Artificial intelligence and machine learning can be used to advance healthcare and accelerate life sciences research. And there are many companies on the market today with AI offerings to do just that.
Derek Baird is president of North America at Sensyne Health, which offers AI-based remote patient monitoring for healthcare provider organizations and helps life sciences companies develop new medicines.
Baird believes some large companies have missed the mark on AI and ultimately dismantled public trust in these types of technologies, but that some companies have cracked the code by starting with the basics. He also believes AI success hinges on solving non-glamorous issues like data normalization, interoperability, clinical workflow integration and change management.
Healthcare IT News interviewed Baird to discuss the role of AI in healthcare today and how the technology can solve common problems and advance research.
Q: How can artificial intelligence and machine learning be used to advance healthcare and accelerate life sciences research? Where is it happening now, and what are the heavy lifts for the future?
A: AI is having a profound impact today on the ways we research drugs, conduct clinical trials and understand diseases. And while I think the current role of AI is just the tip of a very big iceberg, I think that as an industry we need to be much more careful about the ways we describe the promise of AI in healthcare.
There has been so much hyperbole around AI that I think we tend to forget that these technologies are just part of the healthcare equation, not some silver bullet that will suddenly solve everything. Biology is hard, complex and still very mysterious in many ways, but AI is already showing promise by allowing us to sift through vast amounts of data much faster and more intelligently than we could before.
One of the ways we are seeing the promise of AI come to life today is in its ability to help us go beyond symptomology and into a deeper understanding of how diseases work in individuals and populations. Let’s look at COVID-19.
Our big challenge with COVID-19 is not diagnosis, but that when we have a positive patient, we don’t know how sick they will get. We have an understanding of the clinical characteristics of the disease, but the risk-factors underlying the transition from mild to severe remain poorly understood. Using AI, it is possible to analyze millions of patient records at a speed and level of detail that humans just cannot come close to.
We have been able to correlate many of the factors that contribute to severe cases, and are now able to predict those who will most likely be admitted to the ICU, who will need ventilation and what their chances of survival are.
In addition to a powerful individual and population-level predictive capability, these analyses have also given us a great start in understanding the mechanisms of the disease that can in turn accelerate the development and testing of therapeutics.
That is just one example. There are many more. The last 12 months have been a time of rapidly accelerating progress for AI in healthcare, enabling more coherent diagnoses for poorly understood diseases, personalized treatment plans based on the genetics of treatment response, and of course, drug discovery.
AI is being used in labs right now to help discover novel new drugs and new uses for existing drugs, and doing so faster and cheaper than ever before, so freeing up precious resources.
There are many technical heavy lifts, but innovation in the field of AI right now is truly incredible, and is progressing exponentially. I think the bigger lift right now is trust. The AI industry has done a disservice to itself and science by relentlessly overhyping it, obfuscating the way it works, overstating its role in the overall healthcare equation and raising fears around what is being done with everyone’s data.
We need to start talking about AI in terms of what it is really doing today and its role in science and care, and we need to be much more transparent about data: how we get it, use it and generally ensure we are using it intelligently, responsibly and with respect for patient privacy.
Q: You say that some large organizations missed the mark on AI and ultimately dismantled public trust in these types of technologies. Please elaborate. You believe AI success hinges on solving non-glamorous issues like data normalization, interoperability, clinical workflow integration and change management. Why?
A: Companies spent billions of dollars amassing unimaginably vast data sets and promised super-intelligent systems that could predict, diagnose and treat disease better than humans. Public expectations were out of whack, but so were the expectations of the healthcare organizations that invested in these solutions.
I think Big Tech operating in healthcare has not been helpful overall, not just in setting unreasonable expectations, but often in putting profit before privacy, and by bringing the bias of seeing people as products – “users” to be monetized, rather than patients to be helped.
Public and industry confidence needs to be restored, and we need to correct the asymmetry between societal benefit and the ambitions of multinational technology platforms. In order to achieve that, the life sciences industry, clinicians, hospitals and patients need to know that data has been ethically sourced, and is secure, anonymous and being used for the direct benefit of the individuals who shared it.
Patients and provider organizations are rightfully concerned about how their data is secured, handled and kept private, and we have made it a top priority to build a business model with transparency at its center, being clear with stakeholders from the start about what data we are using and how it will be used.
As an industry we have to be clear, not just about the breakthrough medical developments we achieve, but about the specific types of real-world evidence we are using – such as genetic markers, heart rates and MRI images – and how we protect it throughout the process.
Once we get these basics of trust and transparency right, we can begin to talk again about more aspirational plans for AI. This means investing in robust data storage solutions, collaborating with regulators and policymakers, and educating the wider industry on the power of anonymized data.
These kinds of initiatives and partnerships will ensure we are all meeting the highest standards and allow us to rebuild longer-term trust.
Q: In your experience, what kinds of results is healthcare seeing from using clinical AI to support medical research, therapeutic development, personalized care and population health-level analyses?
A: With AI, pharma companies can collect, store and analyze large data sets at a far quicker rate than by manual processes. This enables them to carry out research faster and more efficiently, based on data about genetic variation from many patients, and develop targeted therapies effectively. In addition, it gives a clearer view on how specific groups of patients with certain shared characteristics react to treatments, helping to precisely map the right quantities and doses of treatments to prescribe.
For example, do all patients with heart failure respond the same to the standard course of treatment? Clinical AI has told us the answer is no, by dividing patients into subgroups with similar traits and looking at the variations in treatment response.
You need AI to break a population down based on many traits, and groupings of traits, to get this kind of answer, because the level of complexity quickly becomes too cumbersome for human processing. In this case, sophisticated patient stratification was used to improve clinical trial design and ultimately ensure heart failure patients are receiving the right course of treatment.
Comprehensive analysis of de-identified patient data using AI and machine learning has the ability to drastically transform the healthcare industry. Ultimately, we want to prevent disease, and by having more information about why, how and in which people diseases develop, we can introduce preventative measures and treatments much earlier, sometimes even before a patient starts to show symptoms.
AI is also increasingly being used for operational purposes in hospitals. For example, during the pandemic AI was used to predict demand for mechanical ventilators need.
AI will continue to be a driving force behind future breakthroughs. There are still many challenges that lie ahead for personalized medicine, and still a way to go for it to be perfected, but as AI becomes more widely adopted in medicine a future of workable, effective and personalized healthcare may certainly be achievable.