Who Should Control Patient Data — and Why Has It Never Been Patients?
Shashi Shankar never wanted to be an entrepreneur. After nearly a decade at Genentech and Roche, he was comfortable in pharma — so comfortable that when he kept encountering “API” in technology research, he assumed it meant Active Pharmaceutical Ingredient.
What changed was personal. His grandfather was diagnosed with late-stage gastroesophageal cancer and died in four months. Watching him switch doctors every one to two weeks, accumulating CTs, medications, and lab results with no way to consolidate or understand them, planted a seed. The professional frustration was just as deep: pharma companies were spending enormous sums on real-world data that couldn’t follow a single patient through their treatment journey.
The data landscape is broken in specific ways.
Claims data from companies like IQVIA is broad but shallow — ICD-10 codes, ambiguous patient counts, heavily anonymized. Single-site EMR data from Flatiron is deeper but narrow, capturing only what happens at one care setting. Health information exchanges sound promising on paper but deliver, as Shankar put it, “a subset of a subset of a subset.”
Consider a breast cancer patient — call her Patient Patty — who starts at a community OB/GYN, gets referred to an academic specialist, undergoes biomarker testing at a third facility, receives treatment at a fourth, and still sees her PCP and occasional urgent care. No single data source captures this journey.
Novellia’s approach is deliberately different.
Rather than buying data from intermediaries, they work directly with patients who have a legal right — established by the 21st Century Cures Act — to collect and consolidate their own records. Patients connect multiple patient portals via SMART on FHIR, and Novellia’s AI pipeline normalizes, deduplicates, and harmonizes the resulting data.
“We’re only gonna work with patients,” Shankar told us. “We’re not gonna have smoke-filled room handshakes with data brokers to somehow acquire massive datasets that completely cut the patient out.”
The strategic focus is counterintuitive.
Previous PHR companies — Microsoft HealthVault, Google Health, and many others — tried to build for everyone and failed. Novellia targets the skinny tail of the health bell curve: the most complex, seriously ill patients. These patients are the most motivated to consolidate their records, willing to co-create the product, and generate the richest data. They connect 3-5 portals going back 20 years.
What “engagement” means here is also different.
VCs obsess over daily active users and vanity metrics. For health data, engagement means completeness — and that reframing matters. “When it comes to health, data engagement means something different at the most fundamental level,” Shankar said. “The completeness of that data and the value of that data, both to patients and to pharma… not engagement for vanity metric’s sake.”
The AI layer catches real problems.
In one research project, the system detected a lab value in structured EMR data that didn’t match the patient’s condition. When it consulted the raw XML — unstructured data that patients themselves often can’t see on MyChart — the actual value was off by a single digit. A physician typo that would have moved silently through any other data pipeline.
Shankar reserved his sharpest words for Big Tech’s push into healthcare. “When you have organizations that spent $40 million on dueling Super Bowl ads making fun of one another… and some of these organizations are now talking about dismantling safety protocols — are those really the organizations we should trust with the most sacred data of all, which is our health data?”
It’s a fair question. And one worth asking before, not after, something goes wrong.