Tellicare
Beta · Clinical decision support

Evidence at the point of care.

Physicians are aware of less than 10% of the data relevant to atypical clinical situations. Tellicare turns real-time disease pathway statistics into a single, calm recommendation, at the moment the decision is made.

Live model · Centre Hospitalier de Troyes · Updated regularly

tellicare.ai / epcpass
Patient at admission
A. Doe · 65 y
Mode of entry
Home
Type of stay
Other
EPC cases in unit
2
Antibio. prior stay
No
Recommendation
Could pass
Probability of EPC carriage
31.4%threshold 50%
CH TroyesReims University HospitalHCS Research UnitBayer FoundationInserm · CIC-EC
The gap

Care is decided on a small
sliver of what's known.

Every hour, clinicians make decisions that quietly exclude 90% of the relevant evidence, not by choice, but because the data is in registries, ICU dashboards, lab feeds and electronic health records that don't talk to each other. Tellicare collapses that gap into a single answer.

  • Real-time pathway statistics, not retrospective dashboards.
  • Calibrated to the institution's own population.
  • One calm recommendation, never a wall of charts.
<10%
Of relevant data physicians are aware of in atypical situations
35min
Average time spent searching for context per uncertain admission
3-tier
Recommendation: pass · monitor · screen
<200ms
Median latency from form submit to model response
How it works

From scattered data
to a single decision.

Tellicare is an API-first platform. Every module, starting with EPCpass, is a model served behind a stable, audited endpoint, and a clinical UI tuned for the moment of decision.

01
Ingest

We connect to admission systems, lab feeds and the EHR, and learn the institution's pathway baselines.

02
Calibrate

Models are re-trained on local data so the prior probability matches your patient mix, not someone else's.

03
Decide

At the bedside, fill three groups of fields and get one calibrated probability with a clear recommendation.

04
Audit

Every recommendation is logged and signed, ready for clinical governance, IRB and post-market follow-up.

First module

EPCpass: pre-test for EPC carriage at admission.

Carbapenemase-producing Enterobacterales (EPC) screening is currently triggered by ad-hoc rules. EPCpass replaces that with a calibrated probability on the patient in front of you, saving lab cost on low-risk admissions and catching high-risk ones earlier.

  • Logistic model trained on first & second hospital stays.
  • Returns probability + threshold-based recommendation.
  • Designed for < 60 seconds at the admissions desk.
Predictors
9
AUC (internal)
0.78
Threshold
0.50
POST /api/predict
{
  "birthdate": "1960-03-25",
  "admission_date": "2025-11-25",
  "provenance": "Home",
  "type_of_stay": "Other",
  "n_continuous": 2,
  "binary_var_1": false,
  "binary_var_2": false,
  "binary_var_3": true,
  "binary_var_4": false
}

→ {
  "probability": 0.314,
  "threshold": 0.50,
  "recommendation": "pass"
}
Roadmap

Built to be the go-to
evidence layer for hospitals.

EPCpass is the first of a series of focused modules. Each is a calibrated answer to a decision a clinician already makes 50 times a week. Not a chatbot, not a dashboard.

Q2 2026
EPCpass · public beta
First clinical deployment, weekly recalibration.
Q3 2026
Pathway-of-care companion
Live ICU decompensation risk based on EHR signals.
Q4 2026
CE marking pre-submission
Class IIa medical device pathway, supported by published cohorts.

Try the live beta.

The EPCpass screening tool runs against the trained logistic model. Use the demo patient, or enter a synthetic case to see the pathway.

Or write to jan.chrusciel.md@gmail.com.