Discharge planning
Ranks the inpatient census so case managers focus on the highest-risk patients.
Clinical Intelligence Agent
Patient readmission risk prediction
An agent that scores every admission for 30-day readmission risk from the live FHIR record, explains the drivers, and hands the care team a ranked action list.
Explainable scoringCare-team-in-the-loopFHIR R4Audit Logged
Trusted by global innovators
What is readmission risk prediction?
Readmission risk prediction is a model that scores each inpatient's chance of returning to hospital within 30 days of discharge and shows the factors behind the score. Agnotic Technologies builds AI patient readmission risk models on your FHIR data that rank high-risk patients for the care team, validated against your own outcomes.
Care managers often run a static LACE score by hand. That work does not scale and misses social and utilization signals in the record.
The agent pulls LACE- and HOSPITAL-style features from FHIR data, runs an ML risk model and routes ranked high-risk patients to the care team with a calibrated score and its top drivers.
Readmission risk architecture
The risk agent reads FHIR data, computes features, scores with an explainable model, and routes ranked patients to the care-team worklist.
How does the model score risk?
ML risk scoring grounded in known features, explained per patient.
ML risk scoring
Trains and calibrates a gradient-boosted model on your population for a 30-day readmission probability, revalidated against your own outcomes.
LACE / HOSPITAL-style features
Engineers features from length of stay, acuity, comorbidities (Charlson/Elixhauser), prior-year ED visits, discharge labs, and prior utilization.
Built on FHIR data
Assembles the feature set from FHIR R4 (Encounter, Condition, Observation, MedicationRequest, Procedure) on live chart data, not a nightly extract.
Per-patient explanations
Surfaces the top contributing factors for each score (prior admissions, low discharge hemoglobin, polypharmacy) so the care team sees why.
Where it runs
Where is readmission risk scoring used?
Heart failure & COPD programs
Prioritizes condition-specific cohorts known for high readmission.
Transitional care management
Triggers TCM outreach (48-hour calls, med reconciliation) for high-risk discharges.
Population health
Feeds risk-stratified cohorts into value-based-care programs and outreach queues.
What changes for the care team?
Directional: actual results depend on population, feature availability, and intervention capacity.
Ranked
Census stratified by calibrated 30-day risk
Explained
Every score ships with its top drivers
Monitored
Calibration and drift tracked over time
Compliance-First Healthcare App Development Services Backed by Global Standards
15-Minute Scoping Call01
HIPAA
Health Insurance Portability and Accountability Act
Protect PHI with privacy-first architecture, encrypted storage and transmission, strict access controls, and traceable audit logs.
02
GDPR
General Data Protection Regulation
Implement lawful consent flows, data minimization, retention controls, and secure processing for sensitive health data.
03
FHIR
Fast Healthcare Interoperability Resources
Enable standardized health data exchange across apps, care teams, and systems through standards-based FHIR APIs.
04
HL7
Health Level Seven International
Support enterprise-grade interoperability with HL7-based integrations for records, events, and clinical messaging workflows.
05
HITRUST
Health Information Trust Alliance
Align security programs to healthcare-specific control and risk management practices trusted by providers and ecosystem partners.
06
HITECH
Health Information Technology for Economic and Clinical Health Act
Design with breach notification readiness, digital record safeguards, and operational controls that support regulated care programs.
07
SaMD
FDA Software as a Medical Device
Plan software quality, traceability, and documentation pathways for products that may require SaMD review and submission.
08
MDR (EU)
Medical Device Regulation (European Union)
Prepare EU market-ready processes for risk classification, evidence tracking, and lifecycle governance under MDR expectations.
09
SAMHSA
Substance Abuse and Mental Health Services Administration
Apply confidentiality controls and consent-aware sharing models for behavioral and mental health data experiences.
Standards we build against
Which standards does it use?
FHIRBulk FHIRHIPAAICD-10
How It Works
From admission to a ranked care-team list
FHIR data in, explainable score out, ranked patients to the care team, outcome written back for monitoring.
1.
Capture: FHIR record
On admission and at discharge planning, the agent reads Encounter, Condition, Observation, and MedicationRequest from the live FHIR record.
FHIR R4 features
2.
Processing: scoring + guardrails
Features are computed, the calibrated model scores 30-day risk, and explanations are generated; incomplete cases are flagged, not silently scored.
Explainable ML
3.
Action: rank + route to care team
High-risk patients are ranked into the care-management worklist with drivers and suggested actions; the care team decides, not the agent.
Care-team-in-the-loop
4.
Write-back: score + outcome
The score, drivers, and actions are written to the record and logged; observed readmissions feed calibration and drift monitoring.
Audit logged
Facts and sources
Which rules shape readmission risk models?
Short facts we build and validate risk models around, each linked to its primary source.
Under the Hospital Readmissions Reduction Program, Medicare reduces an applicable hospital's base operating DRG payment for each discharge to account for excess readmissions.
Source: 42 CFR 412.154 (eCFR)In HL7 FHIR R4, the RiskAssessment resource captures predicted outcomes for a patient or population on the basis of source information, such as the risk of a health outcome.
Source: HL7 FHIR R4, RiskAssessmentThe NIST AI RMF lists the characteristics of trustworthy AI as valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed.
Source: NIST AI RMF 1.0, section 3 (NIST AIRC)FDA's Clinical Decision Support Software guidance explains which decision support functions meet the non-device criteria in section 520(o)(1)(E) of the FD&C Act and which remain device software functions.
Source: FDA guidance, Clinical Decision Support Software
Related proof
Read Case StudyLera Health: compliant women's health platform
Related proof of compliant, AI-assisted delivery, not this exact agent. Lera Health shows how we turn clinical data into personalized, actionable guidance on a privacy-first data layer.
Compliance & Guardrails
15-Minute Scoping CallHow are risk scores kept fair and explainable?
A risk model that touches discharge decisions has to be explainable, fair, and audited.
PHI handling
Feature computation and scoring run inside your compliance boundary under a BAA, encrypted throughout, with PHI never used to train external models.
Care-team-in-the-loop
The agent scores and ranks; care managers and clinicians decide interventions. No discharge action follows automatically from a score.
Explainability
Each score ships with its top contributing features, so the care team validates rather than over-trusts an opaque number.
Audit logging & fairness
Scores, drivers, and outcomes are logged immutably, with subgroup calibration monitored so the model doesn't disadvantage a population.
Frequently Asked Questions
Feature computation, scoring and training run inside your compliance boundary under a signed BAA, encrypted throughout, with scores and outcomes audit-logged.
No. It scores, ranks, and suggests transition actions; care managers and clinicians own every intervention. No action follows automatically from a score.
It computes LACE- and HOSPITAL-style features from the live FHIR record, runs a calibrated ML model, explains each score, and routes ranked patients to the care team.
It depends on how many years of outcome data you have, how clean the FHIR or warehouse data is, how many patient subgroups need fairness checks and where the score must appear in the care team's workflow. We give you a timeline after reviewing your data in the scoping call.
Yes. The agent reads Encounter, Condition, Observation and MedicationRequest data through your EHR's FHIR API or a data warehouse, and writes the score back so it appears in the care-management worklist.
Yes. The first release scores one population, such as heart failure discharges at one hospital, and runs in shadow mode so you can compare scores with real outcomes before staff act on them.
AI tools help our engineers draft code and tests, but a person reviews every change, and it passes the same automated tests and security checks as any other code. Which tools may see your code or data is agreed with you at the start.
Ready to stratify your census with confidence?
Tell us your population and EHR. We'll return an architecture review and a plan to validate a readmission model on your outcomes before it drives decisions.
contact@agnotic.com
Partnerships
contact@agnotic.com