How Doctolib uses AI to empower healthcare practitioners

Artificial intelligence will make care delivery even more human

What happens next when I tell people about my work is a bit like flipping a coin. Artificial Intelligence (AI) sits right at the frontier between science and magic, and triggers high hopes, wild dreams, and deep-seated fears. So I know that I can be spending the next few moments discussing how AI will make the world a better place, or how it will eventually tear the fabric of society. Either way, I’m usually in for a passionate conversation.

Even more so when I say that I am working at Doctolib.

These reactions make sense. AI is a transformative technology on the same scale as the printing press: it will change our lives and our minds. The opportunity is as huge as the pitfalls are deep. This is true in every domain, but even more in healthcare where every decision matters.

With that in mind, how do we see AI supporting healthcare practitioners (HCPs) in their activity? What is our approach when developing AI powered products? Here is an overview of how we think about it and the positive impact it could have.

Technology to build more humane interactions

Doctolib has always used technology and innovation to improve the daily lives of care teams and health for all. Our vision is that technology should provide support and relief so that our users can focus on what truly matters: relationships and patient care. We see AI as a fantastic enabler to move from vision to reality.

AI will also contribute to relieving the pressure on our healthcare system. As the population ages, the demand for care increases, leaving healthcare practitioners struggling to give patients the care they need. AI can ease HCPs’ struggle at a scale never seen before.

Care teams are made of superheroes. Doctolib will use AI to give them superpowers. The first superpower we will provide is a medical assistant to support practitioners during and outside of consultations.

• During consultations, the medical assistant will take notes for them and file all patient information. It will also write up the diagnosis and treatment plan. The assistant will provide access to medical resources and support HCPs in suggesting the right prescriptions,
• Outside consultations, the medical assistant will be able to prepare responses and tasks associated with certain patient requests as well as exchanges with other HCPs, as part of care coordination.

This may seem counterintuitive, but artificial intelligence will make care feel more human. We want practitioners to spend as little time as possible in front of their computers, and more time listening to and interacting with their patients. Healthcare practitioners will, at last, be freed from keyboard typing.

High impact, high stakes

This is the vision, but we are already using AI to clear healthcare practitioners’ pain points, enabling them to deliver better care while saving precious time. Our efforts so far have been directed at bridging the gap between:

• practitioners, who work mainly with unstructured data (text, audio, images, etc…), and
• their clinical software, which works best on structured data (codified data in tables) — at least for now as Large Language Models (LLMs) are reshuffling the cards.

In other words, rather than asking practitioners to fill in endless forms, we use Natural Language Processing (NLP) or other techniques to do the heavy lifting for them and deliver a seamless experience. For example, we are categorizing the documents that HCPs receive automatically. We also automatically structure patient vitals from lab results and prescriptions from practitioners to secure them.

This may not seem like much, but running these simple machine learning models at Doctolib’s scale provide humongous time savings, equivalent to adding dozens of HCPs in the healthcare systems of the countries we operate in. You can now imagine how deeply the superpowers we are working on can improve the way HCPs deliver care and increase the performance of the entire healthcare system. The potential is massive and there is a very real opportunity to have most of us live a better and healthier life.

There are also very real risks that we need to keep in check. Doctolib is already well equipped to tackle many of them which are not relevant to AI specifically but to healthcare as an industry. For instance, when it comes to data security and privacy, we already have tools, frameworks, and expertise to address these risks. And we have certifications to prove it.

However, AI consistently presents new regulatory and ethical challenges that we will need to address, especially in healthcare where there are lives at stake. Of course, we want practitioners to remain in the driving seat and ultimately make decisions on how to provide care best. But we recognize that the AI powers supporting them will have an impact on the care their patients receive. And we strive to make this impact overwhelmingly positive.

On the surface, many of these challenges will look new. After all, there has never been anything exactly like generative AI in the past. But parallels can be drawn between AI in healthcare today and what happened in the past when new technologies have been introduced in a mission critical setup. In order to start leveraging CPUs in aircrafts, auditability, explainability and reliability concerns had to be addressed. Similar concerns are expressed on AI today. They may have seemed insurmountable to an observer of that time. Yet, we’ve found a way.

The key ingredients to transform healthcare with AI

In fact, I find as much satisfaction in tackling these challenges as in improving the raw performance of machine learning models. They are not side topics that have to be dealt with. They are what will make AI in healthcare acceptable so that we can improve health for all.

But in order to get there, one must start by being passionate about healthcare and the people who work in healthcare. It is not a simple domain to get into. The best way to learn about it is to work closely with practitioners, secretaries and assistants. Showing curiosity and empathy is the way to build great products for them. This attitude is embedded in Doctolib’s culture. It is how we innovate, and it is how we create the next AI superpowers for them.

In case you’re reading this wondering how to come along in this journey, feel free to look at our current job offers: click here!