By Jason Roys
In 2016, Geoffrey Hinton, a Turing Award winner, predicted the future of artificial intelligence (AI) in healthcare. “We should stop training radiologists now,” he said. “It's just completely obvious that within five years, deep learning is going to do better than radiologists.” There are more radiologists today than there were then.
The healthcare industry’s integration of artificial intelligence (AI) technologies lags other sectors, according to research published in 2022 by the Brookings Institution. Technological, bureaucratic, and regulatory obstacles have hampered integration, the study said.
Meanwhile, investment in digital health startups dropped sharply in 2022 after a bubble in 2021, because rising interest rates made investments riskier. The fourth quarter of 2022 marked the lowest quarterly digital health funding total — $3.4 billion — in the past five years, research and data company CB Insights reported.
At SDV INTERNATIONAL, we’re well aware of these bumps in the road on the way to a digital health future, but we remain bullish on the potential of AI in healthcare to not only save lives but also save the healthcare industry upwards of $350 billion annually if adopted more widely.
The AI in healthcare market is projected to grow from $14.6 billion this year to $102.7 billion by 2028, a compound annual growth rate of 47.6%. Venture capital is sticking with AI in 2023, too. Dr. Justin Norden, a partner with GSR Ventures, said in an interview, "People who are focused on health tech … don't appear to be running away from the sector.”
So, while the bold predictions Hinton and others made have not (yet) become reality, AI technology is alive and well in the healthcare industry. More is in the pipeline, too, focused on clinical practice, medical imaging, clinical trials and administrative efficiency, such as claims processing. Machine learning and data science are becoming key components in developing actionable insights. Innovators still need to invest significant time and effort to create safe, ethical and patient-centered use-cases for the technology.
However, if developed correctly, AI can change patient care for the better.
In this guide, we'll assess the rise of artificial intelligence in healthcare, examine challenges to wider adoption, and look to the future — hopefully with more accuracy than Hinton demonstrated!
HISTORICAL DEVELOPMENT OF AI IN HEALTHCARE
In 1964, Joseph Weizenbaum of MIT created ELIZA (an allusion to Eliza Doolittle, perhaps?), an early natural language processing program. Intended to explore communication between humans and machines, ELIZA simulated conversation by using a pattern-matching and substitution methodology that was based on psychotherapy techniques.
Weizenbaum was surprised when people attributed human-like feelings and understanding to his program.
Artificial intelligence underwent a slow incubation period. In 1978, scientists at Rutgers University developed CASNET, or causal-associational network, as a glaucoma consultation program. The University of Massachusetts in 1986 released DXplain, a rule-based clinical decision support system (CDSS). It could provide a differential diagnosis based on symptoms entered, and it is still used in clinical education.
As computer and network capabilities became more powerful and sophisticated over the last few decades, AI in healthcare began to take hold as an attractive and worthy business and societal goal. IBM tried to ride the wave with its Watson AI technology. Far from being a triumph, though, it became a cautionary tale.
After showcasing Watson on “Jeopardy!” — where it beat Ken Jennings, the winningest contestant ever — IBM decided its AI moonshot would be finding a cure for cancer. However, a series of issues surrounding data collection and interoperability hindered Watson’s ability to revolutionize healthcare. IBM executives later acknowledged that the company had been too optimistic about Watson’s potential. Francisco Partners acquired Watson Health in 2022 and rechristened it Merative.
So, as we saw earlier, innovators and investors like Francisco Partners have not just thrown up their hands. Real-world performance is improving every day as we explore the potential of machine learning technologies and as new clinical and healthcare research use-cases are identified. But we all need to realize that AI in healthcare is not a magic bullet or a moonshot, but a movement.
GAINS FOR HEALTHCARE FROM AI
Now that we’ve set the stage with realistic expectations, let’s look at some areas where AI tools and AI algorithms are making a difference in our healthcare system.
IMPROVED DIAGNOSIS AND TREATMENT
Accurate diagnoses are fundamental to healthcare. In the U.S., approximately 5% of outpatients receive an incorrect diagnosis, with errors being particularly common for serious medical conditions, carrying the risk of serious patient harm.
In recent years, AI and machine learning have emerged as powerful tools for assisting patient diagnosis. Here are a couple of examples:
Scientists at Babylon, a global tech company focusing on digital health, used machine learning to develop new AI symptom checkers that they believe could help reduce diagnostic mistakes in primary care. Their new approach overcomes the limitations of rule-based systems in its machine learning.
Writing in Nature, Dr. Jonathan G. Richens, Babylon scientist, said: “We took an AI with a powerful algorithm and gave it the ability to imagine alternate realities and consider 'would this symptom be present if it was a different disease?’ This allows the AI to tease apart the potential causes of a patient’s illness and score more highly than over 70% of the doctors on written test cases.” Babylon is working to improve health equity in underserved regions.
Early detection of cognitive disorders like Alzheimer’s disease, schizophrenia, and autism spectrum disorder requires analysis of large amounts of data on brain patterns. In 2022, researchers at Georgia State University’s TReNDS Center created an AI-based model that can analyze fMRI data and potentially detect these disorders. (Functional magnetic resonance imaging, fMRI, is a type of brain imaging that shows brain activity and blood flow while patients think or perform activities.)
The AI models were exposed to fMRI data from 10,000 people without known neurological disorders. Then, the AIs were given data from 1,200 people diagnosed with autism spectrum disorder, schizophrenia, and Alzheimer’s. The AI model analyzed the differences between these data sets to determine which patterns or changes were most likely linked to the onset of each.
BETTER PATIENT OUTCOMES
Researchers are finding ways to use predictive analytics to determine clinical risks and courses of treatment.
U.S. Military Treatment Facilities turned to SDV INTERNATIONAL for clinical decision support enhancements to ensure that fetal surgery procedures could be carried out without error. For patients facing a complex fetal condition, procedures offer varying levels of care, including fetal blood sampling, first-trimester fetal echocardiology, fetoscopic laser surgery, and spina bifida repair.
Women giving birth face clinical risks that are subjective and often unpredictable. To mitigate risks, researchers created a tool to leverage patient data collected at the start of labor, such as baseline characteristics, recent clinical assessments and cumulative labor progress from admission. After testing the algorithm on data from thousands of deliveries, researchers concluded that the AI risk prediction model was applicable and accurate.
The more complex a patient’s condition, the more difficult it is to determine the course of treatment. MIT researchers developed a deep-learning technique called G-net that simulates counterfactual, time-varying, and dynamic treatment strategies, allowing doctors to choose the best course of action.
“Our ultimate goal is to develop a machine learning technique that would allow doctors to explore various ‘what if’ scenarios and treatment options,” said Li-wei Lehman of the MIT Institute for Medical Engineering and Science.
ENHANCED EFFICIENCY AND PRODUCTIVITY
Healthcare organizations and medical professionals are often burdened with repetitive administrative tasks. Meanwhile, hospitals must find ways to optimize the use of their expensive and constrained resources. AI to the rescue!
LeanTaaS, a Silicon Valley software company, developed software that combines lean principles, predictive analytics, and machine learning to help optimize how health care systems use valuable resources like infusion chairs, operating rooms, and inpatient beds. The company's long-term vision is to become the “air traffic control center” for a health system, executives said. Its software is in use in more than 500 hospitals.
SDV INTERNATIONAL is making innovations in AI-enabled automated solutions to execute the repetitive and manual tasks that healthcare workers do every day, like prior authorizations and benefits verifications, faster and more accurately. These investments can connect with existing health technology systems, like electronic health records. Health systems are using it in administrative areas like revenue cycle management, finance, accounting, supply chain, human resources, and IT.
Natural language processing (NLP) and speech-recognition are leading to the expanded use of virtual assistants in the near future. Natalie Schibell, vice president and research director for healthcare at Forrester Research, expects virtual assistants to help with triage, appointment preparation, and advising patients how to prepare for a medical procedure.
AI and natural language processing (NLP) assist healthcare organizations in understanding the meaning of clinical data. For example, the Children's Hospital of Philadelphia used AI services to integrate and share genomic, clinical, and imaging data, allowing researchers to cross-analyze diseases, develop new hypotheses, and make discoveries.
BIOMEDICAL RESEARCH
In January 2023, a Canadian company announced that a therapeutic drug against COVID-19 had been designed entirely by generative AI — thought to be the world’s first. It’s been approved for human use and clinical trials. In preclinical trials, it showed efficacy against current COVID variants as well as other coronaviruses.
“The beauty of generative AI is instead of searching for a needle in a haystack, it generates a bunch of perfect needles for you,” Alex Zhavoronkov, co-CEO of Insilico Medicine, told the Toronto Star. “Then you just have to choose from those perfect needles — if you want the longer one or the one that is more stable.”
Decentralized clinical trials, which cast a wider net for participants in a variety of community settings, are being made possible by AI and remote technology platforms. More inclusive, faster and more efficient clinical trials will put better-tested drugs in the marketplace a lot faster.
Walgreens and CVS have jumped on the clinical trial bandwagon. Walgreens launched a clinical trial business in 2022, saying it would create a decentralized platform and in-person locations to recruit for and conduct clinical trials. CVS Health teamed with virtual clinical trial company Medable in 2022.
SDV INTERNATIONAL is supporting scientific research in military medicine and healthcare through its work for the U.S. Defense Health Agency (DHA). SDV INTERNATIONAL has created, used, and provided maintenance for multifaceted, leading-edge technology, including a machine learning initiative in a DHA data warehouse. In addition, SDV INTERNATIONAL is providing health services informatics and computational services.
COST SAVINGS
Health care costs continue to increase faster than the rate of inflation, impacting the budgets of government agencies, employers and individuals. Healthcare providers are under the gun to reduce costs by providing value-based care — judged by outcomes and not just procedures performed — and improving productivity. AI plays roles in both.
A survey by Nursing Times suggests nurses spend roughly one week a month hunting for equipment and supplies. AI assists with supply chain management so hospital staff know what supplies and equipment are available and, more importantly, where they are. The optimization of ordering and managing inventory in the supply chain alone could lead to millions in healthcare cost savings.
Waste, fraud and abuse cost the U.S. $100 billion a year, according to an estimate by the U.S. Justice Department. One estimate puts the cost at as much as $300 billion. A host of software providers are now offering AI products designed to identify errors or unusual activity that may be indicative of fraud.
One example of how fraud detection AI works is through the use of natural language processing. It can use an AI algorithm to identify signs of fraud that employees of healthcare organizations might not notice, such as unusual verbiage, claims being submitted from unusual locations or incoherent data.
Jamie Dimon of JP Morgan Chase shared to a finance summit in Sun Valley, Idaho in February 2023 that the bank records conversations with customers and applies AI-based technology to help determine whether future callers are the actual account holders based on previous verified calls.
Machine learning technology can also enable payers to review compliance on a greater number of claims in a more cost-effective and less burdensome manner. With manual review, only about 1% of claims can be analyzed, increasing the chances for overlooked fraud.
Overtreatment and ineffective treatment are also a major source of waste in health care spending. For example, false-positive mammograms make up about 10% of the tests performed each year, leading to costly and unnecessary biopsies. A convergent AI technology called iBRISK (Breast Cancer Risk Calculator), developed at Houston Methodist Hospital, uses natural language processing, deep learning, image analysis, big data and data mining to deliver a more concise diagnosis and reduce waste.
Out of 14,000 patients tested, 80% did not have to go for a biopsy — potentially saving billions of dollars a year.
DATA MANAGEMENT
RBC Capital Markets estimates that approximately 30% of the world’s data volume is generated by the healthcare industry. “By 2025, the compound annual growth rate of data for health care will reach 36%,” it says. “That’s 6% faster than manufacturing, 10% faster than financial services, and 11% faster than media and entertainment.”
Health data is derived from myriad sources, such as:
Patient data in electronic records
Insurance claims
Genomic and diagnostic tests
Clinical documentation and notes
Medical devices that make up the Internet of Medical Things (IoMT)
AI is being used to make sense of this disparate data and glean actionable insights. Health data is often unstructured, which makes it difficult to access, let alone interpret. AI can seek, collect, store and standardize medical data regardless of the format, assisting repetitive tasks and supporting clinicians with fast, accurate, tailored treatment plans without searching reams of paper or electronic records.
Mendel is an example of an AI company that standardizes unstructured data. Its optical character reader, optimized for healthcare, digitized more than 90,000 images and over 157,000 rich-text format files with only a 6% error rate for a company preparing an FDA submission.
Speech recognition and natural language processing are transforming unstructured data as well. AugMedix transcribes the natural conversation between clinicians and patients into clinical documentation using its proprietary natural language processing.
The process may also be reversed, transcribing clinical notes into language a patient can understand.
HURDLES FOR ARTIFICIAL INTELLIGENCE IN HEALTHCARE
The above topics are a sampling of how AI is reinventing healthcare. A recent SDV article about healthcare technology trends explored personalized medicine, telehealth, remote monitoring, virtual reality and other tech that wouldn’t work as well, or even exist, without artificial intelligence.
The future of health IT appears robust, but several issues must be addressed before wider adoption of AI in healthcare occurs so it can achieve its potential.
INTEROPERABILITY
Without interoperability among medical information systems, clinicians can’t see the whole picture. This can lead to poorer patient outcomes, readmissions, and higher costs. Health systems are moving away from siloed legacy systems, but it’s costly and time consuming to create new systems, and the all-important return-on-investment can be elusive.
In addition to efficiency and cost-effectiveness, interoperability is one reason why consulting firms like SDV INTERNATIONAL help their healthcare clients migrate to cloud services and APIs (application programming interfaces). In addition, AI and machine learning can automate processes to create standard formats that every participant in a system can access and use.
SECURITY, ACCURACY AND TRUST
Data privacy is particularly important as AI technologies collect large amounts of personal health information, which could be misused if not handled correctly. Additionally, proper security measures must be put into place to protect sensitive patient data from being exploited or hacked.
Electronic health records, a key provider of data for AI and machine learning processes, are only as accurate as the data that is entered. Care should be taken in the selection and customization of EHR systems and in the training of the staff using them to ensure accuracy.
Gaining acceptance and trust from medical providers is critical for successful adoption of AI in healthcare. Transparency in how the AI system is making decisions is essential. Clinicians must know that the guidance and recommendations they receive are accurate and appropriate to the patient.
REGULATION
The regulatory landscape for AI is evolving. For example, the U.S. Food and Drug Administration decides which AI-powered medical devices to authorize, such as wearable heart monitors. Last year the agency expanded its scope of regulation to include clinical decision support systems to predict sepsis, identify patient deterioration, forecast heart failure hospitalizations, and flag patients who may be addicted to opioids, among others.
While innovation develops at breakneck pace, regulators play catch-up. Outdated or ill-fitting regulations can hamper the development of, and patient access to, novel technologies that save lives. The European Commission has proposed legislation that will regulate the use of AI in healthcare. However, it doesn’t address how machine learning and artificial intelligence algorithms might be regulated.
Regulation that is too burdensome will stifle innovation. If it's too lax, it could risk patient safety and privacy.
WORKFORCE
The healthcare industry is facing a workforce shortage. By automating routine tasks and creating other efficiencies, AI can contribute to a solution. However, doctors, clinicians, pharmacists, therapists and other healthcare providers must be well-trained on well-designed systems. So that’s one challenge.
The other challenge is the types of employees who will be needed. “Multiple roles will emerge at the intersection of medical and data-science expertise,” predicts McKinsey.
“Designers specializing in human-machine interactions on clinical decision-making will help create new workflows that integrate AI. Data architects will be critical in defining how to record, store and structure clinical data so that algorithms can deliver insights.”
The role of clinical data managers will also change. Rather than collecting and analyzing research data, a data manager may oversee AI processes — or even be replaced by machine learning.
AI AS THE FUTURE OF HEALTH CARE
Today, there are more healthcare workers than ever, but there’s still a shortage. There will be unknown and unforeseen health challenges from climate change, the next pandemic or environmental damage. Can AI fill the gap? Maybe not entirely, but we feel confident that it will make our health care more personal, more accessible and more effective.
The U.S. government is betting on AI. A challenge recently launched by the General Services Administration is using an AI competition to improve mental health, addiction care, health equity, healthcare supply chain and safety, and cancer outcomes.
We at SDV INTERNATIONAL are proud to be creating the health care of tomorrow, working with the U.S. Defense Health Agency, the U.S. Department of Veterans Affairs and other federal and healthcare organizations. Our expertise in machine learning and health information technology can help prevent costly mistakes and help organizations enhance their capabilities with AI.
Contact SDV INTERNATIONAL today at our website, by calling 800-738-0669 or by emailing info@SDVInternational.com.