We built an AI platform that reads clinical notes, matches them to payers' medical necessity rules and records the evidence behind each prior authorisation. 85% of authorisations are now automated, and denials fell 14%.
Client: Ascend CareIndustry: HealthcareService: AI & Automation
Overview
Client
Ascend Care
High-volume utilisation management operation
Engagement snapshot
The challenge
Prior authorisation ran on manual faxes and separate portal uploads. Care decisions took 3-5 days, which added work for providers and raised the risk of denials when documentation was missing.
Nurses checked faxed clinical notes against PDF guidelines by hand, and patient data passed through several manual steps on the way.
Manual reviews held up urgent treatments, lowering patient satisfaction and risking harm to patients.
Simple authorisations cost so much to handle that they were eating into the medical loss ratio (MLR).
Slow medical necessity checks damaged trust with the provider network.
The solution
We set up live clinical data collection through FHIR (Fast Healthcare Interoperability Resources) and HL7 v2 feeds.
AI agents read unstructured clinical notes in the EHR and map them to each payer's medical necessity guidelines.
Each scoring event is saved with a full reason trail that links the specific clinical evidence to the care-level decision.
Key features
AI agents read unstructured clinical notes from the EHR and map them to each payer's medical necessity guidelines.
Business impact
14% fewer denials
Standard requests are approved in under two seconds with reason codes, and complex cases go to clinicians.
Business impact
85% of authorisations automated
One shared set of clinical policy rules, updated by region, keeps decisions consistent across 50 states.
Business impact
Zero privacy violations
Each decision includes evidence-based reason codes that answer provider questions and regulatory audits.
Business impact
100% policy defensibility
Standard care is approved automatically, so clinicians only review complex grey-area cases.
Business impact
↑ 2.1x throughput
Before and after
Before
Static guidelines
Nurses checked PDF guidelines against faxed clinical notes by hand.
After
AI scoring
AI assistants read EHR data against the current medical policies as it arrives.
Before
Manual handling
Patient health information could leak during manual data entry and transfer.
After
Automatic protection
Unneeded patient data is removed and fields are encrypted as soon as data is collected.
Before
3-5 Business Days
Disconnected workflows held up treatment.
After
Sub-2 seconds
An immediate decision for 85% of standard medical necessity requests.
Results
2.1x Throughput
Standard requests that took 72+ hours now get a decision in under 2 seconds, and the team handles 2.1x as many authorisations.
Before: 72+ HoursAfter: < 2 Secs
14% Denial Reduction
Missing documentation is now flagged before a request reaches a reviewer, so fewer requests are denied.
Before: UncertainAfter: 14% Lower
1.8s p95 Response Time
Scores come back fast enough that clinical work is not held up waiting for the system.
Before: < 5sAfter: 1.8s
Testimonial
“Standard authorisation decisions took 3 to 5 days. They now take under 2 seconds, with reason codes we can explain. Coretus helped us automate 85% of authorisations, denials are down 14%, and complex cases still go to our clinicians.”
Julian PierceDirector of Utilisation Management, Ascend CareKey results
Trust and control
Every automated decision shows the clinical evidence behind it, so regulators can see how it was reached.
Compliance readyStrict field-level encryption and removal of unneeded personal data, built to HIPAA and HITECH requirements.
HIPAAThe cloud-based streaming platform scales reliably during open enrolment surges.
Cloud nativeOn completion, the client owns 100% of the IP in the clinical models and health-record connections.
100% ownedFaster delivery
Reusing these components saved about 10 weeks of setup, which the team used to map clinical notes to each payer's guidelines.
Removed unneeded patient data and applied field-level encryption as clinical data arrived.
Provided tested FHIR and HL7 v2 connections to the EHRs.
Tracked scoring times, auto-approval rates and cases sent to clinicians.
Measured the handling cost of each authorisation, so savings against the medical loss ratio were visible.
Requests arrived by fax and through separate portal uploads, and nurses checked faxed clinical notes against PDF guidelines by hand. Decisions took 3-5 business days, which delayed treatment and frustrated providers. Missing documentation also led to more denials, and simple requests cost too much to handle.
It collects clinical data live through FHIR and HL7 v2 feeds. AI agents read the unstructured notes in the EHR and map them to each payer's medical necessity guidelines, with 99.2% recall. Standard requests are approved in under 2 seconds, and complex grey-area cases go to clinicians for review.
Every scoring event is saved with a reason trail that links the specific clinical evidence to the care-level decision, with SHAP values showing what drove each score. Reason codes answer provider queries and regulatory audits. Patient data is de-identified and encrypted at field level when it is collected.
If your reviewers still match faxed notes to PDF guidelines, we can help you automate standard medical necessity checks and keep clinicians on the hard cases.