Brain Health USA: a production-ready EHR roadmap for high-volume, multi-state psychiatric care

When a high-volume specialty care platform like Brain Health operates across hundreds of providers and thousands of daily patient interactions, the gap between a working prototype and a production-grade system is where clinical and financial risk quietly accumulates.

The client had already built a proof of concept using AI coding agents, covering intake, scheduling, charting, billing, and telehealth. What they didn't have was a clear picture of whether it could run at the scale the clinic was already operating at.

That's when BrainHealth engaged Vinta for a healthcare-focused discovery: a 3-week structured process to map end-to-end user flows, assess the existing platform, and produce a validated roadmap before any production build begins.

A 3-week discovery to define the foundation for a psychiatric EHR

Brain Health USA is a multi-location mental health provider offering psychiatric evaluations, medication management, therapy, and telehealth for children, adolescents, and adults across California. Their network spans 200+ licensed providers and handles over 2,500 patient interactions per day, a scale that turns every workflow inefficiency into an operational risk.

What the Healthcare Discovery covered:

• A Value Proposition Canvas mapped across five key roles (providers, intake admins, verification specialists, scheduling admins, and billing admins) to surface the highest-priority pain points and the outcomes the platform must deliver
• An end-to-end user flow tracing how a single patient encounter moves through the entire system: from the first intake link via SMS through clinical visit, post-visit charting, claim submission, and ongoing care
• A Care Service Blueprint mapping handoffs, risks, and dependencies across roles and external integrations (Stedi, DoseSpot, Twilio, Health Gorilla)
• UX design recommendations based on a hands-on heuristic evaluation of the existing AI-generated prototype
• A technical assessment of the existing codebase and a phased MVP roadmap

Understanding the client and their operational reality

Brain Health USA makes mental health care easier to access across California, combining in-person and telehealth appointments for a wide range of psychiatric conditions. 

Their model is high-volume and multi-role: on any given day, intake admins onboard new patients, verification specialists confirm insurance eligibility, scheduling admins coordinate provider availability, providers run back-to-back telehealth sessions, and billing admins manage claims and denials, often across disconnected systems.

When Vinta joined, the team was working across Oracle NetSuite, OfficeAlly, and Excel spreadsheets to move a single patient through intake, insurance verification, scheduling, and billing. Every role touched at least two systems, and no single view connected the full patient journey.

The operational fragmentation created compounding risk: insurance eligibility failures that blocked scheduling, claim denials that had to be caught and refiled within a 3-month window, and providers finishing clinical notes after hours because in-session documentation support was missing.

The platform situation: a functional prototype that couldn't run at production scale

Before engaging Vinta, Brain Health USA had already built a proof of concept using AI coding agents. The prototype successfully validated the target workflows: intake, scheduling, EHR, clinical notes, billing, ticketing, communications, and telehealth. The tech stack was the right foundation, but the problem was the gap between demo velocity and production readiness.

At the target scale of 2,500 daily patient interactions, several architectural patterns in the prototype would create real operational problems. Core read operations fetched entire resource batches from Medplum and filtered them in the browser because status and flag values were stored in custom FHIR extensions rather than standard fields. 

Medplum can filter and paginate standard fields server-side, but it can't filter on custom extensions, which forces the application to pull database contents into browser memory for basic aggregations. Under real concurrency, that pattern means staff dashboards timing out on busy days and clinical operations coming to a halt.

Planning an EHR build or assessment on Medplum?
Build confidently with our Medplum implementation experts.

Concurrent-write safety was the exception rather than the rule. Multi-resource operations (like intake submission, which creates or updates a patient record, a coverage record, an eligibility response, an appointment, a slot, and a workflow task in a single action) were not atomic. 

A partial failure left the chart in a half-written state with no clean recovery path. Audit logging existed but was applied inconsistently, creating a compliance gap that SOC 2 auditors and integration partners would flag. Workflow automation ran inside Next.js cron routes rather than Medplum Bots and FHIR Subscriptions, moving retries and audit logging into the web layer where failures are harder to detect and recover from.

The PoC demonstrated the workflows, but couldn't run them under production load, compliance expectations, and real-world error conditions.

What we built: a validated roadmap and technical foundation for the build phase

Over three weeks, our team executed an in-depth product discovery, where they:

• mapped end-to-end user flows
• identified core personas
• assessed the existing prototype
• aligned on a value proposition centered on provider efficiency and patient experience

The output directly informs scope, architecture, and design priorities for the build phase ahead.

Vision alignment and Value Proposition Canvas

We ran a Vision Alignment workshop with Brain Health's executives, clinicians, PMs, and engineers to establish the care strategy, constraints, and success criteria before any work began. From there, we applied the Value Proposition Canvas across the five key roles to surface the highest-priority gaps.

Three findings shaped everything that followed:

1. The main workflow friction sits between intake, insurance verification, and provider matching, currently spread across Oracle NetSuite, OfficeAlly, and Excel, with no unified patient view.
2. Providers spend excessive time on post-visit documentation. At 30–40 telehealth sessions per day, the absence of AI-assisted charting and in-session documentation support translates directly into after-hours work.
3. Billing lacks automated denial tracking, which creates revenue leakage risk at scale. Centralized claims management with guided workflows is the most direct fix.

Proto-persona workshop and end-to-end user flow

We mapped five personas based on direct operational knowledge from the Brain Health team: Provider, Intake Admin, Verification Specialist, Scheduling Admin, and Billing Admin. The biggest pain across every persona was fragmentation: every role operating across at least two disconnected systems, producing duplicate work, communication gaps, and manual processes that introduce clinical and financial risk.

The end-to-end user flow traced a single patient encounter through the entire system: from the intake link delivered via SMS, through clinical visit with structured psychiatric assessment, post-visit charting and prescription, claim submission, and ongoing care. That exercise surfaced the critical handoff points where the current stack creates the highest risk, the same points that directly shaped the UX recommendations and MVP scope.

Care Service Blueprint and UX design recommendations

The Care Service Blueprint mapped frontstage actions (intake coordinators, scheduling admins, billing admins), backstage actions (provider workflows), and external system integrations (Stedi for real-time eligibility checks, DoseSpot for e-prescribing, Twilio for SMS communications, Health Gorilla for lab ordering) across the full patient lifecycle. This shared view made dependencies visible and gave the team a single reference for architecture and sequencing decisions.

The UX recommendations addressed the existing prototype's most critical usability failures before any new features are built: broken multi-step scroll behavior, navigation patterns that hid entire functional areas from specific roles, missing visual urgency for items requiring immediate action, and a quick-action system that redirected users away from the current page rather than opening in-flow modals.

Technical assessment and MVP roadmap

The technical assessment identified the architectural patterns in the existing codebase that wouldn't hold under production load and made a concrete recommendation: use the existing prototype as reference material for UX flows, note templates, integration payloads, and workflow states, and rebuild on Medplum using its native primitives.

The case for Medplum is practical. The integrations Brain Health depends on already speak FHIR through Medplum's built-in connectors:

• DoseSpot syncs prescriptions directly as MedicationRequest resources
• Health Gorilla writes lab orders as ServiceRequest and ingests results as DiagnosticReport

Medplum's hosted offering also absorbs the compliance and infrastructure surface that would otherwise require dedicated hiring and on-call coverage:

• HIPAA BAA and SOC 2 posture out of the box
• Native AuditEvent and RBAC via AccessPolicy
• Encrypted storage and event-driven automation via Bots and FHIR Subscriptions

Each concrete capability on the MVP roadmap is easier to build on Medplum than off it.

The High-Level Roadmap Proposal organized the build into five sequential phases across an 11-month MVP with a team of three developers:

Phase 1 — Intake (~2 months): magic link intake, 9-step intake form, insurance verification, provider matching with hard and soft filters, and scheduling with day/week/month calendar views for providers and staff.

Phase 2 — Appointment (~2 months): clinical notes with psychiatric templates (Initial Evaluation, Follow-up, Medication Management), unified patient chart, and telehealth with video call and virtual waiting room.

Phase 3 — Post-care (~1 month): clinical assessments including PHQ-9 and GAD-7, lettering and documents via DocuSign integration.

Phase 4 — Billing (~1 month): real-time eligibility checks via Stedi/Candid, claim submission, denial management, and insurance payment processing.

Phase 5 — Ongoing care (~3 months): unified patient timeline, communication systems via Twilio (phone, SMS, in-app), ticketing system, notifications, and role-based access management.

Need to scope an EHR build for a high-volume care model?
We run Healthcare Discovery engagements in 3 weeks.

Next steps: from roadmap to production build

The Healthcare Discovery gave Brain Health USA a validated, human-centered foundation and a build path that's grounded in operational reality rather than prototype assumptions. The AI-generated prototype confirmed the ambition; the discovery defined the path to make it real.

With many patients and a growing provider network, the margin for workflow errors is zero. The scope, architecture decisions, and integration sequencing that came out of this engagement give the team the context to move into the build phase without the guesswork that typically drives EHR project overruns.