←
Atropos Health410 posts
Atropos Health avatar

Atropos Health

@AtroposHealth

Generating publication-grade real-world evidence in minutes to days for precision medicine and clinical decision support.

◷ Joined April 2022▧ 166 Photos and videos
Atropos Health avatar
Atropos Health@AtroposHealth·Oct 27

Our on-demand EHR consultation service took the research team's question and turned it into a propensity-matched cohort study: 205,817 @StanfordHealth patients with normal neutrophil counts and no prior chronic disease, followed for up to ten years.

Atropos Health avatar
Atropos Health@AtroposHealth·Oct 27

A "normal" blood test can still predict disease a decade out. That's one finding from a new @NatImmunol paper on neutrophils, the most common immune cell in your blood: nature.com/articles/s41590-0… 🧵

Atropos Health avatar
Atropos Health@AtroposHealth·Oct 14

Our team is headed to Boston for the 14th Annual IMPACCT Real-World Evidence Summit! We’ll be there to talk about all things novel RWE generation, data fitness, and delivery at the point of care. 📅 Oct 27 - 29, 2026 📍 Boston, MA 5️⃣ Booth #5 #RWE #HEOR #IMPACCTRWE

Post attachment
Atropos Health avatar
Atropos Health@AtroposHealth·Oct 9

For three of the four models, 63–76% of answers using web search alone cited their sources accurately, yet those sources described a different patient group, comparison, or outcome. Others found no matching evidence.

Atropos Health avatar
Atropos Health@AtroposHealth·Oct 9

The result: with web search alone, 13–18% of answers were precision matches. With Alexandria, that rose to 40–56% – a 2.8x to 3.6x increase for every model tested.

Atropos Health avatar
Atropos Health@AtroposHealth·Oct 9

We graded each answer on two points. Does one cited source match the patient, treatment, comparison, and outcome in the question? Does the answer report that source accurately? Answers that did both earned a "precision match."

Atropos Health avatar
Atropos Health@AtroposHealth·Oct 9

We posed 209 patient-specific questions like this to four frontier LLMs. Each model answered every question twice: once with web search alone, and once with added access to Alexandria®, our library of real-world evidence.