Launching an AI start-up for Healthcare

Welcome back to our interview series on all things start-ups. Today we discuss the use of Artificial Intelligence (AI) in Healthcare. We delve into the past, present, and future of AI's role in revolutionizing healthcare, how generative AI changes the landscape, and what founders should think about in entering the healthcare field.

Technology in healthcare

Shayani:

With all the buzz around AI in healthcare, I’m excited to hear your perspective given your unique background as a founder, executive, and advisor to startups. Is AI new to healthcare?

Lusi:

AI has been present in healthcare for a while.

It's used in personalized health, like personalizing health and wellness in B2C plays. Hospitals employ it in various aspects, with radiology leading the way, and now also in pathology and clinical triage. AI is also involved in back-office tasks, such as electronic medical records and health insurance. It's been applied in life sciences, impacting drug discovery, clinical trials, and personalized medicine.

As more data becomes available, AI's usefulness in healthcare grows.

Shayani:

AI is often associated with organizing information, pattern recognition, prediction, and workflow improvements. Are there applications of generative AI, like GPT-3, in healthcare?

Lusi

Generative AI can be useful in organizing unstructured information. It helps us move from transcription to summary and understanding, which is new and adds on top of that the ability to spot trends more easily than a human can.

For instance, some companies use it to extract insights from patient records to assist physicians and nurses during patient interactions. This helps maintain a personalized touch while having essential records at hand.

That's always been kind of an age-old question of balancing record keeping with patient engagement and physician time. I'm sure we've all been to medical appointments where the doctor is looking at the computer the whole time rather than looking at you. We all want that one-to-one interaction, but at the same time, we also want the records there. I just went to a physical therapy appointment this morning, and he reminded me that I had some work done on my right ankle five years ago that I didn't even remember. So I am thankful that there are records in there that help them understand not just a point solution, but a holistic picture. Some of the generative AI solutions I've seen are then taking that record to respond to patient prompts, about past visits, prescriptions, etc.

Shayani:

So in essence, generative AI helps create more signals in the data than previously available. I remember when iOS came out with its more strict privacy settings, but Facebook countered by creating more signal from the noise to continue to target audiences without the information previously collected by iOS and doing a better job at it.

Are there specific areas in healthcare where AI can be extremely valuable and where its potential hasn't been fully tapped yet?

Lusi

Imaging analysis is a well-explored area, but there's always room for improvement, especially in combining multi-modal information. Currently, your EMR data, your images, your genomics, and your labs are all siloed. Some systems allow you to analyze one or the other, but not many that look at the whole picture.

Founders should also look into addressing hidden, labor-intensive, error-prone processes across healthcare workflows. This includes back-office tasks, billing, electronic medical records, and more. AI can streamline these processes, allowing humans to focus on valuable patient interactions.

Shayani:

In a world where we’re focused on productivity, I like that founders should also look into solutions that can help reduce errors, which can lead to life-or-death decisions.

Given the complexity of healthcare, what should founders consider before diving into this space?

Lusi:

Founders should understand that healthcare is NOT a short-term endeavor; it takes time. Healthcare is mostly likely an enterprise B2B solution versus B2C solutions. You might be able to hack together an app and have an MVP out in a couple of days for a consumer product, but that's not going to be the case for healthcare.

There are 4 considerations:

1. Regulatory approval and reimbursement processes can be lengthy and challenging. You can shortcut FDA clearances by finding like solutions and going through a 510K route. You can also shortcut the reimbursement process by creating workflow solutions that pay for themselves in the ROI saved/gained, as opposed to at a procedure level which needs to be submitted and approved by every insurance carrier.

2. Data access is another hurdle; healthcare data is valuable and tightly controlled due to privacy. Hospitals will not let any start-ups access their data, so you need to figure out what’s your data advantage to get started, and how to keep that once you’re in. We all know that AI is only as good as its training set, how can you get a large amount of representative data to make your AI useful. (For more reading about healthcare data, here's a previous blog interview with Su Huang at Datavant)

3. Integration with existing systems, like electronic medical records, can be complex. Epic is the big gorilla in the field and they don't make it super easy for startups to integrate with them.

4. Selling a SaaMD (software as a medical device) is different than traditional medical device sales. In some ways, it’s easier, as you don’t need trunk stock and a lot of feet on the ground, so you’re able to distribute without relying on large channels. However, you add a whole new stakeholder/decision maker into the mix that you need to convince, which is the IT buyer, in terms of security, integration, etc.

Passion for healthcare is crucial. Success in this space takes time, and you need dedication. It's not ideal for first-time founders unless they already have extensive healthcare networks and experience.

Consider gaining experience in other startup areas before tackling healthcare. Maybe get your startup chops doing a B2C startup, learn things such as product-market fit, MVPs, rapid testing, etc., and then take those learnings to healthcare. I love to see founders get trained elsewhere, and then really have a passion for healthcare and the longevity to stick with it before they enter.

Shayani:

How can founders validate their healthcare AI products effectively given the long cycle to market?

Lusi:

This is where it's really important for a problem to be looking for a solution rather than a solution looking for a problem. Just because you're not going to have as many “at bats” to use your solution to find a problem. Observe what is the pain point. Does it need a painkiller or a vitamin? Why is it painful?

Also really understanding the workflow what's being done now? Fully understanding who are all the stakeholders, why does something take so long? What systems are involved and understanding the context can go a long way.

Essentially start with thorough market research. Understand the pain points, workflows, and stakeholders involved.

Then, I’d try to get into usable pilot programs quickly, perhaps with existing institutions where you have access, even if it’s small and not statistically significant, getting doctors, nurses, patients, and techs interacting with your solution is much more valuable than whiteboarding in your head. Don't worry about immediate revenue; focus on testing, measuring, and demonstrating impact.

Fortunately, many hospitals and pharma companies do host incubators where they define the problems they’re having and allow start-ups fast-track access to be able to pilot on a limited scale. Look into joining one of those programs.

Shayani:

I spoke to a new founder recently, and they were telling me that they are swimming in advisors for all sorts of things, and not all of them are helpful. What types of advisors should founders have to increase their chances of success in healthcare?

Lusi:

The type of advisor you look for should match your target audience. If you're selling to consumers, seek advisors with B2C experience. If you're selling to regulated environments, find advisors with regulatory expertise. Domain expertise in the specific healthcare sector you're targeting is invaluable. If you’re selling to hospitals and clinics, find someone who’s done that before. If you’re selling to pharma companies, find someone who’s done that. Advisors can provide valuable guidance and help navigate the complex healthcare landscape.

I previously wrote a four-part blog about the different target audiences for healthcare start-ups here.

Advisors are different than your first full-time salesperson as we discussed in our previous blog.

Shayani:

Would you be open to advising healthcare startups?

Lusi:

I'm passionate about working with startups and helping them understand the challenges and opportunities in healthcare. It's a field with immense potential to improve patient lives, and I'm excited to be part of that journey. Feel free to reach out!

About:

Shayani and Lusi are two of the partners of The Thirteens, a start-up consultancy of fractional executives covering multiple functions and industries, super charging your start-up to the next level.