This article is part of the Forum Network series on Digitalisation.

Tremendous advances have been made in many areas in healthcare. Last year, the United States drug regulator approved an all-time record of 62 novel medicines, raising the average success rate to over 50 per year during the last five years. Cell and gene therapies are now able (and available) to cure aggressive hematologic cancers or other very serious conditions, and even to avoid blindness. Genomic sequencing is much better and more cost effective than even a few years ago. Most recently, artificial intelligence (AI) in healthcare has reached a level of maturity that opens unprecedented perspectives in terms of precision and performance for medicine. Medical algorithms that equal or outperform physicians are becoming the norm and dozens of clinical trials using AI products are published every year in all medical specialties. 

On the other hand, many health indicators are deteriorating in the Western world, and particularly in the United States, including reduced life expectancy and high infant, childhood and maternal mortality. Explanations are manifold, and probably vary among countries, yet it does not change the fact that several key health outcomes are worse today than they were previously. This is not linked to care rationing since expenditures have increased almost everywhere, with also more human capital dedicated to healthcare. 

AI is the new frontier

Artificial intelligence is about to transform all healthcare industries and could be part of the solution to some of the challenges encountered by healthcare systems. Its potential is huge, leading to realistic opportunities to increase access and enhance quality, safety and the patient experience. While these promises are widely accepted, we must also acknowledge that other pieces of the puzzle are more difficult to place. In particular, whether or not these advances will lead to a more efficient healthcare system is unknown. We also have very little idea of the final picture resulting from this likely transformation. The importance of respective stakeholders, either new entrants or historical players, may be reshuffled. For instance, the fate of physicians has been challenged by some, although I believe it is overblown to imagine fully replacing doctors by machines. 

The future role of drug developers is also not obvious. Some technological leaders have claimed that we would be less useful in this redesigned landscape. I am far more optimistic about the continuing contributions of pharmaceutical research companies, not least because we do not want to become less active in the global medical business. At least four lines of arguments support my view. 

We have an unparalleled expertise in clinical development of medicines

Over the years, we have learnt to conduct more and more sophisticated clinical trials, tailored to the increasing complexity of medicine. The development of AI products is not trivial, and they are often intertwined with other types of care. The external validity of already existing AI products remains to be determined, and many other challenges remain.

• The Technology Pendulum: From hope to fear – and back?

We have the necessary capabilities

Our scale, infrastructure and means place us in a strong position to help building novel products with a high level of evidence regarding their effectiveness and safety.

We know the healthcare ecosystem very well

We have partnered with patients, care providers, customers and regulators for a long time and we work hard together to maintain trust and develop solutions. The best way to disrupt healthcare markets with new technology is to engage with key stakeholders to understand their needs, and the needs of the systems in which they operate.

Many initiatives from talented companies have failed to disrupt healthcare markets due to a lack of engagement with and understanding of key stakeholders, and the systems in which they operate.

We have vast amounts of underused data

Even the most modern AI algorithms are data-greedy and need enormous amounts of data to be trained for optimal performance. We have data regarding biology, molecules, patients, clinical centres, diagnostic tests and the like. Of course, we already use those data in our day-to-day business. But these data need to be merged with other types of information to realise the full potential of AI, opening the way for collaboration with others.  

Find out more about the OECD Principles on AI

Let’s take for example the enormous burden of mental health, and particularly the 350 million people around the world battling depression. There is potential for AI to lend support to both the affected patients and their clinicians. Various tools that are in development include digital tracking of depression and mood via speech, voice, facial recognition, and the use of interactive chatbots. Machine learning has been explored for predicting successful antidepressant medication, characterising depression and predicting suicide, as recently described by Eric Topol in Nature Medicine, paving the way for new promises. 

We do not ignore that many significant companies are explicitly in the race to develop critical AI products in healthcare, and there is indeed a risk for pharmaceutical companies to invest in this not-so-familiar field. But we are used to managing risk; it has been an integral part of our business model for decades. 

The greater risk for us as an industry would be not to engage at all. 

Continue the conversation and help us co-create the agenda

All of the discussions you have on the Forum Network inform our thinking for the OECD Forum event each year – join to respond to Emmanuelle's article and comment to help us co-create the agenda! 

Related Topics

Banner image: Photo by Piron Guillaume on Unsplash