How to make use of your healthcare data? Just ask.
How to make use of your healthcare data? Just ask.5 February 2020 | 8min
Healthcare data is often siloed and 80% of it is unstructured (ie. can’t be easily stored in a traditional column-row database or spreadsheet, such as text and image files)
Making sense of how we can access and gain insights from this data is the next main challenge facing healthcare
For a healthcare AI company to be successful, it needs to bring together technology, domain expertise & data, and the tools it delivers have to make our lives easier
Healthcare data is key to discovering new insights to improve patient outcomes and efficiency. However, they are trapped in different data silos inaccessible to end users. What if you could just ask questions of your data and instantly get answers? – that is Huma.
We sat down with the cofounders of Huma, Greg Kostello and Dr. Lana Feng, to understand how they see the transformative potential of artificial intelligence (AI) and data in healthcare and what is needed to succeed in the healthcare space.
Breaking down the silos
HT: Tell me a little bit about the problem you are trying to solve in the market with your solution.
Lana: Healthcare data is everywhere, and it’s scattered all over the place. Finding data is like finding a needle in a haystack, let alone trying to find insights across these. What if you could just ask questions of your data using natural language and instantly get answers? That’s huma, We combine the simplicity of Alexa and the power of business intelligence.
We want to fundamentally change how people interact with data, just like Alexa. Domain experts such as scientists or clinicians; they know how to ask questions. Huma is so simple that they can use it – don’t have to know how to write code or understand structured query language (SQL). Most importantly, it doesn’t matter whether it’s unstructured or structured data, it could be Word, PDF or even free text. This is really powerful because 80% of our data in healthcare is unstructured.
Simplicity is key
HT: How does this work exactly?
Lana: It works by you asking a question, for example, ‘In this data set, show me cancer distribution?’. Huma immediately comes up with the cancer distribution in the data . It also suggests questions like, ‘What about lung cancer with secondary tumours?’ or ‘What about mutations?’ and ‘How many patients have epidermal growth factor receptor (EGFR) mutations?’ and ‘Show me EGFR mutation distribution across different cancer types?’.
Greg: You don’t have to be a data scientist, you just have to know human language and be a subject matter expert. It’s interesting because data scientists are often times hired because of their enormous skill but they end up writing reports or SQL queries, and now this frees them up to focus on the hard problems because all of these standard questions that the user wants is available to them through Huma.
HT: What unmet need did you set out to overcome with your start-up?
Lana: We want to enable users to make split-second decisions based on data.
The status quo is that when users want a specific question answered they go to a data scientist. Data scientists then get the data, crunch the numbers and come back with a report. The whole process takes several weeks, so it is highly inefficient. To make matters worse, data scientists typically don’t have healthcare domain knowledge, so the answer/report may not be exactly what clinicians or scientists want, so there is really a disconnect and we try to fill that gap.
Bridging tech and domain expertise
HT: Being a start-up entering the healthcare market, what do you see as the biggest challenge?
“For a healthcare AI company to be successful, you need three things; technology, domain expertise and data.”
Lana: Most of the AI companies trying to get into the healthcare industry purely come from tech. But, for a healthcare AI company to succeed, you need all three: technology, domain expertise and data. Five years ago, we saw a lot of healthcare AI start-up companies that failed because they ignored the domain expertise piece and were relying on technology to solve all the problems. Finally data really drives your machine learning platform – it is the new gold. There is few healthcare data in the public domain due to privacy protection, which is a good thing. But it does present challenges to AI companies.
HT: From your perspective, who are the people or the industry players that need to work together to really transform healthcare.
Lana: I think all the stakeholders. We came back from the HLTH Conference in Las Vegas – the point of that conference was to bring all the stakeholders in the healthcare innovation ecosystem together, from startups like us to the big tech giants like Google, Microsoft and Oracle, hospitals and health systems, the life sciences and pharma industries to the FDA to the consumer-patient side. You have to provide this kind of venue to bring everyone together.
Data will improve healthcare – the rate of change is less clear
HT: What do you think the healthcare industry will look like in 10 years?
Greg: So much is going to change in the next even five years, we probably won’t recognize where we are. There’s a lot of things that are happening concurrently. For example, the cloud is everywhere – a lot of companies are at least either partially in the cloud or majority in the cloud and taking advantage of that. Machine learning is becoming a commodity and is transforming how we work with data, in particular around language.
80% of data in healthcare is what is called unstructured – this is documents, PDFs, Word, PowerPoint, etc. Those are opaque to normal analytical systems, you just can’t see them because they want things in nice neat rows and columns. Suddenly with new tools you’ve transformed this and now you can work the way you want to work.
You don’t have to change your behaviour or the way you do your work, but the system will become smarter and things will become smarter and adapt to you. All those things are coming into play. Being able to deal with these huge amounts of data is effectively going to transform the industry.
Lana: I’m not as optimistic probably because I come from healthcare and healthcare never really warrants a radical change because it’s highly regulated and patient lives matter and so it’s always incremental. The last 10 years were about gathering data, like electronic health records (EHR). The next 10 years are going to be about how are we going to get the insights out of data? How are we going to use data better?
For example, in hospitals in the United States, you don’t see notes or physician scribbled prescriptions anymore, it’s now digitized in the electronic health care system. The next step is all about making these data interoperable and making sense of the data. Once we solve that problem, there are going to be a lot of transformative changes. For example, on the life sciences side, the accessibility and insights out of data could really drive acceleration on drug developments. That will increase accessibility of life-saving drugs to patients and could also reduce drug prices, which could be hugely advantageous. On the hospital side, you can diagnose patients a lot faster.
Another side of using this data is being able to predict the best treatment for a patient. From an epidemiology standpoint and even patient progression standpoint, being able to anticipate is a huge thing.
Lastly from a regulatory standpoint, the FDA is also addressing this data interoperability problem. Once they do that, they could approve drugs faster. Then all of a sudden, this whole ecosystem is going to be quite transformative for healthcare, but it’s going to be incremental.
Greg: Now you see the different opinions. Lana comes from the industry and I’m from tech, and the truth probably lies somewhere in the middle. It’s very hard to see how things are going to transform the industry at the beginning. Everything sort of looks the same but over time we look at tools on the consumer side like Siri and others, suddenly these things are part of our lives.
A few years ago, you wouldn’t imagine walking and talking to a machine, and now it’s everywhere. So, it’s hard to predict where these changes are going to happen. I agree with Lana, we have to be really careful. This is not like advertising, these are lives at stake so we have to be really careful.
Where I see the first steps is helping you do the job you already do even better. Think how hard it is to find information these days, and if you just had a tool that made it easier to find information, how much more effective would you be at your job? Just that small step could have a really big impact.
The more data the better
HT: Is there something that healthcare leaders could do today to kickstart this further?
Greg: The more data sets we have access to in the system, the more complex the questions we can have that can be answered. A lot of the times it’s getting other groups to think about what are the questions that we can ask the system. Once we get that buy in, it goes faster and faster from that point forward
Humans first, plus machines
HT: Why did you choose Huma.ai as the name of your company?
Greg: Huma.ai stands for Humans plus Machines, where humans always come first. We always wanted to take humans into account for any machine learning we did.
Lana: Many tech companies start with the data, let data tell the story, and we say, ‘No, there’s tons of knowledge in our brains, why don’t we start there’? So, we start there, with human intelligence and bring that to the machine intelligence, from a user perspective.
Greg: It goes back to my days starting at Apple where we really tried to focus on how do we make technology accessible to the average person? At the time when I started, it was really a tool for techies. Computers were technical things, and now computers are part of everyday life.
What we noticed was that information and systems were getting more complex and our lives were getting busier and busier, so where are the tools to make our lives easier, not harder. Easy to access, easy to use, no barriers to entry.
Huma.AI successfully participated in Startup Creasphere, a leading digital health accelerator that strives to transform healthcare together with startups.
Lana Feng, PhD is the cofounder and President of Huma.AI. She has over 20 years of experience in biotech and pharmaceutical industries, mostly in precision medicine. Dr. Feng came from Novartis Oncology Business Unit where she established international partnerships for their precision medicine programs.
Greg Kostello is the cofounder and CEO of Huma.AI. He “loves building products that other people love using” and he brings over 35 years’ experience in executive, technology and development management. Greg is a tech veteran who has developed consumer products at Apple and Netscape and built large SaaS platforms used by NASA, Cisco and ESPN.