Product Clinic #9: Bringing about the right kind of change with AI

“AI is going to change the world.” 

‍

I’m not quoting anyone in particular, because I don’t feel like I have to. We’ve all heard that statement, or a variation on it, countless times. And it’s undeniable. We’ve witnessed the power of ChatGPT firsthand. As a coder, AI-assisted programming has already changed how I work.

‍

But is something trained only on existing data truly capable of bringing about change?

‍

Having a conversation with ChatGPT, while often brief and transactional, is occasionally philosophical. It makes you question the meaning of consciousness. Am I just an advanced automaton who is trained on hours of observing my surroundings? Or is there something special about being human that differentiates me from this machine?

‍

The distinction between human consciousness and AI that I always come back to is original thought. Undeniably, humans are capable of genuinely original thought. All our technological achievements are a testament to that. As it stands, however, AI can only combine and regurgitate ideas, not create them. As ChatGPT puts it: “AI can generate new combinations of ideas and simulate certain aspects of human thought, it does not possess genuine original thought in the way humans experience it.”

‍

On that basis, I’d like to challenge the statement “AI is going to change the world”. This statement is only true for some definitions of change. Fundamental change is impossible with AI alone. It requires original human thought. 

‍

I would argue that a better statement is “AI will optimise the world”. And it would be optimising a world full of broken systems and misaligned incentives.

‍

One example of change wrought by AI is its applications in making healthcare decisions. Some algorithms will assign Black patients a lower risk level even if they are sicker than their white counterparts, resulting in worse care. Since the AI was trained on biased data, the outcome exacerbates that bias. Hardly the change we’re looking for.

‍

What does this have to do with clinical trials?

‍

It’s well established that the current clinical trials ecosystem is broken. Before considering how AI can be applied to clinical trials in an effort to fix them, it’s worth examining exactly how clinical trials are broken. For those not familiar with how clinical trials are often run, the state of affairs can be shocking. Two of the key issues with clinical trials today are:

‍

1. Broken Incentives: The classical CRO bill-by-the-hour model provides no incentives to run trials efficiently. In fact, quite the opposite. CROs are incentivized to slow trials down and produce as many billable change orders as possible! In the words of an anonymous CRO executive: “We make more money the worse the trial goes.”
2. Broken Processes and Tech: A trial is complex and has many moving parts. In the classical CRO model, each piece is subcontracted, therefore splitting data across multiple technology stacks and making oversight more challenging. Looking beyond the terrible experience this creates for patients and clinicians when so many vendors and technologies are involved, the opportunity for error here is enormous. Tragically, it's not unheard of for trials to fail because of entirely preventable problems.

‍

It’s naive to assume that AI could resolve any of these issues alone. At best it could optimise these broken systems. AI is just a supercharged next-generation band-aid, but the cure lies in re-imagining how trials are organized through improved processes and superior systems.

‍

Does that mean there’s no use for AI? Of course not. Given the correct application, AI can help us move at an unprecedented speed. But it can only optimize so far as the system allows it. 

‍

At Lindus Health, we approach the problem of inefficient clinical trials in two ways. We fix the systems and apply AI/ML to everything that isn’t broken.

‍

Fix the systems

‍

• When Lindus Health takes on a trial we act as an owner, not a subcontractor. It’s part of our DNA to care deeply about the success of our trials because our mission is to deliver treatments, not invoices.
• At Lindus Health, we are a CRO, but we put technology first. We’re deeply invested in building end-to-end clinical trial software that collects the right data and stores it in a submission-ready format. This allows us to deliver at a level that a normal tech-agnostic CRO cannot.
• Our efforts in implementing digital data flow mean that we can massively reduce trial execution risk and avoid unnecessary delays due to human error or incompatible systems.

‍

Apply AI/ML to everything that isn’t broken

‍

• Predicting study success early can save a massive amount of resources. Our model, published in Nature, allows us to optimize trial design and enable treatments to reach patients faster.
• Protocol writing is a slow but important process. With the help of the AI protocol generation tool recently demoed at the CNS summit, we are able to save our clinicians countless hours while ensuring higher quality and more standard ICH M11 compliant protocols.
• Identifying issues in data is an area where AI excels, and we’ve used it to augment our risk-based monitoring capabilities. We recently won “Best in Stream” for a talk delivered at PHUSE EU for demonstrating how we used AI to identify incorrectly entered data.

‍

These are just three examples from a list that keeps growing as we find more and more suitable use cases.

‍

Starting by recognizing a broken system doesn’t generate as many headline-grabbing statements as shiny new AI initiatives, but it does mean that we can bring about fundamental change. And fundamental change means patients get the treatments they need sooner.

‍

Amiel Kollek, Senior Software Engineer at Lindus Health

‍

—

‍

This is part of our product clinic series, where we discuss how we build at Lindus Health. Check out our previous posts on how we prioritize and use technology, processes, planning, and rituals to build great products.

‍

About Lindus Health

‍

Lindus Health is an anti-CRO running radically faster and more reliable trials for life science pioneers – bringing ground-breaking treatments to patients more quickly.  Lindus Health does this thanks to a commercial model that aligns incentives (fixed-priced quotes per study, with milestone-based payments), marrying a world-class clinical operations team with its unique software platform, and access to 30 million Electronic Health Records. Clinical trials are the biggest bottleneck to advances in healthcare and by removing this constraint they aim to improve health for everyone. They handle the end-to-end execution of clinical studies, including design, patient recruitment, clinical data capture, monitoring and project management.

‍

Lindus Health has to date delivered more than 90 trials across the US, UK and Europe to tackle a range of conditions including diabetes, asthma, acne, social anxiety, major depressive disorder, hypertension, chronic fatigue syndrome and insomnia.

‍

The company was named after James Lind, who pioneered the first clinical trial and treatment for scurvy, and co-founded by Michael Young, a former Special Adviser to the UK Prime Minister on Life Sciences, and Meri Beckwith, a former life sciences investor.

‍

The company has raised over $24M from investors including Peter Thiel, CREANDUM, Firstminute Capital, Presight Capital, Seedcamp, Hambro Perks, Amino Collective and Calm/Storm.