Revenue Cycle Automation at a US Health System — 14% Net Patient Revenue Uplift Across 22 Hospitals
Revenue Cycle Optimizer + Coding Assistant + Medical Records Parser delivering an integrated revenue-cycle platform from charge-capture to denials management.
The challenge
The client — a US health system with 22 acute-care hospitals and an associated network of ambulatory clinics across multiple states — was operating a revenue-cycle function that the CFO described as 'a slow leak'. Net patient revenue was running 8-12% below the system's calculated capture potential, with the gap distributed across under-coding (charges not captured because clinical documentation did not support full coding), first-pass denials (claims rejected by payers that required appeals or were ultimately written off), and slow AR cycles (accounts receivable averaging 56 days against an industry best-practice of high 30s).
The revenue-cycle function was structurally fragmented across the 22 hospitals — each hospital had its own coding team, its own denials-management process and its own AR-followup workflow, with limited cross-hospital standardisation or learning. Coding accuracy varied widely by hospital, denials patterns were addressed reactively rather than systematically, and the central revenue-cycle leadership lacked the cross-hospital visibility needed to drive improvement.
The constraints were operational. Wholesale EHR replacement was off the table — the system had only recently completed a multi-year Epic deployment and the CFO had no appetite for another platform programme. The HIPAA compliance posture meant any AI processing of clinical data had to happen inside the system's HIPAA-compliant environment. The coding and revenue-cycle teams — approximately 1,400 staff across the 22 hospitals — could be reskilled but not replaced en masse.
The approach
MindMap deployed an integrated revenue-cycle platform composed of Revenue Cycle Optimizer (Rc) as the cross-cycle analytics and denials-management layer, Coding Assistant (Cd) as the coding co-pilot, Medical Records Parser (Mp) for the clinical-documentation analysis, Claims Router (Cr) for the payer-specific submission optimisation, and AR Automation (Ar) for the receivables-follow-up workflow.
Phase one was the coding-assist build. Coding Assistant was integrated with Epic as a coding co-pilot — for each encounter, the platform analyses the clinical documentation, suggests the ICD-10 and CPT code set, and surfaces the documentation excerpts that justify each code. The coder confirms, modifies or rejects each suggestion with the coder's action logged as the system-of-record. The platform's suggestions are not auto-submitted; the coder remains the decision-maker.
Phase two was the clinical-documentation improvement (CDI) layer. Medical Records Parser analyses the clinical documentation and identifies coding-opportunity gaps — clinical findings that, if documented more specifically by the clinician, would support more accurate coding. The CDI specialists use the platform's gap-identification to query clinicians for documentation enhancement during the encounter rather than after coding.
Phase three was the denials-management layer. Revenue Cycle Optimizer's denials-prediction module evaluates each claim before submission, flags claims at high risk of denial, and surfaces the specific risk factors with remediation recommendations (e.g. missing prior-authorisation reference, payer-specific coding-modifier requirement, documentation insufficiency for the requested procedure level). Claims that go ahead and are denied are routed through an appeals-management workflow with the platform's draft-appeal-narrative as a starting point.
Phase four was the AR-follow-up automation. AR Automation handles the systematic follow-up on aged receivables — payer-specific follow-up scripts, automated re-submission of correctable claims, escalation routing for the cases that require human intervention.
The pre-built building blocks
Rather than commission a ground-up build, the engagement leaned on MindMap's pre-built accelerator library — production-tested components that compress what would otherwise be a six-to-nine-month build into weeks.
Revenue Cycle Optimizer
Cross-cycle analytics, denials prediction and AR-follow-up workflow
Coding Assistant
Epic-integrated coding co-pilot
Medical Records Parser
Clinical-documentation analysis for CDI
Claims Router
Payer-specific submission optimisation
AR Automation
Receivables follow-up workflow
The architecture
The platform runs in the system's HIPAA-eligible AWS environment, with all PHI processing inside the system's dedicated VPC and the system's HIPAA encryption-and-access-control posture maintained throughout.
Coding Assistant and Medical Records Parser are integrated with Epic via Epic's standard FHIR-based clinical-data APIs (the system is on the Epic-hosted Epic deployment and the integration uses Epic's SMART-on-FHIR pattern for the in-Epic co-pilot UX). The clinical-documentation analysis uses a fine-tuned Llama 3.1 70B variant trained on the system's de-identified clinical corpus, with the coding-suggestion output structured around the ICD-10-CM and CPT/HCPCS code sets the system uses.
Revenue Cycle Optimizer's denials-prediction model is a gradient-boosted-tree ensemble trained on the system's historical claims-and-denials data. Features include payer-specific patterns (different payers have different denial profiles), encounter-specific factors (procedure type, documentation completeness, prior-authorisation status), and provider-specific patterns (specific providers have specific documentation patterns that drive specific denial categories).
Claims Router handles the payer-specific submission optimisation — formatting the claim per the payer's specific requirements, including the payer-specific coding-modifier rules that have become a structural feature of US payer-provider interaction. The system's existing claims-clearinghouse relationship was preserved; the platform sits upstream of the clearinghouse and improves the claims that flow through.
AR Automation handles the receivables workflow with payer-specific follow-up scripts, automated re-submission of correctable claims, and escalation routing. The platform's actions are logged for the system's standard internal-audit and external-audit purposes.
The system's standard HIPAA audit trail covers every PHI access, every model inference and every workflow action.
The numbers behind the story
Net patient revenue across the 22 hospitals has risen approximately 14% on a same-store basis (controlling for service-mix change and payer-mix change). The uplift is distributed across the three main contributors: more accurate coding (capturing approximately 6% of the uplift), reduced first-pass denials (capturing approximately 5%) and faster AR cycles (capturing approximately 3%).
AR days have dropped from 56 to 47, with the system's CFO targeting continued reduction to high 30s over the next 18 months as the AR-automation workflow matures. The denials first-pass rate has dropped 62% — the platform's pre-submission denial-prediction layer has caught the bulk of the previously-denied claims at draft time.
Coding accuracy has improved on the system's internal coding-quality sampling. The improvement is most pronounced on the high-complexity encounters where the previous coding process had been most error-prone; the platform's specific clinical-documentation analysis catches the secondary diagnoses and the procedure-specific coding nuances that drive accurate DRG assignment.
Cross-hospital standardisation has been a strategic outcome. The platform has provided a consistent coding, denial-management and AR-follow-up workflow across the 22 hospitals, with the central revenue-cycle leadership able to see cross-hospital patterns and drive systematic improvement. The performance gap between the best-performing and worst-performing hospital has narrowed materially.
The 1,400-person revenue-cycle team has been restructured rather than reduced. The coding role has shifted toward higher-judgement work (the AI handles the routine coding suggestions; the coder focuses on the complex cases). The denials-management role has shifted toward systematic improvement (the AI handles the rote re-submissions; the denials specialist focuses on root-cause work with the payers and the clinical teams).
“Our revenue-cycle was leaking eight to twelve per cent of our calculated capture potential across coding, denials and AR. MindMap's integrated platform recovered fourteen per cent in eighteen months — across all twenty-two hospitals, without replacing Epic, with our revenue-cycle teams doing higher-judgement work. The platform paid for itself inside the first year.”— Chief Financial Officer· US Health System
Why MindMap was chosen
The system had evaluated several healthcare-revenue-cycle vendors. Most were point-solution providers (coding only, or denials only, or AR only) with limited integration across the revenue cycle. The few integrated platforms required wholesale workflow replacement that the system's revenue-cycle leadership considered operationally unacceptable.
MindMap's accelerator-composition approach — bringing Revenue Cycle Optimizer, Coding Assistant, Medical Records Parser, Claims Router and AR Automation together into an integrated platform that augmented the system's Epic deployment rather than replacing it — was the structural differentiator. We could demonstrate the integrated pattern at a peer US health system.
Our embedded US healthcare-revenue-cycle expertise on the delivery team (three former revenue-cycle directors from peer US health systems and a former coding-specialist trainer) was the third factor. The system's CFO felt that the team understood the operational reality of US healthcare revenue cycle, not just the modelling.
Related deployments
Prior Auth Acceleration
Prior Auth Accelerator + DocGenie automated 70% of payer prior auth submissions end-to-end, reducing turnaround from 3 days to 4 hours.
Medical Records Processing
Medical Records Parser processing 14,000 patient documents per day across nine hospitals, lifting coding accuracy from 87% to 99.2%.
Clinical-Trial Document Workflow
DocuMage + Clinical Trial Matcher accelerated study-start document cycle 5x — letting trial sites enrol patients weeks sooner than the prior baseline.
Want an outcome like this?
Start with a 2-week AI Readiness Sprint. We deliver a prioritised use-case backlog and business case grounded in what's actually buildable with our accelerator library.