How AiCure is bringing computer vision to health care: an interview with CEO Adam Hanina
This interview is also published in The Wharton Journal here.
As AI progresses in the health care industry from buzzword to use at scale, applications of machine learning and neural networks are consistently raised. Far less frequently mentioned is the use of computer vision, and analysis of image data. AiCure has built one of the healthcare industry’s first computer vision algorithms (founded in 2008), with numerous patents, and recently full focus on improving medication adherence. Below is a conversation with AiCure’s CEO and fellow Wharton alum, Adam Hanina.
Adam, thanks for speaking with me!
To start off, what do you think the impact of AI in healthcare will be for patients?
The human impact will be vast. I think AI will prevent tens of millions of hospitalizations and accelerate drug development by decades. With an estimated one million nurse shortage in the US by 2020 and a rapidly aging population, there is a huge need to fill the care gap: technology and adaptive tools will step in. The automation of healthcare tasks traditionally performed by humans and accessible to a select few will now be accessible to a much broader patient population.
Like all technologies, the impact of AI will be felt incrementally in the beginning but eventually will be adopted systemically across all of healthcare. As AI works up the value chain from decision support and analytics to full diagnosis and treatment, applications will become seamlessly integrated into treatment and the benefits quantified and clinically validated. They will be our new standard of care.
Regarding the role of AI in healthcare, can you elaborate on what the role of clinicians will be once we reach full adoption?
For the time being, AI will only augment clinical care — it will not replace the clinician. Unlike autonomous vehicles, clinical care requires a great deal of human interaction and clinical evidence. The regulatory environment is much slower to adapt than in other industries. Greater insights will evolve clinical roles to focus on better patient care, oversight and a more granular analysis of the clinical problems at hand. AI will become collaborative with clinical teams.
How do you feel computer vision and the algorithms AiCure is building will be different from the machine learning and predictive analytics that typically get discussed within health tech circles?
Computer vision or visual recognition allows computers or smartphones to see. They offer an inbuilt platform for patient observation and interaction. When a nurse cares for a patient, they use visual cues to assess, interact, and assist patients, ensuring high standards of care. At AiCure, we are automating this form of visual observation and interaction through intelligent software. The core platform offers interactive assistance to patients on their smartphone, performs visual dose confirmation, allows for coordination of care, and is even starting to visually assess disease progression over time.
AiCure is capturing one of the largest proprietary visual asset libraries by indication, patient demographic, geography, and over time. It offers a profound view of human behavior that was previously only accessible via a care provider. AiCure is unique in this way. Other companies typically have a build-and-integrate approaches to accessing data through external electronic medical record systems, imaging scanners, or more commoditized data sets.
Beyond the technology, it is critical to demonstrate clinical merit and validation. A 2017 randomized controlled trial published in the American Heart Association’s Stroke journal, showed a significant delta between patients using AiCure and treatment as usual. According to plasma levels, 100% of the patients in the AiCure group were within therapeutic range compared to only half the patients in the control group. What these data show is that without AiCure, patients on anticoagulation therapy were at serious risk of stroke and bleeding — a $65 billion-dollar economic burden on the health care system. AiCure is replicating similar results across multiple indications.
AiCure is one of the few venture funded AI startups in healthcare that has grown past pilots to sell contracts to large healthcare players. What allowed you to successfully scale?
We developed an economic pain map and went for the bullseye. If you can target groups whose business model is directly impacted by a problem you can solve and for which you have a solid clinical evidence base, they will then become dependent on the technology. When we first deployed for a top pharmaceutical company, I felt a real adrenaline rush. They were not only using our platform but their trial depended on the data we were generating.
Given the recent entry of large non-health care players such as Amazon, JP Morgan, and Berkshire Hathaway, what do you believe the role of startups in healthcare will be?
Generally, it bodes well for the startup environment and defines clear opportunities. Larger companies build platforms and are exceptionally proficient at process and efficiency. Smaller companies that are able to demonstrate clear improvements in health outcomes and that are clinically validated over a multi-year period will offer an attractive M&A opportunity. No one company, no matter how large, will do this alone.
And it is not just Amazon entering the space — the ecosystem is changing rapidly. The acquisition of Flatiron Health by Roche is just the beginning of pharma stepping up their commitment to the tech space as they realize that health outcomes are the real currency. Perhaps in 2022 Verily will buy Novartis.
What’s preventing some of the larger incumbents from building these AI tools? How is this business defensible in the long run?
AiCure is adopting a long-term strategy centered around four key pillars: 1) proprietary technology (close to 100 patents pending/awarded in the US and globally); 2) largest video asset base by indication and longitudinal data; 3) clinical evidence and validation; and 4) first in class both in drug development and population health.
There is a lot of talk around big data being the panacea to global health. A simplistic proposition that somehow collecting data for data’s sake will magically solve everything. The problem with this approach is that it forgets about the patient and the caregiver — and the incredibly complex web of behaviors that feed into disease and treatment. Technology, AI tools, and big data cannot be one size fits all. When they are, the value proposition is vastly reduced. AiCure is unique in its quest to track, interpret and impact human behavior. As a result, the National Institutes of Health and the World Health Organization are looking to the platform to be transformative; a sentiment shared by our investors and advisors including Baird, New Leaf, Pritzker Group, Tribeca Capital, and Bill Gates via Biomatics Capital.
Which players have you worked with that are excited vs. reticent about tools like this?
The CDC has listed AiCure as one of the go-to platforms for infectious disease control. The NIH has funded and supported AiCure. We are now working with small and large biopharmaceutical companies. And governmental organizations like LA County or the State of Illinois are using us for high risk population health management. We are in discussions with payers and providers and working with PBMs. The opportunities continue to build.
It is important to frame the applications of the technology to the stakeholder. Reticence to adoption may stem from concern regarding patient burden or a belief that the status quo in clinical research is acceptable. That is why the clinical evidence is so fundamentally important. In scientific and clinical communities, scientific evidence or what we call scientific truth cuts through the noise.
Smart adherence tools (pill bottles, etc.) have historically had some trouble proving wide scale success, how are you changing this?
Dr John Urquhart, a brilliant man and a pioneer in the adherence space passed away in March 2016. When I first met him, he spoke passionately about the need for quality research and understanding the dose-response relationship in drug development. He was adamant that the the electronic pill bottle — which he invented in the late 1970s — was never intended for wide-scale use or to increase adherence. He saw it as a simple tracking tool for use in clinical research. It was not designed to change behavior. There is a growing body of evidence showing supporting Urquhart’s thesis: electronic packaging is unreliable and ineffective at increasing adherence. Part of the reason is that patient monitoring, psychology, training, and intervention go far beyond a binary view as to whether a pill cap has been opened or closed.
At AiCure, we are working towards the automation of 30% to 40% of what a nurse does when they interact with a patient. The intelligence of the platform tackles areas of assistance, efficiency, and assessment.
How did your Wharton MBA prepare you to be a founder and CEO?
I think Wharton equips you with the vocabulary to understand a broad array of business challenges. As you build a business, the requirements to solve tactical, strategic, psychological, intellectual and emotional challenges are unrelenting and ever-changing. The degree touches upon a surprisingly broad remit and does actually equip you with a framework for solving problems. Call it: Wharton Think. The knowledge gained allowed me to have a universal dictionary across finance and business.
Wharton also reinforces the value of relationships. The Wharton network has been incredibly helpful to facilitate conversations that would have been difficult to access otherwise. It is also where I met one of the co-founders, Gordon Kessler.
Biographies of the interviewee and interviewer
Adam Hanina is co-founder and CEO of AiCure, a venture and NIH-funded company that has developed artificial intelligence to confirm medication ingestion by clinical trial participants and high-risk patients in real time. Mr. Hanina has been responsible for much of the vision and strategy at AiCure and is the primary inventor of over 20 patents. He is a passionate advocate for the use of healthcare software as a population health tool and has directed much of his previous work to this end. Prior to AiCure, Mr. Hanina contributed to the development of Cerner Corporation’s European strategy and was a Visiting Fellow of eHealth at Imperial College in London, UK. He has acted as a subject-matter expert on patient monitoring technologies for the National Institutes of Health (NIH). He holds an MBA from the Wharton School of Business.
Dhruv Vasishtha is a second year MBA student at Wharton in the Health Care Management department. His prior experiences have spanned management consulting, founding startups, product management, and advising high growth AI startups.