharges that better predicted readmission risk than standard methods. And there's now the potential to blend predictive insights with programs like the one at Johns Hopkins Medicine that sends nurses to the homes of the medically fragile who live by themselves. The right AI tools will identify vulnerabilities that are not as obvious.

Another factor that will keep us from an AI dystopian future in health care is that the best way to influence patient behavior toward better health choices is to work with human emotion, which technology cannot do. I’ve been reading Sapiens: A Brief History of Humankind, by Yuval Noah Harari, which is about the uniquely human ability to create stories through imagination. I believe tying this capability to motivational interviewing could change patient behavior. Expressing empathy and creating stories around the benefits of weight loss, for example, will be more effective if a patient sees healthier habits as a means to something they already cherish, like being able to play with their grandchildren.

I respect IBM Watson Health for its ambitious approach, but it's also important to note those making incremental progress elsewhere in the industry. Whether progress is a small step or a giant leap, there’s tremendous opportunity to harness AI for a healthier world.


Read the rest of this article here