0
MODEL CATALOG

‍One solution.
20+ outcomes predicted.

Training and validation on de-identified 1.3M+ surgical case cohort — full metrics shared under NDA and in forthcoming peer-reviewed publications.

↳ Model 01. Global Risk Score

20+ OUTCOMES · ENSEMBLE
‍
One calibrated score combining patient and procedure risk.
Output: 0–100 score, risk tier and drivers
‍AUROC range: 0.90–0.96
Check our AI model cards
Risk Score  65/100
MINS → High Risk
Cursor icon showing a hand pointer
AKI → Low Risk
Pain → Low Risk
Transfusion → Low Risk

↳ Model 02. RBC Transfusion Volume

‍Predicts expected red blood cell requirements per procedure.

Output: predicted RBC units per procedure
‍Model Type: LGBM
‍MAE: 0.43 units
Check our AI model cards

↳ Model 03. ASC Eligibility Prediction

‍Predicts whether a case is suitable for ambulatory evaluation or requires additional review.‍

Output: ASC candidate, review or inpatient
‍Model Type: LGBM
AUROC range: 0.90–0.96
Check our AI model cards

↳ Model 04. Hospital & ICU Length of Stay

‍Predicts total postoperative hospital and ICU days.‍

Output: expected hospital and ICU stay
‍Model Type: LGBM Tweedie
‍MAE LoS Time: 0.3—0.5 days
Check our AI model cards

↳ Model 05. Postoperative VTE Prediction

‍Predicts venous thromboembolism following surgery.

Output: probability, risk tier and drivers
‍Model Type: LGBM
‍AUROC: 0.89

Check our AI model cards

↳ Additional Outcome Models

15+ ADDITIONAL PREDICTIONS

‍Pain, AKI, sepsis, pneumonia, readmission, ICU admission and other surgical complications.
Check our AI model cards
MODEL CATALOG

‍One solution.
20+ outcomes predicted.

Training and validation on de-identified 1.3M+ surgical case cohort — full metrics shared under NDA and in forthcoming peer-reviewed publications.

↳ Model 01. Global Risk Score

20+ OUTCOMES · ENSEMBLE
‍
One calibrated score combining patient and procedure risk.
Output: 0–100 score, risk tier and drivers
‍AUROC range: 0.90–0.96
Check our AI model cards
Risk Score  65/100
MINS → High Risk
Cursor icon showing a hand pointer
AKI → Low Risk
Pain → Low Risk
Transfusion → Low Risk

↳ Model 02. RBC Transfusion Volume

‍Predicts expected red blood cell requirements per procedure.

Output: predicted RBC units per procedure
‍Model Type: LGBM
‍MAE: 0.43 units
Check our AI model cards

↳ Model 03. ASC Eligibility Prediction

‍Predicts whether a case is suitable for ambulatory evaluation or requires additional review.‍

Output: ASC candidate, review or inpatient
‍Model Type: LGBM
AUROC range: 0.90–0.96
Check our AI model cards

↳ Model 04. Hospital & ICU Length of Stay

‍Predicts total postoperative hospital and ICU days.‍

Output: expected hospital and ICU stay
‍Model Type: LGBM Tweedie
‍MAE LoS Time: 0.3—0.5 days
Check our AI model cards

↳ Model 05. Postoperative VTE Prediction

‍Predicts venous thromboembolism following surgery.

Output: probability, risk tier and drivers
‍Model Type: LGBM
‍AUROC: 0.89

Check our AI model cards

↳ Additional Outcome Models

15+ ADDITIONAL PREDICTIONS

‍Pain, AKI, sepsis, pneumonia, readmission, ICU admission and other surgical complications.
Check our AI model cards

First-class AI models for surgery & anesthesia

License Hoopcare's validated risk models and ship clinical prediction.
One solution, 20+ predicted outcomes — global risk, MACCE, AKI, transfusion, length of stay, ICU, ASC eligibility — deployable in your cloud, your customer's VPC, or fully on-prem.
Test our models →
Schedule a meeting
Test our surgery risk calculator
Lap cholecystectomy Total knee Total hip Bypass Colectomy Appendectomy Hernia repair
Hoopcare preoperative evaluation shown inside an electronic health record
INNOVATION

Gradient boosting on ICD-10 codes is what everyone has.
We read the notes to catch key unstructured data.

Portrait of Dr Yassine Moussali, cofounder of Hoopcare
"Three years of notes. 150,000 characters per patient. H&P, consults, prior anesthesia records — read on every case." — Dr. Moussali, Anesthesiologist
Functional status
"Walks two blocks, stops for dyspnea" is a feature, not a sentence.
Cardiac reports
LVEF, stress test results captured from cardiology consults.
Prior anesthesia events
The difficult airway from 2019 lives in a note, not a field.
Frailty signals
Falls, assistance at home — scattered across notes, never coded.
Social context
Escort, distance from hospital, home support: the ASC-eligibility deciders.
+ More
Learn more
OUR UPCOMING PUBLICATIONS

Validated AI models trained on 1.3 M+ procedures to predict complications.

Model development and analysis are complete.
Final manuscript details are being prepared ahead of submission.
AI-Enabled Preoperative Chart Review for Surgical Complications Prediction
PRE-SUBMISSION - 2026
AUTOMATIC CHART REVIEW
JAMA Surgery logo
AI-Enabled Preoperative Triage for ASC Eligibility  Prediction
PRE-SUBMISSION -  2026
ASC ELIGIBILITY
Anesthesiology journal logo
Machine Learning Model to Predict RBC Transfusion Volume
New England Journal of Medicine AI logo
PRE-SUBMISSION - 2026
TRANSFUSION RISK
Preoperative Prediction of Postoperative Hospital & ICU Length of Stay
Annals of Surgery logo
PRE-SUBMISSION - 2026
LENGTH OF STAY PREDICTION
LLM driven M&M review: Cause, Mechanism, and Preventability
Nature logo
PRE-SUBMISSION - 2026
AI M&M REVIEW
Machine Learning Model to Predict VTE after Surgery and Guide Thromboprophylaxis
British Journal of Anaesthesia logo
PRE-SUBMISSION - 2026
VTE RISK
Machine Learning Model to Predict High Pain Level after Surgery
British Journal of Anaesthesia logo
PRE-SUBMISSION - 2026
PAIN RISK
AI-EKG Model to Predict Low Ejection Fraction before Surgery
PRE-SUBMISSION - 2026
CARDIAC EVALUATION
MODEL CATALOG

‍One solution.
20+ outcomes predicted.

Training and validation on de-identified 1.3M+ surgical case cohort — full metrics shared under NDA and in forthcoming peer-reviewed publications.

↳ Model 01.

Global Risk Score

20+ OUTCOMES · ENSEMBLE
‍
One calibrated score combining patient and procedure risk.
Output: 0–100 score, risk tier and drivers
‍AUROC range: 0.90–0.96
Check our AI model cards
Risk Score  65/100
MINS → High Risk
Cursor icon showing a hand pointer
AKI → Low Risk
Pain → Low Risk
Transfusion → Low Risk

↳ Model 02.

Number of RBC Required

‍Predicts expected red blood cell requirements per procedure.

Output: predicted RBC units per procedure
‍Model Type: LGBM
‍MAE: 0.43 units
Check our AI model cards

↳ Model 03.

ASC Eligibility Prediction

‍Predicts whether a case is suitable for ambulatory evaluation or requires additional review.‍

Output: ASC candidate, review or inpatient
‍Model Type: LGBM
AUROC range: 0.90–0.96
Check our AI model cards

↳ Model 04.

Length of Stay (total & ICU) & OR time Prediction

‍Predicts total postoperative hospital and ICU days.‍

Output: expected hospital and ICU stay
‍Model Type: LGBM Tweedie
‍MAE LoS Time: 0.3–0.5 days
Check our AI model cards

↳Model 05.

Postoperative VTE Prediction

‍Predicts venous thromboembolism following surgery.

Output: probability, risk tier and drivers
‍Model Type: LGBM
‍AUROC: 0.89

Check our AI model cards

↳ More AI models

Additional Outcome Prediction

15+ ADDITIONAL PREDICTIONS

‍Pain, AKI, sepsis, pneumonia, readmission, ICU admission and other surgical complications.
Check our AI model cards
Why it works

Trained across 1,500+ procedure types.

Lung tumor
Pancreatic surgery
Heart valve procedures
Hip replacement
Hysterectomy
CABG
Gastric Sleeve
Bunion
Colectomy
Breast cancer surgery
Appendectomy
Cardiac surgery
Plastic surgery
Spine surgery
Shoulder replacement
Brain tumor
Knee replacement
Vascular procedures
Colonoscopy
Cataract surgery
Hernia
C-section
Thyroidectomy
HYBRID MODEL

Everything the models need: structured data and unstructured notes.

Structured Data
Light Model
100-feature models
FEATURES
  • Problem list & coded diagnoses (ICD-10)
  • Medication list — active prescriptions
  • Procedure & CPT code
  • Essential labs — CBC, BMP, coagulation
  • Demographics, vitals
Structured Data
Standard Model
500-feature models
FEATURES
  • Everything in Light, plus:
  • Full lab history & vitals trends
  • Coded comorbidities & prior encounters
  • Prior utilization — admissions, ED visits, procedures
  • Functional & social flags
Clinical Notes Analysis
Boost Performance with Clinical Notes Analysis
Clinical Notes AI Review
FEATURES
  • Reads up to 3 years of notes — 150k characters per patient
  • H&P, consults, prior anesthesia records
  • Extracts functional status, frailty, cardiovascular symptoms
  • +20 note-derived features on top of structured models
  • Available on any tier — same API, one extra flag
BUY VS. BUILD

Your clinical AI layer.  Deployed in weeks, not built in years.

Three years of work, one integration.

A validated clinical-risk layer needs labeled surgical outcomes at scale, clinical expertise to define the endpoints, and evidence your sales team can cite. We've done that part: a de-identified cohort of 1.3M+ surgical patients, models built with anesthesiologists, publications on the way. You integrate an API; your product gains a clinical brain.

→ Differentiate your platform — scheduling everyone has; risk-aware scheduling nobody has.
→ Evidence included — peer-reviewed citations your enterprise deals can lean on.
→ Maintained, recalibrated, monitored — model drift is our job, not your backlog.
Book a consultation
Hoopcare emblem
SHIP FASTER
CITE THE EVIDENCE
DEPLOYMENT

Runs where the data lives. Answers where you need them.

Containerized models with a REST + FHIR interface, deployed in your cloud, your customer's VPC (Azure, GCP, AWS), or fully on-prem — no PHI ever reaches Hoopcare. Score a patient from structured data alone, or attach the LLM note-analysis layer where notes are available. Our engineers support your integration from first call to production.
Book a technical walkthrough →
API-first
REST + FHIR endpoints, versioned models, and a sandbox with synthetic patients to build against.
Zero PHI egress
Inference happens inside the deployment perimeter — your BAA story stays clean.
Diagram of Hoopcare deployment inside a hospital environment
PLANS

One goal. Prediction, wherever your product runs.

Per endpoint or full catalog
For OR tech & analytics platforms
100 Features Model
Contact us
/ annual license
FEATURES
  • All core endpoints or à la carte
  • Hybrid deployment — your cloud, customer VPC, on-prem
  • Sandbox + synthetic test patients
  • Model updates, recalibration and drift monitoring
  • Evidence pack: validation summaries and citations as published
Book a demo
Per endpoint or full catalog
For OR tech & analytics platforms
500 Features Model
Contact us
/ annual license
Includes
  • All core endpoints or à la carte
  • Hybrid deployment — your cloud, customer VPC, on-prem
  • Sandbox + synthetic test patients
  • Model updates, recalibration and drift monitoring
  • Evidence pack: validation summaries and citations as published
Book a demo
Advisory
For teams building in-house
Consulting & Services
Contact us
/ advisory
WITH OUR CLINICAL DATA SCIENCE TEAM
  • Clinical endpoint and feature design reviews
  • LLM note-extraction architecture guidance
  • Validation and calibration methodology
  • Regulatory and publication strategy
  • Led by our anesthesiologist-founders
Book a demo
THEY TRUST US

Built with the world's
‍top teams.

Mayo Clinic and Arizona State University logoJohns Hopkins logoTechstars logoRamsay Sante logoMayo Clinic logoArizona State University logoIllume logoQuincy logoCornell Tech logoMayo Clinic and Arizona State University logoMACSF logoMayo Clinic and Arizona State University logoTechstars logoIllume logoCornell Tech logo

We handle the hurdles of shipping clinical AI.

From security review to validation study, we've done this inside health systems — so your integration doesn't start from zero.
Integration engineers included
Our team works alongside yours — API, deployment, monitoring — from first call to production.
HIPAA & BAA covered
Zero-PHI-egress architecture keeps your compliance story simple — BAA signed where needed.
Validated, and validatable
Performance measured on your customer's population before go-live, with clinicians in the loop.
Evidence, not promises
Every model is headed for peer review — your sales team gets citations, not claims.

Contact us

Curious about our solution or have inquiries? We're eager to assist! Just complete the form below, and we'll respond promptly.
THEY TRUST US
Mayo Clinic and Arizona State University logoJohns Hopkins logoTechstars logoMayo Clinic and Arizona State University logoRamsay Sante logoMayo Clinic logoArizona State University logoQuincy logoCornell Tech logo
Name
email
message
Send message
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
‍



‍
Hoopcare logo
Hoopcare is a medical technology company founded by doctors to make surgery and anesthesia safer.
PATIENTS
HomeTerms of UsePrivacy PolicyNotice of Privacy PracticesHIPAA AuthorizationInformation CenterCookie settingsContact Us
HOOPCARE
Terms of UseTelehealth PolicyGo to Hoopcare France
© Copyright 2026 - Hoop Care, Inc. - INTELLIGENCE ANESTHESIA - HOOPCARE MEDICAL OF FLORIDA P.A.
Disclaimers
Please be aware that HSA/FSA eligibility, contribution amounts, rollover provisions, and other key features may change annually and differ between plans. We recommend consulting your plan administrator, HR department, or a qualified tax or financial advisor to understand how these accounts apply to your specific situation. We are not responsible for any errors or changes in regulations that may affect your FSA or HSA benefits.
‍
Hoopcare makes available virtual care management programs for eligible individuals, and works with the following organizations who coordinate to deliver their respective services through the technology solution developed by Hoop Care Inc.: Hoopcare Medical of Florida P.A. and other affiliated professional entities delivering clinical care; wholly owned subsidiaries of Hoop Care Inc. Hoopcare features may change over time. All health records may not be available. ‍Hoopcare is not a medical device and has not been evaluated by the FDA. Please talk to a healthcare provider if you need help.

SOC 2 Compliant - HIPAA Compliant