Jan 17, 2019 · 35m · a16z
a16z Podcast | Dark Data in Healthcare
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the a16z Podcast, experts Susanna Fox, Anil Sethi, and Vijay Pande discuss how unlocking 'dark data' in healthcare empowers patients, removes operational friction, and transforms clinical trial matching and public health outcomes.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 24.5% of the talking time here. How this is scored →
The host as informed peer 3.9 Guest teaching 3.7 Guest disagreement 1.6 The host pushing back 2.6
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 33:33 Susanna Fox rejects Vijay Pande's stance on Dr. Google
Susanna directly challenges Vijay's assertion that patients need to ship data off to real doctors, forcefully defending patient expertise and self-education.
Hardest push from the host ▶ 2:06 Sonal Chokshi challenges consumer interest in health dataSonal explicitly rejects the premise that patients want control over their health data, using a car registration comparison to press the guests on practical utility.
Biggest teaching moment ▶ 7:48 Susanna Fox corrects Sonal Chokshi on HIPAA definitionSusanna corrects Sonal's assumption that HIPAA is purely a restrictive privacy barrier by pointing out that 'P' stands for Portability, prompting Sonal to acknowledge she had no idea.
The host holds their own ▶ 20:49 Sonal Chokshi introduces permissionless innovation modelSonal leverages her background in tech history to frame data portability as permissionless innovation built on top of internet platform infrastructure.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| The Origin and Problem of Dark Data | 3 | 4 | 1 | 6 | Sonal challenges the premise that everyday consumers actually want or need access to their health data, comparing it to basic vehicle maintenance tasks. Vijay and Susanna reframe her skepticism by pointing out how dashboard alerts and diagnosis triggers drive consumer action. | |
| Generational Shifts, Oral Histories, and System Friction | 4 | 2 | 1 | 3 | Sonal shares an insightful observation about how oral histories in immigrant families lead to missing context in doctor visits. The guests agree with her framing and discuss lowering friction in data transfer. | |
| Demystifying HIPAA: Portability and Patient Rights | 4 | 7 | 1 | 2 | Susanna corrects Sonal's belief that HIPAA is primarily a restrictive privacy law, revealing that 'P' stands for Portability. Sonal admits her lack of knowledge before offering a clear conceptual model of the patient at the center of the ecosystem. | |
| Operationalizing Deep Clinical Data: Tanya's Story | 3 | 5 | 1 | 5 | Sonal questions the value of health data by drawing an analogy to passive wearable data and asking why oncologists wouldn't already share records. Anil educates the room with a deep personal story about his sister's care across 14 facilities. | |
| Three Friction Points: Requests, Refineries, and APIs | 3 | 3 | 1 | 1 | Anil and Vijay break down the three primary friction points in health data processing. Sonal facilitates the discussion by summarizing key technical transitions like converting unstructured documents to structured data. | |
| Macro-Level Impacts and Permissionless Innovation | 7 | 2 | 1 | 1 | Sonal demonstrates strong host expertise by connecting patient data control to computing history and the concept of permissionless innovation. The guests enthusiastically validate her framework. | |
| Public Health Surveillance and the Opioid Crisis | 3 | 3 | 1 | 2 | The group explores how unstructured dark data like death certificates can serve as an early warning system for public health crises. Sonal prompts Susanna to explain how payer reimbursement models contributed to the opioid epidemic. | |
| Longitudinal Baselines and Clinical Trial Matching | 4 | 3 | 1 | 1 | Vijay illustrates the power of longitudinal baselines through Ben Stiller's prostate cancer diagnosis, which Sonal effectively compares to pediatric growth curves. Anil adds how patient health summaries can automate clinical trial matching. | |
| Small Data at the Edges and Peer-to-Peer Healthcare | 4 | 4 | 6 | 2 | Susanna directly pushes back on Vijay's dismissal of 'Dr. Google', asserting that empowered patients hold genuine expertise and should operate at the top of their license. Sonal ties the debate back to community platforms like PatientsLikeMe. |
Statements from this episode (21)
Fox: Healthcare institutions operate believing they own patient clinical data
“We're in this position because most of healthcare has grown up believing that they own the data. So professionals generally collect and hold the clinical data that drives a lot of clinical care.”
Pande: Electronic medical records are basically just digitized paper doctor notes
“We have electronic medical records now, but it's really not radically different. It's basically a doctor writing or provider taking down some notes.”
Fox: Most people avoid confronting their personal medical data
“Most people don't want to engage in their health. Most people don't want to know what's going on. They're not necessarily ready to look into the mirror of data. I call it standing naked in front of the mirror of data.”
Fox: Personal health crises trigger immediate patient engagement with medical data
“But then when something happens, you're diagnosed with something, or your child or your mom is diagnosed with something, That will trigger engagement, let me tell you, and then people should have immediate access to the data and information that they need to m…”
Sethi: Retiring boomers will be far more engaged in their healthcare
“So people are going to retire and become more engaged than their parents were in their own health care.”
Pande: Patient data sharing must drop below one second to be useful
“If you can bring the friction from 30 minutes to a third of a second, that's when the doctor now is actually able to use that information.”
Sethi: Patients can legally demand their medical records via API or email
“Patients are actually outside of HIPAA, so you can ask, as a patient, a HIPAA-qualified entity to do something for you and share your information, whether it's through an API or whether it's through a PDF or email, and they have to do it.”
Pande: Empowering patients with data creates a massive healthcare startup opportunity
“Once you realize that the patient's at the center of it, you realize there's a huge opportunity there. If companies or startups empower the patient, then the patient can drive this whole thing.”
Pande: Wearable health metrics like 10,000 steps lack clinical value
“In the case of wearables, you're dealing with something that's very broad and very shallow, and also often not very clinical, like 10,000 steps versus 9000 versus 11,000, that's largely just made up.”
Sethi: Apple Health Records is built on acquired Glimpse technology
“Tanya used Glimpse, which is a former company. Apple acquired it in a transaction, and some of the health records that we see them releasing it to the world, Is some of that technology.”
Sethi: Health data products must target highly self-incentivized patient groups
“We actually think in order to operationalize health data, you have to find self-incentred people, and that is cancer patients, lupus, HIV, autoimmune disease.”
Sethi: Most released medical data is unstructured documents, not coded data
“Most of the information that's released is not in any coded form. It's documents and PDFs and XML files.”
Sethi: Cancer pathology data remains inaccessible through standard electronic health APIs
“A lot of the cancer information is in the pathology report which doesn't come out of APIs yet.”
Pande: Real-world evidence will surpass clinical trials in utility
“And what I think we'll start to see is, especially as new statistical methods come online, that real world evidence will be very useful, in some ways will be more useful than what you can do in a clinical trial, both due to power and due to the fact of real li…”
Pande: The commercial hurdle for new drugs is reimbursement, not FDA approval
“In a world where things have gone past clinical trials, but really now the new barrier is not getting past the FDA, the new barrier is reimbursement.”
Fox: Researchers first detected the opioid crisis through death certificate data
“We started seeing signs of the opioid crisis in death certificates, and it was actually public health researchers who started looking at data and seeing, wow, that we're really seeing an uptick in this sort of death among young people.”
Sethi: Longitudinal health data will become far more important than episodic data
“Longitudinal data is going to become ever more important as compared to episodic data. And that's because we're moving from acute care to chronic care.”
Sethi: Over 90% of clinical trials fail to recruit enough patients
“Most clinical trials, upwards of 90% or worse, don't get filled, and it's because there's no frictionless way for a clinical trial's inclusion-exclusion criteria to detect a candidate patient.”
Fox: Patients with data access will form coalitions to demand clinical trials
“When patients all have access to their data, They could form a coalition and ask for a certain clinical trial.”
Sethi: Top EHR vendors hold only 6% of individual digital health data
“If you took the top few vendors For electronic health records, and you reduce the entire U.S. Population down to those vendors. Those vendors, in aggregate, will hold about six percent of all data being generated digitally on you today.”
Sethi: Digital health data fragmentation will outpace EHR vendor consolidation
“Even if there's a consolidation of vendors in the marketplace, which we predict there will be, and I think that's going to be a good thing, the consolidation is going to be outstripped by the runaway fragmentation in digital data.”