Bayesian Health
Runs risk models inside hospital record systems to flag deteriorating patients early, with sepsis as the founding case. A Johns Hopkins spinout.
Highlights
- Real-time clinical-risk machine-learning models running inside hospital EHRs
- Sepsis early-warning detection — hours before traditional vital-sign-based scoring
- Deterioration detection across non-sepsis adverse events
- Founded by Suchi Saria (Johns Hopkins, Bloomberg Distinguished Professor of ML and Healthcare)
- Johns Hopkins spinout commercializing 10+ years of clinical research
- $30M+ raised; Andreessen Horowitz Bio Fund lead with Suzanne Heywood
- Peer-reviewed validation in Nature Medicine showing reduced sepsis mortality
External link — opens bayesianhealth.com in a new tab. Bayesian Health is a third-party product; we are not affiliated with it.
About Bayesian Health
What it is
Bayesian Health runs machine learning models in real time inside hospital electronic health records to detect patients who are deteriorating, with early sepsis warning as its founding case. It is a Johns Hopkins spinout with backing from a16z and its bio fund.
Why it's different
Sepsis detection is a well-chosen problem because outcomes degrade by the hour and the signs are present in data nobody is watching continuously. It is also a field littered with failures: the best-known hospital sepsis model was independently found to perform far worse in practice than its vendor claimed, which is the reason to ask about evidence rather than accuracy here. Bayesian's academic lineage and published validation are the reason it is worth taking seriously. The other honest caveat is alert fatigue — a deterioration model that fires too often gets ignored, and an ignored alert is worse than none because it buys false assurance.
How people use it
It is deployed inside the electronic health record so alerts reach clinicians in the workflow rather than in a separate system nobody opens. The measures that matter are time to treatment and outcomes, not model accuracy in isolation. Any health system evaluating this should ask for independent validation in a population resembling their own, because performance does not transfer between patient populations as readily as the marketing implies.
Written by the n3os team. We are not affiliated with Bayesian Health.
This listing was written from public information, without Bayesian Health’s involvement. If you own it and something here is wrong — or you would rather not be listed at all — email us and we will correct or remove it.
Get the ones worth knowing about
We write one of these for every tool worth the trouble. Get the new ones, plus what we have found genuinely useful lately.
Your address goes to Buttondown, who send the emails on our behalf. One click unsubscribes, and the list is never sold or shared.