Dr Michal Tzuchman Katz, a paediatrician and co-founder of Kahun – the company behind the first AI chatbot for clinicians – is uniquely placed to understand the role that AI can play in the future of healthcare.
Coming from a background as a software engineer and later taking the leap to train as a paediatrician, Dr Katz saw firsthand the potential AI has to improve patient outcomes.
Her time in the ER showed her how valuable these tools can be in improving the quality of care through standardisation, alleviating the burden on healthcare professionals and eliminating bias,
This ultimately led her to co-found Kahun, an AI clinical reasoning tool, which recently announced it is integrating ChatGPT into its XAI chatbot – the only XAI chatbot for physicians, helping doctors make better-informed decisions based on the company’s proprietary map of more than 30 million evidence-based medical insights.
Dr Katz spoke to HT World about the potential of AI to enhance healthcare for physicians and their patients.
Can you tell us about the motivation for founding Kahun?
I’m a physician so I know how we work and how the burden of documentation and cognitive overload is something that really impacts how we practise medicine. It came from the pain point of understanding that and realising that there is an opportunity to bring technology tools to the point of care to help providers practise better medicine. Even if it’s starting by just improving their well-being, this will cause a chain reaction that will eventually improve the outcomes of patients.
Can you talk us through how the platform actually works in practice?
When we started the company, we weren’t the first ones to build an AI tool for physicians – many companies have come to the conclusion that these tools need to be in the work of care.
We knew we were going to build tools that would be in the hands of doctors making decisions, so we knew that whatever technology it is would have to be trusted.
It would have to be explainable, transparent and traceable to the sources that are used to help physicians make decisions.
As a physician, I’m not going to trust a technology or a software solution that will just tell me what to do, if I cannot understand what’s behind the recommendation.
This is what we set out to do.
At the core of Kahun is a knowledge graph, it’s a structured representation of evidence-based knowledge, which is the basis of every decision that we as physicians need to make.
When used as a decision support for the providers, the recommendations can fully be traced back to the guidelines, protocols and references.
It’s almost impossible to do a literature search for every patient that you are seeing, with the shortage of staff, the overload of patients and comorbidities, you have very little time in which to provide the best quality care for your patients.
It’s bringing all this information, all this evidence, together to help you analyse it and make a decision.
And how does this trickle down to benefit patients?
The main idea and the main incentive for all of these solutions is to create a more standardised and better quality care for as many patients as possible.
When physicians are empowered with these tools, they are able to make a decision for their patient that is unbiased and non-anchored and that makes the best use of the evidence.
This could mean earlier diagnosis. It could mean eliminating unnecessary tests. It could make sure that they’re getting the right treatment for their complex presentation, taking into account their background, their demographics, risk factors and so on.
It means being able to make a personalised decision about how to treat them.
There are many outcome improvements these tools will make to improve the lives and quality of the patients.
You mention bias, how does this tend to impact decision making in healthcare?
We are all human and we all have a tendency to collect data from our patients and decide what we think could be the reason or the possible cause for their presentation and continue anchoring on that path, making diagnostic tests on that path, making decisions and treatment based on that path and not questioning the initial position.
By bringing empowering, comprehensive tools to the point of care, helping physicians to double check and make sure that our decisions are not biassed and not anchored, but rather are looking at our patients in a holistic manner, then we eliminate some of these human limitations.
What has your background in software engineering taught you about the role of AI in healthcare?
I think the unique position that I am in is that I understand both the technologies and what they can offer and also the clinical expertise that is needed at the point of care for us to deliver the best care.
I think that also gives me an advantage for realising the limitations of both.
Generative AI, for example, is an amazingly powerful tool and has the potential to have a lot of impact, but these technologies too have limitations.
I understand how these models work and their ability to predict the next word in a sentence.
But medicine is so much more complex and therefore the combination of leveraging the scientific knowledge together with these text generative models is something that has the potential to be both impactful but also trusted and safe and used in a way that is compliant in healthcare.
Do you find there is still scepticism among healthcare professionals around the use of AI?
Of course, there has always been hesitation and scepticism about the adoption of AI in healthcare.
The physician takes the responsibility for any decision that they make, so these tools have to be transparent and explainable for physicians to adopt them in scale.
[However], over the last five years there has been a great change in the willingness to adopt these technologies.
I think as an industry, healthcare realises that it is very conservative but equally, in such need of these technologies, so regulators and legislators have to make room to enable them to come to life at the point of care.
These key decision makers and stakeholders of the industry are advancing the adoption, which is great.
There are obviously risks when AI is not used properly. What do healthcare providers need to consider in order to integrate it safely?
Any healthcare organisation that opens itself for these models needs to take a very stepwise approach.
They should examine and analyse the way that these systems are influencing the decision-making process and the outcomes of patients, making sure that we are always aware of the limitations of these models and examining what type of models will be needed to reduce the risk.
Kahun recently announced it was integrating ChatGPT into its XAI chatbot, how do you expect this to enhance the platform?
For Kahun, it’s a very natural collaboration because leveraging the fluency of the generating models into our solutions makes total sense with the combination of a clinical reasoning engine that is transparent and explainable and leverages the knowledge in a way that physicians can trust, together with generative fluency LMS.
We enable the providers to use systems that reduce both the documentation and administrative burden, but also the cognitive burden.
Physicians don’t need to have a prompt engineering background to optimise their prompts so that they will get the best response, this is what Kahun takes care of for them.
It feels like we are at the tip of an iceberg. What do you think the future holds for AI in healthcare?
If I imagine what a physician’s work day would look like in three years, AI is going to be a game-changer.
Systems will be increasingly adopted into workflows and physicians will be able to spend more time with patients and less time on the computer and documentation.
They will be more empowered with knowledge so that they feel they are really working at the top of their licence.
