Healthcare AI Startups: Key Challenges and Opportunities

AI Innovations in the healthcare industry are rising exponentially, as the market presents huge opportunities. However, healthcare is a complex marketplace, and while technological innovations are important, they are not sufficient to create a market.

To succeed commercially, startups must clearly understand the needs of their key stakeholders—that is, the investors and early customers. Below are some key considerations the company should focus on to differentiate itself.

Define your market size

Investors want to know the size of your market. This helps them decide whether or not it is their game. Yes, US healthcare is a big market—$4.5 trillion (2023)—but it is divided into numerous segments and sub-segments.

Hospital care, clinical and physician services, and prescription drugs account for a significant portion of this expenditure. As an example, if your AI product is focused on the home health segment, it comprises roughly ~3% i.e. ~$133 billion. Still a huge market, but then almost 85% of this market is focused on services such as nursing, etc. That leaves ~15% for medical devices and other related expenses. Deconstructing the market to the right category may lead to a smaller opportunity but it also clearly demonstrates what is the real market available for the given solution. Investors appreciate this level of granularity. This data also allows the company to stay focused on its initial target market and drive market adoption.

Whose problem are you solving?

To be successful, you need to position your solution correctly so that it aligns with the customer's pain points and priorities. Progressive health systems want to work with innovative companies to bring new solutions. However, it is important to do your homework before presenting your solution.

Health System executives are constantly bombarded by startups. A senior healthcare executive once shared his frustration that he was surprised by how a significant number of vendors often don’t do their homework before pitching their solution.

While it may be tempting to go for a broader audience, invest time in carefully segmenting and sub-segmenting your target customers. If you are selling a solution for cardiac monitoring, it makes sense to first go after those health systems whose priority is cardiology and who are investing resources in this area. This information gathering requires time and resources, however, it also hugely improves the odds of a successful customer engagement and eventually a sale.

How your solution integrates within the existing workflow

Healthcare is complex. One of the biggest challenges is the status quo. Technology can be great; however, if it requires a major change in workflow, its adoption will likely be difficult and slow at best. AI Startups should ensure that their solutions easily integrate with the existing workflow. They should also clearly define how this integration will be accomplished and what the hospital and vendor's responsibilities are. This upfront clarity helps health systems plan and allocate appropriate resources to ensure successful integration.

Post-integration monitoring is a new requirement for AI tools

AI is poised to change healthcare, however, there is a lot more monitoring that is required for AI-based solutions. Unlike an approved device or drug whose performance does not change over time, the performance of an AI-based tool is bound to change over time due to a variety of local factors. This phenomenon is called ‘Drift’.

A drift is a degradation of the model performance over time as it encounters real-world data that is different from the data it was trained on. A drift can significantly impact the model’s accuracy and usefulness.

For AI-based solutions, constant post-market monitoring is required to evaluate their performance. However, it is unclear who should be responsible for this. This requires additional costs, which the health system may not be willing to incur. Vendors must find a collaborative way and set up a process with the health systems to ensure their models can be updated as needed to meet their performance benchmark.

Define your business model

One of the most important barriers to adoption is – who will pay for it. If an AI solution does not have a way to make money directly or bring significant savings indirectly, there will not be a taker. For an AI- solution that is embedded in the medical device, it is easier to factor in the cost as part of capital equipment, however, for an independent AI-decision support system this can be a challenging and long process.

As a startup, you must consider if you are going for the longer path to reimbursement that includes clinical validation, FDA clearance/approval, reimbursement pathway selection, and market access strategy, or once you get the regulatory approval, you would like to sell the solution to an established organization who has the expertise to take the technology to the market.

You still need awareness and education to drive adoption

Even when you have crossed all these milestones and have an AI application with a CPT code, the adoption can still be a challenge. According to a recent report published in NEJM AI (https://ai.nejm.org/doi/10.1056/AIoa2300030), medical AI device adoption is very slow, limited to a few privileged demographics, and mostly driven by a small number of applications. There is a need to provide more education and equitable access to increase its adoption.

To conclude, these are some of the high-level points, that your leadership team should define and regularly refine as the technology, competitive, and regulatory landscapes are changing at a faster pace. Despite these challenges, it is a highly rewarding path, especially when you can bring a solution that addresses an unmet need and helps all stakeholders including the patients achieve better outcomes.