Company
Medical evidence moves faster than the institutions that act on it.
We reconstruct how a medical claim actually moved through the public record — when it appeared, what strengthened it, where it spread — and rank the ones worth your attention. Whether those same trajectories can forecast what a regulator or guideline body does next is a separate question, and we are testing it in public rather than asserting it.
Can these decisions be forecast at all, ahead of time — and do the probabilities mean what they say? Nobody has answered that with a public record. We would rather find out than assert it.
Claim trajectories, reconstructed from public sources and ranked against your portfolio. Every ranking shows its working, so you can check the evidence rather than trust the order it came in.
Founder
Kanav Jain
Kanav built products at Epic, Doximity, Transcarent, and CancerCompass — systems used by clinicians, patients, and healthcare organizations to manage care access, communication, and navigation. He co-founded Andwise, a physician financial platform, then spent a sabbatical building governance infrastructure for algorithmic healthcare systems. NextConsensus emerged from a pattern he saw across every layer: the higher the stakes, the harder it was to trace, question, or undo a decision.
The problem
Medical evidence moves faster than the institutions that rely on it, and the cost of that lag lands on teams with dates: an evidence review window set against a guideline the public record had already overtaken, a launch plan that learns about a coverage change the day it publishes. The lag is first a visibility problem: the movement is already in the public record, but reconstructing it by hand is days of work per claim, so most teams read the snapshot and infer the rest.
So we rebuild the movement instead, and rank what it implies for a given portfolio. Whether those same trajectories also forecast what a regulator or guideline body does next is a separate question, and one we are testing in public rather than assuming.
How the system stays honest
Every trajectory we hand over is rebuilt from dated public revisions and carries a hash, so you can reconstruct it yourself and compare rather than take the ranking on trust. The research program adds a second discipline on top: each probability is frozen against a dated evidence cutoff and registered before the outcome is known, so the record is the one that was made rather than the one that reads best afterwards. The full standard lives at method and limits.
What builds up over time
Two records accumulate, and they are not the same asset. The first is the reconstruction corpus — every claim trajectory rebuilt from public revisions, reproducible by anyone who wants to check it. That is what a program delivers today.
The second is the scored record, still being built on research questions. Over time it holds:
- Transitions that occurred
- Plausible transitions that did not
- Probability trajectories over time
- Authority behavior patterns
- Timing and lead time
- Calibration across cohorts
- Baseline-relative performance
That record gets built in public first, on research questions, before any of it is sold into a customer's decision. The order matters: you should be able to see how the forecasts have held up before you depend on one.
Where it gets used first
The first place this pays for itself is claim review — seeing which of your approved claims are being contested in public before you are asked to defend them. That is an application, though, not the company. The company is the trajectory reconstruction underneath it.
Evidence planning, market access, and competitive assessment all draw on the same object. Where we additionally register a forecast — naming the authority, the action, and the date, then scoring it against a public source — that is the research program running in the open, not a capability you are being sold.
How Refract fits
The machinery that watches the public record and tracks how it changes over time is Refract, and it is open source — anyone can inspect how the evidence is captured. What we keep is the forecasting layer on top and the scored record it produces.