Dialogue · Healthcare · Care Platform

Leave of absence, logged at the source

Doctors at Dialogue wrote the same leave of absence twice, and the care team interpreted it by hand. I replaced both with one structured form inside the consult note, so signing it triggers everything that used to be manual.

RoleProduct Designer, research to handoff

Team1 PM, 1 designer, 2 developers

Timeline4 months

What I owned: planning and running the research, concept development, the final solution design, and the training documentation, written with medical operations and the product manager.

Care Platform with the Medical leave form in the consult note and the automated 'LOA event triggered' message in the episode, beside the generated leave of absence PDF

60→100%of leaves logged to the patient recordData Analytics, project brief

6→2doctor steps per leave (67% fewer)Workflow comparison

~4 hrsof practitioner documentation removed every weekTime saved per leave × volume

The tension: For 80–100+ leaves a week, doctors filled out an attestation form and then retyped the same details in free text. The care team had to read every note to work out what to do. Anything unclear meant missed follow-ups and unlogged leaves, and patients waited on a busy queue for the documents they needed to start their leave.
The turn: Research showed the real failure was the handoff between doctors and operations, not the doctors' workflow. So I designed one structured Medical leave form inside the consult note, using a pattern doctors already knew from prescriptions.
Where it landed: Doctors went from 6 steps to 2 (fill in, sign). The PDF, patient message, internal note, record and care team tasks now happen automatically, and every leave is logged.

Problem

Written twice, handed off, read by hand

Every leave crossed two teams, and the details were copied, handed off and reinterpreted along the way. Doctors prescribed 80–100+ a week. Shadowing both teams, I mapped the full chain.

Workflow map before. Doctor: start consultation, select LOA PDF template, fill out LOA details in free text, write Next Steps note, save consult note, assign case to Care Team queue. Handoff, where a case can take hours to get picked up. Care coordinators: open case from queue, read Next Steps, send LOA doc to patient, use r/ to log dates in internal chat, track manually in a task or Google Sheet. — The workflow before. Highlighted: the doctor writing the same details twice, and the coordinator's logging step, where every coordinator did it differently.

Duplicated informationDoctors wrote the same details twice, and inconsistently: some filled out both, some skipped Next Steps, some only partly filled it in.

Inconsistent loggingCoordinators recorded leaves their own way: protocol, other platform features, Google Sheets or Slack. Leaves went unlogged and follow-ups were missed.

Patients waitingPatients needed the documents to start their leave, but got them only when a busy queue reached their case.

On top of that, enterprise clients got incomplete reporting, with no reliable data on their workforce's health.

"There's no reliable way to know what to act on. Every practitioner does this differently."Stephanie M., M.D.

Research

The break was in the handoff

I started out treating this as a doctor problem: too many manual steps, too much room for error. Shadowing both teams and interviewing doctors and operations changed that.

Job shadowing, both teams5 physician interviews3 clinical ops interviewsBefore / after workflow maps

1
Doctors weren't skipping steps out of carelessness.
The system gave them no structured path, so each one documented leaves their own way.
2
The care team needed to know when to act, not just what to do.
They had built informal workarounds just to keep track of active leaves, and those workarounds didn't scale.
3
The handoff was where it fell apart.
The gap between what a doctor documented and what operations could act on was the core failure point.

The reframe. This wasn't about fixing a doctor's workflow. It was about creating one source of truth that served both sides of the care team, captured at the moment the leave was prescribed.

How might we capture a leave once, in a form the system can read, without adding work for doctors?

Exploration

Pull data out, or put it in?

I explored three concepts. The first two tried to pull structured data out of the documents doctors already made.

Concept 1: a Leave of Absence section with fixed fields added to the Medical Attestation PDF template, which the system reads to fill in Next Steps

Structured PDF template

A Leave of Absence section with fixed fields in the Medical Attestation form, which the system reads to fill in Next Steps.

Structured input beats free text
Parsing data out of a PDF proved unreliable for our system
Doctors had to judge when the section applied

Not chosen

Concept 2: a structured Medical leave or attestation section in Next Steps, with attestation type, continuation and duration, generating the Next Steps text

Structured Next Steps

Free-text Next Steps replaced with a "Medical leave or attestation" section: type, continuation, dates.

Easy for the system to read
Still redundant: doctors fill out the attestation form, then enter the same details again

Not chosen

Medical leave form in the consult note

One structured form that generates the PDF and everything else.

One entry for the doctor
One source of truth for the system
A pattern doctors already knew from prescriptions

Chosen

The chosen design reversed the flow: capture the structured data once, and generate the documents from it.

The pushback

The technical team questioned whether the automation could be ready for launch, or whether the solution was too complex. I worked closely with our engineers and the data team to shape a workflow that kept the core of the idea, the system recognizing the leave as an event inside the doctor's workflow, and made sure every automated output was correct before launch.

Solution

Fill it in, sign it. The rest happens on its own.

One structured Medical leave form, inside the consult note, built on the prescription pattern doctors already used every day. Try it: it plays on its own, and takes over the moment you click in.

Open the Medical leave section of the consult note.

Playing demo

Generate the leave of absence PDF
Medical_leave.pdf
Bilingual, signed, ready for the employer
Send the document to the patient in chat
Your medical leave document is ready. You can share it with your employer.
📄 Medical_leave.pdf
Post the internal message for the care team
Internal Clinic (Auto)
LOA event triggered:
Type:
Continuation of leave of absence:
Duration:
Follow-up:
Record the leave in the patient's profile
Event Dates Employer Status
Leave of absence Logged
Create tasks for care coordinators
☑ Book follow-up ·

Event	Dates	Employer	Status
Leave of absence			Logged

What it produces

Generated bilingual Medical leave of absence PDF with prescriber, patient, type, continuation, duration, follow-up, details and signature — The bilingual PDF the patient gives their employer, generated from the form.

Internal Clinic (Auto) message: LOA event triggered, with type, continuation, duration and follow-up — The automated message in the episode, so the care team sees what happened without reading a note.

Care Platform: the Medical leave form in the consult note on the left, the automated LOA event message in the episode chat — The automated message in the episode, so the care team sees what happened without reading a note.

Before and after

The doctor's part shrinks to starting the consultation, adding the Medical Attestation form, and completing and signing it. Everything the care team used to do by hand now happens the moment the form is signed, with no handoff or queue in between.

Workflow map after. Doctor actions: start consultation, add Medical Attestation form to consult, complete and sign LOA form. Automation: generate LOA PDF, send doc to patient via chat, send internal message outlining system actions, record LOA information in backend, generate tasks for care coordinators. — The workflow after: three doctor actions, then automation. The handoff and the queue are gone.

Impact

Captured once, at the source

Every leave is now captured at the moment it's prescribed, with nothing left for the care team to interpret.

100%of leaves logged to the patient record, up from 60%

6→2doctor steps per leave (67% fewer)

~4 hrsof practitioner documentation removed every week

~$80ka year in care coordinator time (estimate)

The care team's manual leave workflow is fully automated. For the business, enterprise clients now get reporting built on structured, real-time data instead of incomplete records. And with care coordinators no longer reading notes and logging leaves by hand, the time saved on the task came to roughly $80k a year.

Reflection

Follow the cost, not the complaint

The person feeling the pain isn't always where the problem starts. Doctors felt the paperwork, but the cost landed on operations, and only fixing the handoff between them solved it for both.

What I'd do differently

Talk to operations first. I scoped this as a doctor problem and only found the real one in interviews. Starting from the downstream team would have reached the reframe sooner.

Design the reporting with clients in the loop. Clients were the end consumers of this data, and their input on the reporting format would have shaped which fields the form captures.

What's next

Show leave history and status in the patient's profile, reuse the same event-logging pattern for other clinical events, and tie employer reporting to care outcomes.