Innovara Digest: How the Marriage of Data Analytics and AI Will Transform Healthcare
As healthcare providers struggle to meet changing patient demands while keeping costs down, the usefulness of artificial intelligence (AI) in healthcare is becoming ever more apparent. Personal health assistants such as Your.MD and Babylon Health, that base their diagnoses and advice on aggregated data, have made their way into the consumer market already. In the near future, more robust apps that diagnose patients in real time using aggregated data and customize treatment plans based on their lifestyle, medical history, and family history will be widely available. In this week’s digest, we will explore the possibilities and limits of machine learning in data analytics.
Analyzing data is easy for AI.
According to Eyeforpharma’s “Thinking Machines: From Magic to Normal”, the amount of medical data being created is expected to reach 25,000 petabytes (that is 25.6 million terabytes) by 2020. It would be extremely inefficient for humans to fully undertake the task of analysis—on average, a doctor reads 6 medical research papers in a month. IBM Watson can read 500,000 papers in 15 seconds. “AI could save healthcare professionals hundreds or even thousands of hours of analysis each year, enabling them to reach more patients,” the author states. “Such developments could also bring down the overall costs of treatment…remov[ing] localized healthcare poverty and offer[ing] places without good healthcare professionals access to the help, analysis, and treatment they require.” Given the global healthcare workforce shortage of over 7 million, AI will bring the relief that the healthcare world needs.
Well-trained AI algorithms are very precise in early diagnosis.
The ability of AI algorithms to analyze large amounts of data in a short period of time is impressive, but their real usefulness lies in their ability to discover patterns in these data and make recommendations. In “How Artificial Intelligence is Revolutionizing Healthcare”, TheNextWeb’s Ben Dickson highlights some of the ways that AI is revolutionizing early diagnosis of common diseases and disorders:
- Stanford researchers trained an AI algorithm to identify skin cancer by sight, using 130,000 images of blemishes. Its efficiency is said to “rival that of professional doctors.”
- DeepMind is developing an algorithm to detect eye conditions at early stages. This “might eventually be able to prevent 98% of most severe visual loss.”
- Morpheo, which will be available in the fall, automates the identification of sleep patterns to assist doctors in diagnosing sleep disorders.
Machines alone aren’t perfect.
Humans are both detrimental and necessary to the success of AI in healthcare. According to Eyeforpharma’s “When the Machines Take Over,” standardization of and access to data have been major struggles in learning algorithms. “If we could freely link data across data silos, we could develop deep insights much faster, but…we have strict data governance and privacy rules to follow,” says Robin Murray, Director RWI Technology at Quintiles IMS. “Work is being done on common data models and coding schemes but it is a long, arduous process.”
Despite the roadblocks that humans pose in the advancement of AI, they are necessary to its success. Researchers from the Radiological Society of North America claim that a combined effort between humans and AI leads to higher accuracy in the detection of tuberculosis in x-rays. Pairing AlexNet and GoogLeNet, two high-performing AIs, netted 96% accuracy. Adding a human to the model resulted in 99% accuracy. In the world of infectious diseases – especially with the major challenge of highly infectious, resistant strains of TB – 3% more accuracy is hugely important.
…And then there are cybersecurity concerns.
“Healthcare is ground zero for cyberattacks,” says Russell Branzell, President and CEO the College of Healthcare Information Management Executives (CHIME). Merck & Co. just found out the hard way—the drugmaker confirmed it has been affected by a cyberattack. The value of medical data ($7-10 per medical identity, vs. $1-3 for a financial identity), coupled with a tendency to overlook the importance of cybersecurity, makes the entire healthcare industry ripe for these attacks.
Branzell estimates that nearly 15% of all documented attacks are carried out by insiders. Just ask St. Joseph Health, which had to pay $15 million in a settlement to patients affected by a 2012 data breach, and $21.5 million on upgrading security systems and credit monitoring for affected patients.
As Apple, Google, IBM, other tech giants, and a flood of startups form collaborations with healthcare and research institutions, AI’s entrance into healthcare is still considered to be lagging. Accenture reports that clinical AI applications will save the industry $150 billion over the next 10 years. According to mobihealthnews’s “Where We Are Now and What’s Next,” about half of US hospitals intend to adopt AI in the next 5 years. In the world of data analytics and AI, that is an eternity. The pharmaceutical industry may be lagging even further behind in marrying data analytics with the adoption of AI, in large part due to risk aversion. “I work in an industry where it takes 12 years to launch a product,” says Judy Sewards, Vice President of Digital Strategy and Data Innovation at Pfizer. “In our industry, you need to be 100 percent. Error is someone’s life.”
Learn to leverage advances in technology with Digital Healthcare Marketing™, available from the Innovara Marketing Academy:
The Innovara Digest is a weekly compilation of timely analyses and opinion pieces of interest to marketing, medical affairs, sales and other managers from pharmaceuticals, medical devices, biotech, healthcare IT, and life sciences industries.