No shared mental model
Providers, the coordinator role, and specialist offices each assumed someone else was responsible for follow-up.
Service Design · Health Equity
Closing the
Care Gap
Created a documented workflow at a free community clinic so that underserved patients are actually seen.
Average referral processing time
Before
12+ months
After
2 weeks
Mercy Health Center · Athens, GA
A free clinic for uninsured and underserved neighbors across Northeast Georgia.
Project Snapshot
The context, at a glance
Organization
Mercy Health Center
Role
UX Researcher + Product Designer
Target Users
Care coordinators responsible for managing specialist referrals across a free community clinic.
Duration
3 months of iteration
Team
Care coordinators and healthcare stakeholders
Tools
eClinicalWorks · Epic EHR · Microsoft Excel
The Problem
A process existed on paper,
but not in practice.
Patients weren't failing to receive specialty care because providers didn't care. They were getting lost inside a referral process that no one truly owned.
Research
Understanding a system
that no one could explain.
There wasn't a documented workflow I could study. I reconstructed it by following referrals from beginning to end, interviewing providers, and learning how each specialist office actually accepted patients.
Understanding specialist workflows
In my first two weeks I called specialist offices Mercy referred to. I asked the same questions to each:
01.What is your financial aid application, and what counts as complete?
02.What clinical documentation do you require with each referral?
03.What is your preferred method of contact — fax, phone, email, portal?
04.Who specifically should I reach for financial aid questions, and what is their direct line?
Understanding internal workflows
The investigation revealed that referral failures were driven less by patient behavior than by inconsistent processes. Every specialist office operated differently, ownership shifted between people, and no one had visibility once referrals left the clinic.
Research Findings
The referral journey, before and after
Same patient, same clinic, same specialists. What changed was where ownership lived and whether anyone could see the whole path.
Mercy
Patient
Specialist
No owner
Stage One
Referral written
A provider writes a specialist referral order in the chart.
MercyStage Two
Handoff
The order enters the office with no defined owner downstream.
⚠ No ownerNoneStage Three
Packet sent
The same generic packet is faxed to every specialist, regardless of requirements.
MercyStage Four
Bounced or buried
The office returns it for missing items, or it simply sits unread.
SpecialistStage Five
No follow-up
No one chases status. The referral goes silent and falls off everyone's radar.
⚠ No closure pointNoneStage Six
Patient waits
12+ months pass. Symptoms worsen while the referral sits unclosed.
⚠ Learned helplessnessPatientStage One
Referral written
A provider writes the order.
MercyStage Two
Coordinator owns it
The SOP assigns one accountable owner the moment the order lands.
✓ SOPMercyStage Three
Docs assembled
An in-person session builds the financial aid packet together with the patient.
Mercy + patientStage Four
Tracked daily
Every row carries a status, a next action, and a date. Nothing stalls silently.
✓ TrackerTrackerStage Five
Expedited
A named partner contact accelerates scheduling from weeks to days.
SpecialistStage Six
Seen & closed
The appointment is kept and the referral is formally closed.
✓ Trust rebuiltPatient
What I learned
Across interviews, referral reviews, and conversations with providers, the same patterns surfaced repeatedly. These weren't isolated mistakes—they were predictable symptoms of a process with no shared ownership, no visibility, and no standardized workflow.
One-size-fits-all packets
Mercy was sending the same materials to every specialist, regardless of what each office actually required.
No tracking layer
Once a referral left Mercy, no one could see where it was or whether it had stalled.
◎ Invisible systems
Patient effort assumed, not supported
Applications expected a level of independent document-gathering most patients couldn't meet without help.
◎ Cognitive load
No closure point
There was no defined moment when a referral counted as "done" — so referrals never reached one.
◎ Learned helplessness
These five failures became the design brief for what came next.
Solution
No worflow existed, so I built one.
Five components. One coordinator. A process that turned lost referrals into closed loops.
01
A Standard Operating Procedure
◎ STEP-BY-STEP PROCESS DOCUMENT
The SOP defined who does what, in what order, and where the handoffs live. It also formalized the Referrals Coordinator role as the single point of accountability — referrals were no longer "everyone's job and therefore no one's job."
02
A Master Tracking Spreadsheet
◎ SINGLE SOURCE OF TRUTH
A centralized Excel tracker that gave every active referral a visible status. The tracker became the engine that drove the daily workflow — no referral could stall silently because every row carried a date and a next action.
Lo-fi · first paper sketch
04
Patient-Facing Materials & In-Person Document Assembly
◎ CREATED REUSABLE PATIENT-FACING MATERIALS
Rather than handing patients a list of documents to gather on their own, I sat down with each one during dedicated in-office appointments. We filled out applications in person. For patients who couldn't easily return to the clinic, I made the calls and gathered the documents on their behalf. This was the part of the system that explicitly absorbed the cognitive and logistical load that the old process had silently offloaded onto the patient.
05
Daily Follow-Up & Specialist Partner Relationships
◎ BUILT TRUST BETWEEN PATIENTS AND PARTNER RELATIONSHIPS
Each morning: triage every open row by next-action date. Standing meetings with specialist partners turned transactional contacts into relationships that moved referrals faster.
Refined · the working tracker
Mercy Referrals Tracker — Sanitized Preview
All patient data fictionalized for portfolio display.
| Patient | DOB | Specialty | Referral Date | Financial Aid | Date Sent | Appt Scheduled | Next Action | Status |
|---|---|---|---|---|---|---|---|---|
| A. Martinez | 04/12/72 | Cardiology | 11/14/24 | Approved | 11/18/24 | Yes — 12/02 | Confirm appt w/ patient | Scheduled |
| T. Brown | 07/03/68 | GI — Hepatology | 10/22/24 | Approved | 10/28/24 | Yes — 11/15 | Patient seen ✓ | Closed |
| D. Williams | 09/15/81 | Orthopedics | 12/01/24 | Pending DOL letter | — | No | Walk pt through DOL request | Awaiting Docs |
| R. Thompson | 02/28/55 | Optometry | 11/30/24 | Approved | 12/02/24 | Pending | Follow up w/ St. Mary's | Sent |
| M. Davis | 06/19/79 | Dermatology | 12/05/24 | Not yet started | — | No | Schedule in-office FA session | Open |
| J. Wilson | 03/08/62 | Neurology | 11/08/24 | Approved | 11/12/24 | Yes — 12/18 | Reminder call 12/16 | Scheduled |
| L. Garcia | 08/22/70 | Mental Health | 10/15/24 | Approved | 10/20/24 | Yes — 11/04 | Patient seen ✓ | Closed |
The reusable building blocks
The system held together because its parts were standardized. These are the components I built once and reused on every referral — the closest thing this operation had to a design system.
Status taxonomy
One unambiguous state per referral
OpenLogged, not yet started
Awaiting DocsWaiting on financial aid paperwork
SentComplete packet faxed to the specialist
ScheduledAppointment booked with the patient
ClosedPatient seen — referral complete
Specialist contact card
One per partner office · illustrative
St. Mary's — GI / Hepatology
Preferred contactFax, then call to confirm
Financial aid repDirect line on file
RequiresClinical notes + FA approval
TurnaroundDays, once expedited
Per-specialist checklist
Built from each office's actual requirements
- Provider clinical notes attached
- Financial aid application complete
- Proof of income (DOL letter)
- Office-specific intake form
- Fax cover sheet to named contact
Patient prep guide
The cognitive load, broken into steps
- Book a short in-person session — bring any ID you have
- We gather patient information and necessary documents
- We fill out the financial aid application line by line
- I send the packet and call you the day your appointment is scheduled
The first signal it was working
By the end of the first month, the new process produced what the old one couldn't: closed referrals. Patients who'd waited over a year started getting scheduled — the cirrhosis patient among them. On the tracker, rows finally moved through statuses instead of stalling at "Open," and the backlog shrank week over week.
Decision log
The reasoning, not just the result
Three decisions carried this project. Each one traces the same loop — what the evidence showed, what it meant, the call I made, and the tradeoff I accepted.
01
Patient-facing app, or an internal tracker?
Finding
Hundreds of referrals lived on paper in a drawer. No one could see what was open or how old it was.
Insight
The bottleneck wasn't effort — it was invisibility. Cases stalled because nothing forced an open referral to demand its next action.
Decision
Build one shared tracker that surfaces every open case and its next action daily — not a new patient-facing app.
Outcome
The backlog became burnable. Processing time fell from 12+ months to ~2 weeks.
Tradeoff accepted
Less visible "design" than a patient app — but it's the tool the team could actually run every day without me.
02
Who owns the handoff to the specialist?
Finding
Referral packets were sent into a void. Specialist offices had no point of contact and often never responded.
Insight
The failure was structural, not clinical — an ownerless handoff between the clinic and each specialist.
Decision
Establish a single named point person holding each specialist relationship and its financial-aid contact.
Outcome
Specific cases moved in days; partners gained a reliable coordinator — cited by name in Mercy's public staff feature.
Tradeoff accepted
Concentrating relationship knowledge in one role created a single point of failure — which decision 03 was built to resolve.
03
Clear more cases now, or make the process outlive me?
Finding
A process that worked only because one person pushed it would collapse the moment that person left.
Insight
Durability is a design requirement, not a nice-to-have. A win that can't survive a handoff doesn't really count.
Decision
Encode the work into an SOP, per-specialist checklists, and a daily follow-up cadence anyone could run.
Outcome
The system stayed in use after I left — the outcome I'd point to first.
Tradeoff accepted
Spent time documenting instead of clearing a few more cases short-term, to buy long-term continuity.
Outcomes
Three months of work, measured
The change showed up in the only place that mattered: how quickly patients were seen.
Within three months of stepping into the role, the referral process at Mercy was no longer the same system. It was being actively managed, daily, against a tracker that made every open case visible — and the change showed up in the only place that mattered: how quickly patients were seen.
12+ months → 2 weeks
- Backlog when I started
- Hundreds of stalled paper referrals, most over a year old
- Closure rate on referrals I personally managed
- ~90% reached scheduled care
- Processing time
- Reduced from 12+ months to 2 weeks
- System continuity
- The redesigned process remained in use after my departure
Backlog burn-down
Open referrals fell week over week as the daily cadence cleared the year-plus backlog.
Processing time
Roughly a 26× reduction in average referral processing time.
Closure rate
Of the referrals I personally managed, about nine in ten reached scheduled specialist care.
Illustrative — reconstructed from the tracker records that ran the process. Exact patient figures are withheld and the shapes are approximate.
What changed structurally
The shift wasn't only in throughput. Mercy's relationships with its specialist partners changed too. Where the clinic had previously sent packets into a void and waited for a response, it now had a named point person who held those relationships, knew each office's financial aid contact by name, and could move specific patient cases through in days rather than weeks. The referrals function itself became something Mercy's leadership and outside partners could rely on as a stable part of the clinic's operation — not a recurring source of breakdown.
The strongest outcome — the one I'm most proud of
The system outlived my role. When I left Mercy, the SOP, the tracker, the per-specialist checklists, and the daily follow-up cadence stayed in use. That continuity is the outcome I would point to first if asked whether the work was real or just personal heroics. A process that depends on one person to function is fragile. A process that keeps running after that person leaves is a system. Mercy now has one.
In Mercy's words
"Jenny Negrete came to Mercy in October of 2024 and serves as our Patient Representative and Referral Specialist. Since her arrival, she has tremendously improved our patients' access to specialty care around Athens and has given our partners, like St. Mary's, a point person for care coordination. Jenny has been a wonderful addition to Mercy and we love all that she does to advocate for them to get the care they need."
Mercy Health Center · Public staff feature, 2025What it meant for patients
The patients who had waited longest were the first to benefit — including the cirrhosis patient, scheduled and seen in the first cycle of the new process. Many told me they had stopped expecting the system to work at all. The redesign was, in part, built to earn that expectation back.
Most case studies I read in school were about increasing conversion rates. This one was about whether a patient with cirrhosis would see a hepatologist before their disease progressed further. The work I did at Mercy taught me to take design seriously because the people on the other side of it are taking their lives seriously. That's the standard I bring to every product decision.
Reflection
How this work shaped what I bring to product
01
Walking into ambiguity
No training, no documented process, a year-plus backlog. The discomfort of not knowing what to do next is not the same as having nothing to do — the work was to ask, observe, and map until the system's shape appeared. Shipping before you understand is the same instinct that built the broken process I inherited.
02
Designing for people, not users
At Mercy, the word "user" never fit. I sat with sick, exhausted people and asked them to gather bank statements; I called the Department of Labor while they waited beside me. Empathy isn't a research method or a value statement, it's a practice — the person on the other side of every screen has a body and a limit, and the design either respects that or it doesn't.
03
Building systems that outlive you
The outcome I'm proudest of is that the system kept working after I left. Early on I could have been the system — holding every relationship and follow-up in my head and calling it competence. Writing it down so someone else could inherit it was the harder, better choice, and the difference between a fix and a system.
What surprised me
I expected the fix to be about chasing specialist offices harder. It wasn't. The breakdown was structural — no one owned the handoff — so the highest-leverage move was making every open case visible in one place, not working faster. Visibility changed behavior more than effort did, and that reframing is the part I carry into every product problem since.
Design tradeoffs I weighed
- A lightweight tracker now, not a perfect tool later
- An Excel master tracker shipped in days and made every open case visible. Waiting on a built-in EHR feature would have meant patients kept waiting — I chose speed and visibility over elegance.
- One named owner first, then documented out of fragility
- Centralizing accountability in a single point person fixed the void immediately, but a one-person process is brittle. Writing the SOP and checklists traded short-term effort for a system that could outlive me.
- Triaged by clinical urgency, not arrival order
- A year-old backlog can't be worked first-in-first-out. Sequencing sickest-first was less "fair" by queue logic, but it was the only defensible choice with a cirrhosis patient in the stack.
- Documented while clearing, not after
- Slower throughput in the first weeks bought a process that still ran months later. I treated durability as the real deliverable, not the cleared backlog alone.
What I'd build next · how the system could evolve
The process works, but it still leans on manual follow-up. If I carried it forward, the next moves are clear:
- Automate the follow-up cadence — status-based reminders so a stalled referral surfaces itself instead of waiting for a person to notice.
- Behavioral SMS reminders — multilingual and rewritten to remove shame triggers, carrying the same psychology lens to the patient-facing edge of the system.
- A referral-health dashboard — so leadership can see closure rate and aging cases at a glance, without depending on one person to report it.
- Proper baseline instrumentation — capturing processing time and closure rigorously from day one, so the next round of impact is measured, not reconstructed.
- Fold the SOP into the EHR — moving the checklist into eClinicalWorks so the process lives where the work already happens.