While we’ve examined the issues that are keeping the majority of buyers from immediately adopting healthcare AI solutions, there are still pockets of potential buyers who could become the early adopters. These buyers can see an immediate ROI and don’t need to wait for additional evidence or reimbursement.
One such group consists of independent radiology groups and teleradiology businesses, which can differentiate themselves by improving their output in terms of quality and speed. If the AI algorithms can make them more accurate when reading scans and add more quantification, such as the progression of a tumor over time, it will allow them to make their end-product of a higher quality than that of their competitors. Also, providers value the speed with which they receive these reports. If providers can get reports faster and offer more expedient care to their patients, they’ll most likely direct more business to the radiology group that offers the fastest reports.
The diagnostic imaging provider RadNet, which maintains a U.S. network of 350 outpatient imaging centers, has its own subsidiary that develops AI solutions that can improve the quality and speed of the radiology workflow. They’ve also secured two software clearances from the FDA to help radiologists to spot breast and prostate cancer.15 This highlights the fact that if you work for a healthcare AI company that’s developing radiology algorithms, RadNet could be an ideal customer thanks to its radiology centers around the country. However, if your potential customers are developing and gaining approval for their own algorithms, how can you stay in business? Can you be accurate with your forecasts if you don’t know which target customers are developing their own solutions?
For teleradiology businesses, their value proposition is that they can quickly and accurately read scans. If these tools allow them to better triage and to contact medical centers for the acute cases that need immediate attention, they can become a more highly valued partner. Teleradiology vendors often provide evening and weekend coverage. This means that there are fewer resources in hospitals to take care of acutely ill patients. The timely mobilization of those resources could save lives and create a less chaotic environment. Because of that, the ability to quickly identify those cases and to notify providers will be of high value to those providers.
Academic Medical Centers (AMCs) could provide another important group of early customers. AMCs usually handle large volumes of patients as they’re often the destination for the uninsured and underinsured. They’re trauma centers which receive complicated patients sent to them from other centers. If AI technologies are showing promise as assistive devices to improve or hasten diagnosis, to make their clinical staff more productive and to improve operations, they’re more likely to be first- movers. If they see an immediate benefit, they’ll work with the solution provider to generate the real- world evidence that can accelerate commercialization of that solution (probably for a discount!)
Of course, these medical centers can develop their own models. They have plenty of data and the resources to use that data to develop high-quality models. So, what’s stopping them? In a lot of cases, nothing! Some are busy developing models for various clinical and administrative issues and gaining experience. Some are trying to commercialize their models to other providers, although they’re in the minority. Most will use models that are developed by outside entities since they don’t have the bandwidth to fully execute them on their own. My discussions have shown that most medical centers say that the key benefit of externally developed models is that they save time. Building a model from scratch takes a lot of time and effort, and so finding a relevant model from an outside vendor and then training it on local data is a much faster option.
My interviews with AMC leaders have suggested that they’ll continue to invest in AI and to learn from their experiences. At the moment, most of them are using hybrid models which combine self-developed algorithms and commercially available algorithms from industry vendors. This approach allows them to focus on key niches in healthcare through their own algorithms while simultaneously using the commercial algorithms to tackle broader opportunities. Because of that, they’re able to harness their own internal expertise as well as the wider expertise of the industry.