Skip to main content
A referral is a medical decision, but the tools meant to support it often fall short. The shallow decision trees and the incomplete, inaccurate directories that should point a patient toward the right provider too often point them toward the wrong one: someone who doesn’t perform the procedure, doesn’t take the plan, or can’t see them for months. In our own work, we’ve found that roughly 70% of referral obstacles could have been prevented with better information available at the moment the referral was made. That’s the root cause we’re after. The information that prevents those obstacles is diverse: what a provider actually does, where they actually practice, what patients actually experience, whether they can actually be seen. So we’re building it one domain at a time.

​
Grounded in practice, not principle

Our models exist to improve referrals for patients, and that purpose defines their boundaries. We are not trying to describe what could be true in principle based on a license, training, or registration. We are trying to estimate what is true in practice at the moment a referral is made. That means grounding each model in observed behavior and other real-world evidence, then defining its output around a concrete referral decision. E.g. a retina specialist is an ophthalmologist, and in principle is capable of cataract surgery; in practice, a routine cataract referral does not belong with them. Our methods are built to capture the second judgment, not the first. That framing also means we will sometimes not know the answer. The available evidence may be incomplete or conflicting. Where that uncertainty exists, we say so rather than guess.

​
The models

Each model page details its methodological approach, validation, and known limitations. Click into any model to learn more.

Referral Scope Match

Open betaRoutine referral categories matched to what a provider actually does, derived from observed practice patterns.

Stale Location Affiliation Risk

Open betaRisk that a practitioner-location affiliation is stale or not patient-facing.

Availability

Closed betaHow soon a provider can be seen at a specific location.

Patient Experience

In developmentWhat patients actually report about access and experience.

In Network Probability

In developmentLikelihood that a provider will still be in network in the next 90 days.

More on the way

Coming soonWe’re building toward every piece of information that prevents a referral obstacle. Tell us what would help most.