Is AI the Right Prescription for the Health Care Industry?

Artificial intelligence (AI) has given a fillip to cost-cutting efforts in a wide range of industries, including health care, an area in which rising patient costs remains a hot-button issue. For investors, opportunities could come from companies that can capitalize on AI as a way to cut costs via improved workflow and more efficient administrative tools. In addition, AI may lead to attractive growth in companies that can use it to accelerate R&D and/or be more productive in diagnosing patients. In our view, the health care space is ripe for disruption, especially in the biopharma segment, which is highly profitable but extremely inefficient. In this piece, we detail some of the most important AI trends for investors to watch in health care.

Background: Health Care Costs on the Rise

Rising health care costs are a challenge being faced by every country, but it’s a particularly acute issue in the United States (Figure 1). US health expenditures represented 18% of GDP in 2021, the highest among wealthy developed nations, and is growing at a 5% compound annual growth rate (CAGR).¹ This is unsustainable.

On a per capita basis, the US spent nearly twice as much on health care as comparable countries ($12,197 vs $6,345),¹ but we’re not getting much bang for the buck. Americans have a lower life expectancy and the gap has actually widened as our spend has increased (Figure 2).

How AI Can Help

• Administrative Tasks AI is already being deployed in familiar ways to help automate tasks and reduce the administrative burden (e.g. call center chatbots). McKinsey estimates that roughly a quarter of US health care expenditures is on administrative tasks; simplification and automation could result in 25-30% savings. This would also free-up clinicians to spend more time with patients, while reducing a major contributor to burnout.
• Radiology Perhaps where AI has made the biggest splash is in Radiology. AI is simply better at processing images than the naked eye. AI is particularly well-suited to radiology because medical images contain vast amounts of data. It’s basically a pattern recognition exercise and AI began outperforming humans nearly a decade ago. By leveraging machine learning, AI-enabled equipment can more quickly and accurately process medical images. This could lead, for example, to earlier cancer detection, resulting in a much better outcome for the patient and lower costs for the health care system.
• Biopharma While Medical Devices may be more penetrated, Biopharma is where AI stands to have the greatest impact. On average it takes 10 years and $2.6B to bring a drug to market.² AI can have a profound impact here, both in terms of speed and cost. Traditional drug development involves trial and error to discover molecules, followed by long, expensive clinical trials, in which over 90% of drugs fail. Typically, thousands of molecules are tested before one is taken to clinical trial. And it takes 10 years, on average, from the start of the clinical trial to approval, if you’re lucky enough to make it that far. Only 8% of candidates cross the finish line.³ Advancements in generative AI can help. 
• Molecules aren’t intuitive to humans. While we can understand and manipulate English better than AI can, for biology, it’s the other way around. To generative AI, proteins and RNA are no different than French or Spanish. It’s just another language, and AI has the potential to understand this language much better and faster than we can. This can help predict the shape of proteins, how they may interact with a molecule, and whether there’s an optimal receptor to target—again, observations that aren’t intuitive to humans.
• Clinical Trials It's not just in the discovery phase where AI can help. Clinical trials are the most time consuming and expensive part of drug development. AI can help speed site selection and patient recruitment. And for pharmaceutical companies, time very much is money. Patents are filed before the clinical trial starts and only last 20 years. Half of the patent life is consumed by clinical trials. A better hit rate and faster development means more time on market. Lower costs means better margins. And hopefully some of this will be passed on to the consumer.

Case Study: Exscientia

Exscientia is an example of an AI-first biotechnology company. By using successive AI-driven design cycles, company scientists were able to identify an AI-designed kidney cancer drug by testing 80% fewer compounds (160 instead of 2,500), taking 90% less time (9 months instead of 54) than the traditional approach. This is the company’s third AI-designed drug in clinical trials. Exscientia thinks AI can reduce drug discovery time by 70%, decrease the number of compounds synthesized by a factor of 10, and lower the capital required to go from finding a biological target through completing preclinical toxicology studies by 80%.⁴

Hurdles to AI Adoption in Health Care 

While AI is likely to significantly impact the way we receive care, it’s not a silver bullet and its impact will be gradual. There are several headwinds to the industry’s adoption of AI.

First, the human connection remains vital in many facets of health care. You build up a level of trust and comfort with your doctor. Health care is still very personal. According to a Pew Research Center survey, the majority of Americans (60%) are uncomfortable with health providers relying on AI to make clinical decisions. Women, in particular, expressed discomfort (66%).

That said, the level of patient comfort does vary based on what AI is being used for: There was widespread support for skin cancer screening (65%). Conversely, people were very opposed to mental health chatbots (79%) or using AI to prescribe pain medication (67%).

Additionally, in the Pew study, Americans were evenly split on whether AI would make the quality of health care better or worse. They did believe it would result in fewer mistakes and less racial bias. However, there were data security concerns and a belief relationships with providers would deteriorate.

Second, certain parts of health care require physical interaction. An in-person check-up is still more effective in certain instances. And AI can’t draw blood or perform a biopsy.

Third, the health care industry tends to be very conservative and is slow to embrace change. Health care practitioners have been conditioned to base decisions on evidence that usually comes from multi-year clinical trials. They need to see the data before they’re willing to support new approaches.

Finally, whether they want to admit it or not, health providers are financially motivated. You can have a great product that’s FDA approved, but if providers aren’t getting paid to use it, they won’t. Reimbursement is the single largest hurdle to AI adoption. According to a Boston Consulting Group/UCLA Biobank survey, 72% of respondents cited reimbursement as the barrier preventing patients from accessing novel medical devices in the US. Only 16% cited FDA regulation.

The Bottom Line

Health care costs are an increasing global burden. AI can help bend the cost curve, but it’s not a panacea and will take time. Even so, for most large health care companies, rather than being a differentiator, it will be table stakes. There are some small AI-first biotechs or biosimulation software providers, but large cap companies will be increasingly inclined to adopt AI in one form or another to keep up with industry innovation. A company’s adoption may simply be on the cost/administrative side, or it can be more transformational, such as new work in drug discovery and development. Still, importantly, the large health care stalwarts will not be the only adopters of AI, with peers left behind; in our view, all health care companies will be leveraging AI.

For the Life Science Tools industry, AI may represent a longer-term risk given the potential wealth transfer from instruments and consumables to semiconductors and super computers. On the flipside, the consumer should be the ultimate winner. AI promises more life-saving drugs and devices that come to market sooner and are offered at lower cost. AI also could spawn improved practitioner productivity, which will broaden access to health care, and a healthier population, which should drive economic growth.