Dr. Wendy Nilsen knows precisely what smart health should look like, and in her eyes, this isn’t it.
Despite billions spent on healthcare innovation and endless talk about AI, the essential metrics (heart disease, diabetes, stroke) have barely budged.
As NSF’s Deputy Division Director for Information and Intelligent Systems, Nilsen asked the room at The Katz School of Science and Health at Yeshiva University this week, Is smart health smart enough yet?
Because if sophisticated sensors, advanced analytics, and AI-driven data aren’t actively reshaping real-world outcomes already, they’re failing.
Technology's potential isn't Nilsen’s question, its meaningful application is. Incremental progress isn’t sufficient, she argues, and real transformation takes a full rethink how healthcare gathers, integrates, and uses data.
Breaking Health Data Silos
Nilsen’s point is that healthcare doesn’t lack data, but what it has drowns in fragmented pools.
Hospitals overflow with isolated data islands (pathology reports, radiology images, wearable output) that clinicians navigate manually, leaving critical insight hard to spot. Nilsen notes the irony that AI should be seamlessly fusing these datasets, yet often remains another disconnected data island.
Take anesthesiology, for instance. Real-time decisions can be life-or-death. Nilsen points to an NSF-backed surgical control tower that fused scattered patient data (vitals, medication doses, surgical timelines) into one analytical cockpit.
This integrated approach sped up decisions and measurably improved patient outcomes. Yet, she says, most healthcare facilities remain stuck in fragmented frameworks, missing this clarity.
Nilsen pushes the industry to look beyond traditional markers (glucose, cholesterol reads for example) to biological signals healthcare habitually overlooks.
Cytokines, molecular inflammation markers found in numerous chronic diseases, are often dismissed as too complex for real-time measurement. Similarly, she says, insulin remains elusive for continuous monitoring, despite its central role in metabolic health. Nilsen insists these markers aren't impossible to work with, just overlooked.
Further, she adds, nano-sensors, minimally invasive wearables, and implantables are no longer futuristic concepts, they’re just still awaiting adoption. Which makes no sense since expanding sensing capabilities beyond easy metrics can be so important.
Real-World Data Is Everywhere. In the Real World.
When healthcare discusses real-world data, it too often means patient-generated data (think Fitbits and health apps) ignoring an infinitely richer data universe surrounding everyday life.
True real-world health data encompasses passive streams from environments, homes, cars, and daily behaviors, far surpassing intentional patient logs, she explains, pointing to how consumer industries use IoT so seamlessly already.
Nilsen points to how cars can monitor cognitive decline through subtle driving patterns, homes infer wellbeing from occupancy data, and nevermind the FitBit or Apple Watch, commercial data revealing dietary behaviors far more accurately than self-reports.
Yet, she says, healthcare rarely taps these reservoirs, despite obvious predictive potential.
And really, all of those more patient-oriented devices point to another topic Nilsen dug into. Healthcare systems routinely overlook essential contextual variables. Think of things like indoor temperatures during heat waves, social environments around dietary choices. These are all subtle data points that can shape patient outcomes if we only intelligently tracked and implemented the insight into a broader platform.
Otherwise, we’re back AI in isolated data pools, floating unanchored, limiting clinical understanding.
Slashing the Administrative 30%
But let’s come back to the medical enterprise itself for a moment. Because Nilsen makes some points about the future of healthcare as a business and where it’s ripe for innovation.
Healthcare’s most stubborn figure, 30% administrative costs, is a glaring inefficiency AI is uniquely suited to address.
Nilsen argues for automating repetitive tasks currently burdening clinicians (scheduling, documentation, patient intake) that consume precious clinical time.
She points to hospital scheduling platforms, notably at Mayo Clinic, that demonstrate AI’s transformative operational potential. They’re able to optimize patient flows, clinician workloads, and dramatically reduce wait times. She emphasizes that every minute clinicians spend on paperwork and archaic systems detracts from direct patient care.
Securing AI’s Healthcare Revolution
While healthcare races towards ambitious AI transformations, it remains alarmingly complacent about privacy and security. Nilsen says current cybersecurity frameworks, designed for simpler data environments, are dangerously inadequate in today’s complex, multimodal AI healthcare ecosystems.
Every connected device, sensor, and analytic layer expands vulnerability. Robust, AI-native security and privacy frameworks have to keep pace with new tech adoption or healthcare AI risks devastating breaches and irrevocable patient trust loss.
Validating Continuous AI
And there’s never a conversation about medical tech that doesn’t invoke regulation.
Current frameworks treat AI models as static devices, yet modern AI continuously evolves. Nilsen notes the challenge and asks how regulators will validate a system that constantly learns, refines, or potentially degrades?
Her solution? Continuous, real-time validation frameworks that audit, benchmark, and ensure transparency.
Reinventing Healthcare’s Architecture
Incremental steps frustrate Nilsen. Real healthcare transformation demands radically integrated, context-aware, secure, and continuously validated AI architectures.
Her vision calls for breaking data silos, capturing neglected biological signals, leveraging expansive real-world data, embedding critical contextual insights, automating inefficiencies, and rigorously validating evolving AI.
The whole conversation cuts sharply through complacency. Nilsen makes it clear that truly smart healthcare requires fearless reinvention and incremental evolution won't suffice.
To support the future outlined here, healthcare organizations need an architecture aligns seamlessly with this vision of integrated, context-rich, dynamically validated AI systems.
VAST breaks down healthcare’s persistent silos, delivering real-time, multimodal data fusion at unprecedented scale and speed. Its approach natively supports continuous validation and security, ensuring trustworthiness at the data infrastructure level itself.