Sean Cassidy, CEO and cofounder of Lucem Health, with his wife Lynette, son Carter and daughter Keira
Photo: Sean Cassidy
Artificial intelligence is being applied to an increasing number of areas within healthcare. And many hospitals and health systems are finding early successes.
But what about applications that have not yet made major inroads in healthcare? For example, can health systems use AI to fix specialty margins and enrich the mix of patients in high-value service lines? Or, can AI-driven patient identification and proactive outreach flip the referral dynamic? And can shared-risk AI models replace some traditional software sales and potentially reshape some aspects of the vendor-provider relationship?
Sean Cassidy has some thoughts on these questions. He is CEO and cofounder of Lucem Health, a healthcare technology company – founded in partnership with Mayo Clinic Platform – that develops AI-powered systems for early disease detection, finding undertreated patients, and identifying high-risk patients before diseases become life-threatening.
Healthcare IT News sat down with Cassidy to discuss the aforementioned questions and more.
Q. How can health systems use AI to fix specialty margins and enrich the mix of patients entering cardiology, hepatology, GI and other high-value lines?
A. We're learning that many health system specialty departments are flooded with low-acuity patients that could have and should have been treated in primary care. These patients take up high-value exam room time and cause understandably frustrated specialists to practice below the tops of their licenses.
Specialists and the unit economics of their service lines could benefit greatly from an upstream process that proactively and programmatically routes patients into care pathways tailored to disease stage, progression and acuity.
By surfacing hidden risk signals in electronic health record data, AI can "front-end" disease-specific programs that ultimately lead to earlier diagnosis and treatment. Such programs would feature proactive patient identification, engagement that motivates action, accelerated screening and diagnosis, and optimized treatment pathways that take individual patient circumstances and specialty care access constraints into account.
This higher yield, higher impact approach to care delivery would increase specialty care margins by enabling specialists to focus their attention on the sickest patients – in other words, patients who require the most complex and expensive care.
A more proactive approach powered by AI wouldn't just deliver more service line revenue in a fee-for-service context. It could also deliver meaningful, long-term savings for patient populations subject to fully capitated contracts. In value-based arrangements, health systems must optimize and hopefully minimize utilization across all their services lines.
Earlier diagnosis and treatment ultimately uses fewer healthcare resources than current practice: the evidence for that is overwhelming. Advances in AI now make it much easier to capture that benefit at scale.
Q. How can AI-driven patient identification and proactive outreach flip the referral dynamic, and what does that mean for relationships between PCPs and specialists?
A. Our current reactive approach to healthcare mostly waits for patients with undiagnosed diseases to become uncomfortable or alarmed enough to pick up the phone and schedule appointments. Even when healthcare tries to be proactive – for example, with colorectal or breast cancer screenings, and annual physicals and checkups – patients often don't respond to proactive engagement from their providers.
This leads to more complex and invasive treatments, worse patient outcomes, and greater physical and emotional distress. And, because diseases that progress to later stages are often harder to diagnose and treat, they can require more touchpoints and more coordination of care.
In an ideal world, we would eliminate or sharply reduce the number of interactions and modalities patients would have to navigate to get positive outcomes that improve their lives. That world would feature a great deal more certainty for patients and their primary care physicians that they are navigating to settings of care – for example, specialists – that can make or confirm their diagnoses and accelerate their treatments.
Disease-specific patient identification, outreach and diagnosis that sits between primary care and specialty care is, by design, proactive. It creates, in effect, a new kind of referral pathway – one that is initiated and powered by AI and has tailored workflows that route patients to the appropriate care setting for follow up.
This model dramatically reduces the "low-acuity patients in specialty care" problem and facilitates a more harmonious relationship between primary and specialty care because it takes maximum advantage of their respective and unique remits. And, because it features diagnosis at earlier stages, it can reduce treatment and "primary to specialty care navigation" complexity.
Q. In 2026, you suggest shared-risk AI models might replace some traditional software sales and potentially reshape vendor-provider relationships. How?
A. Healthcare technology economic models are changing rapidly. For decades, vendors have sold software via usage or volume-based licenses and subscriptions. These models ask providers to absorb 100% of the risk related to whether they actually get value from the products they purchase. The next wave of AI-powered digital health products will upend those models.
Instead of paying for access or usage – regardless of whether they get value commensurate with that payment – health systems will insist on value-oriented partnerships with their vendors in which the price they pay is tied explicitly to measurable clinical or financial benefits – or both – that are objectively enabled by the vendors' offerings.
This approach puts the onus on software suppliers to demonstrate value through performance. It also requires them to pay close and constant attention to customer success.
Aligning incentives in this way builds accountability and shifts the focus from feature checklists to results such as clinical yield, throughput, cost savings or margin impact. Vendors become vested and embedded rather than just software suppliers.
Those that are unable to demonstrate clear, repeatable results will fade quickly. Those that deliver quantifiable value will establish long-term relationships grounded in transparency and trust.
Shared-risk commercial models also lower innovative technology adoption barriers for health systems. Innovation now can be funded directly from incremental revenue or savings as it hits the P&L, rather than as a substantial up-front capital expense.
By the end of 2026, shared-risk relationships between health tech vendors and their customers will become the standard. Vendors that embrace this model and deliver compelling value will survive and thrive. Those that don't will be left behind.
Follow Bill's health IT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.
WATCH NOW: In 2026, AI capabilities will boost RPM, virtual care expert says
