As the healthcare and pharmaceutical industries rush to tackle a wide range of COVID-19 challenges, from developing treatments and sourcing medical supplies to mitigating outbreaks and racing towards a vaccine, the use of AI-driven solutions is accelerating rapidly. News about AI efforts in healthcare breaks daily: Hospitals in the U.K. are turning to AI to help triage COVID-19 patients. The CDC hosts an AI-driven bot on its website to help screen for coronavirus infections. The FDA recently cleared several AI-driven devices to screen COVID-19 patients for heart problems.
This need for AI’s algorithmic modeling power is potentially reshaping the future of a highly regulated industry that has historically been conservative about implementing new technologies. In fact, AI spending in the healthcare and pharmaceutical industries is expected to increase from $463 million in 2019 to more than $2 billion over the next five years, according to a recent study by global tech market advisory firm ABI Research.
This signifies maturing capabilities and growing investor confidence in that domain, says Lian Jye Su, principal analyst at ABI Research. “AI, particularly machine learning, enables the ingestion of large volumes of data, be it time series data, images, videos or treatment records, and shows patterns and correlations in the data,” he says. “The ability to infer insights from large amounts of data now allows doctors to make better diagnoses and prescriptions based on statistics.”
Before the COVID-19 crisis, AI and machine learning technologies had just begun to move beyond the pilot stage at large healthcare companies and were considered “nice-to-have” ideas at small and midsize organizations. Certainly, some aspects of AI were making progress, particularly around speech and natural language processing. But true data analytics insights were weak, says Kali Durgampudi, chief technology and innovation officer at Greenway Health.
“The healthcare industry was very behind the curve on AI compared to other industries,” he says, adding that access to data was and still is a huge reason for the slow uptick. “Patient data has been unstructured and locked inside disparate databases,” he explains. “It was anticipated that regulatory reporting requirements would help support data gathering and AI, but unfortunately, that did not come to bear prior to COVID.”
AI in healthcare: moving from behind the curve to the forefront
In the past, healthcare companies had looked to AI solutions from a reimbursement angle as they searched for ways to reduce cost and boost efficiency — especially as companies transitioned from fee for service to value-based fees with the implementation of the Affordable Care Act in 2010, says Justin Richie, data science director at digital consultancy Nerdery.
With COVID-19, AI is now being tapped for epidemiological research around how things happen, as well as its ability to run millions of simulations to understand behaviors and trends. But a tremendous learning curve remains, as organizations still try to demystify what AI even is, says Richie. “There is still an inordinate amount of time spent explaining what AI might mean to organizations,” he explains. “The bottom line is that AI will solve a lot of problems through sheer computational force, to identify patterns that the human brain cannot comprehend.”
Still, the COVID-19 pandemic has made it clear that response times are critical to reducing the spread of disease as well as finding a cure and reducing death tolls, which is where AI comes in, says Gonzalo Raposo, life sciences tech manager at IT and software development company Globant.
Collaborations between healthcare professionals and AI models can work to predict contagion rates, forecast for needed medical supplies, research potential cures or treatments and even provide new diagnosis methods. “Who would not appreciate a faster and more accurate diagnostic?” he points out. “Who would not appreciate a reduced time for researching new drugs and vaccines based on a more accurate prediction of the potential outcome?”
While experts agree that AI and machine learning will be game-changing in the healthcare and pharmaceutical space, they emphasize that data is the common and most important factor to even enable model-building. “Both algorithms and data are driving outcomes and transforming healthcare; they go hand-in-hand,” says Dr. Rama Kondru, CIO/CTO of Medidata. “AI and machine learning are only as good as the data fabric they are built on.”
However, the pandemic also uncovered AIs current limitations, such as poor-quality data, unavailability of the right data sources, lack of AI skills sets and the explainable or accurate results of AI tools. Besides access to data, one of the biggest challenges facing AI adoption is not technological. Instead, the infrastructure to aggregate and analyze the data is missing, says Durgampudi. “There needs to be a central entity where all the data can be aggregated, or a set of standards by which to correlate all the data,” he says.
In addition, there needs to be interoperability of data. “For example, there’s been a push to report COVID cases directly from electronic health records to the CDC to support the development of vaccines and treatments,” he explains. “This requires interoperability of data, with the layering on of AI and machine learning to mine the data for meaningful insights, and then accelerate vaccine and therapy creation.”
Now, though, more life sciences and healthcare companies have begun to focus on gathering the right data sources to address the right business problems at the right time, says Angela Radcliffe, research and development lead, life sciences at Capgemini. “That’s the key in delivering the top-line growth with AI technologies,” she says.
The biggest areas of AI opportunity in healthcare
The opportunities for AI in healthcare, both during and after the COVID-19 crisis, are vast. AI is now truly at the forefront during this COVID-19 era, emphasizes Radcliffe, with tremendous potential in accelerating the race towards bringing faster access to treatment. These include an AI-enabled digital platform for antiviral drug repurposing for faster access to COVID-19 treatment; identifying the AI-enabled drug biomarkers for a potential COVID-19 drug; AI-enabled virtual trials with connected IoT devices for real-time monitoring of health outcomes; automated drug safety analysis for pharmacovigilance, and much more.
According to Radcliffe, the use and adoption of AI in the clinical research and development space — including gene analysis, pre-clinical trials, patient recruitment and drug safety analysis — holds some of the biggest opportunities. In addition, more healthcare companies are adopting or getting serious about AI-based solutions in areas such as FWA (fraud, waste and abuse), followed by other commercial pharmaceutical areas including forecasting supply chain issues.
Accelerating the discovery and development process of new drugs through AI is a tantalizing possibility, adds Su. “Companies such as Benevolent AI, Insitro and Zymergen offer a large library of chemistry, biology and material engineering knowledge, producing enormous structured, curated and qualified proprietary ‘data lakes’ of dynamic usable knowledge.”
AI will also ultimately support areas such as personalized treatment, leveraging advanced analytics and machine-learning algorithms to provide a platform that discovers treatments for rare genetic disease and other illnesses. And screening and detection will become more automated, accurate and non-invasive thanks to AI, using deep-learning models to perform medical image diagnosis.
Certainly, AI will help accelerate the search for and the creation of a COVID-19 vaccine, says Durgampudi, with more access to data overlaid with AI to help create new alterations of vaccines that are better targeted to individuals. “For example, by leveraging data on individuals who received a seasonal flu or SARS vaccine — and how they responded to it — will inform how we conduct clinical trials and the development of next-generation vaccines,” he says.
In addition, as the FDA eases regulations to put technology in place that would normally take months or even years to approve and roll out, AI will have a big role in changing the way clinical trials are run, says Dr. Kondru. Synthetic control arms (SCA), for example, leverage advanced AI analytics with patient data from historical clinical trials to mimic the results of a traditional control arm (or placebo). “This is already in use, such as in an ovarian cancer drug trial, but AI-enabled synthetic control data can also determine safety for certain experimental treatments that may be under review for cross-indication use, which is important in the search for COVID-19 treatments,” he says.
What the future holds for AI in healthcare
Over the next six to twelve months, as more and more COVID-19-related data is collected; activity promoting the use of technology in clinical trials increases; and regulators and stakeholder groups demonstrate pragmatism and flexibility, there should be more AI model and simulation adoption across the entire clinical trials landscape, says Dr. Kondru.
Over the coming year, there is no doubt that both startups and established companies will continue coming up with innovative ideas and solutions around the use of AI in healthcare and pharma, adds Raposo. “Something that we all learned from this pandemic is that we have to be prepared to respond fast,” he says. “We can’t wait five years for a vaccine for COVID-19, for example —I like to think that isolated cases of laboratories using AI in pharma will become something widely used in the mid-term.”
Yet, given the slow adoption of innovation in healthcare generally, there still may be a lag in significant AI-driven development, cautions Su. “Once we get past COVID-19, though, healthcare regulators and agencies may start to pay more attention to the benefits of AI in the automation and augmentation of existing medical and R&D processes,” he explains.
The key to success, says Dr. Kondru, is a data-driven approach and building robust AI and machine learning models and analytics in a transparent and clear manner. “This will help regulators and pharmaceutical companies be more willing to adapt to new technologies, helping to overcome some of the barriers and challenges that COVID-19 has created for clinical trials operations,” he says.
Over the long haul, healthcare companies will have no choice but to implement or strengthen AI solutions to stay ahead of the curve, Radcliffe predicts. “The status quo won’t be the same post-pandemic,” she says. “We’re going to see the healthcare industry being proactive and establishing more stringent AI and data strategies enterprise-wide.”
With the availability of ever-increasing diverse datasets and technologies, availability of niche AI skill sets, increased collaborations, and faster time to market vision, AI-enabled solutions will become extremely critical, she adds. “AI will be necessary to improve financial outcomes, to improve member experience, to intelligently process claims, and most importantly, to prepare for similar critical situations in the future.”