responda

Responda

Request access

Voice-to-chart ePCR for EMS

Written on scene. Reviewed by QA. Checked against the state’s own rules.

The crew records the call, Responda drafts the chart from their own words, a QA reviewer accepts it or sends it back, and billing works from the same record. Every value is approved by a clinician before it enters the chart.

Responda builds and validates the NEMSIS record. It does not submit to any state.

Request access
Backed by the Johns Hopkins University Ignite programNative iOS · iPad first, iPhone for field capture
The Responda chart editor on an iPad. A source-inputs pane on the left holds Transcript, Dispatch and Notes tabs and notes that transcripts stay in memory for the session only and are never stored on the device. The main pane shows the Vital signs section of a chest-pain chart with a recorded set of vitals: blood pressure 148 over 92, heart rate 104, respiratory rate 20, oxygen saturation 95 per cent, and a sinus tachycardia ECG rhythm. A section rail down the side marks which parts of the chart already have content, and the footer shows section 6 of 12, one item outstanding before submit, and a Review and submit button.
The chart itself — every section, with the source pane beside it. Sample chart; the agency, patient and crew are fictional.

Accepting suggestions

Nothing enters the chart until you accept it

The crew records the call and the chart is drafted from what was said. Every drafted value arrives as a suggestion, next to the sentence it came from and the person who said it — and it stays a suggestion until a clinician accepts it, one at a time, in the real form.

“Nothing here is in the chart. Each suggestion is accepted one at a time, by you.”
Not a slogan — the app’s own words, at the foot of the review panel in the screenshot below.
The Responda review panel open on an iPad beside the chart it is proposing values for. The panel is headed Awaiting review with a count of four and lists the suggestions grouped by chart section: a dispatch complaint of chest pain, a chief complaint of chest pressure, a history of present illness, and a first set of vitals reading systolic 128 and heart rate 88. Each suggestion shows the speaker it came from and the sentence they said — the medic's “dispatched for chest pain, patient is awake and talking”, the patient's “it's more pressure than pain, right in the middle” — and each carries an empty status ring, because none of them has been decided yet. On the chart beside it, the dispatch complaint field is still blank. Below a Start reviewing button the app states: nothing here is in the chart, each suggestion is accepted one at a time, by you.
The review panel, beside the chart it is proposing values for. Nothing on the left has changed: every suggestion is still waiting on a decision, and the field it belongs to is still empty. Sample chart; the agency, patient and crew are fictional, and the quoted speech is invented.

A suggestion is not a chart entry

The panel counts what is waiting; the field it belongs to stays empty until you decide. Accept, edit, reject or skip — the chart changes on your decision, not on the recording.

The evidence travels with the value

Each suggestion carries the speaker and the exact phrase behind it, so a reviewer opening that field a year from now reads the sentence that produced it rather than a score on its own.

Decided one at a time, by a clinician

Every accept, edit and reject is logged against its field, and an existing value is never silently overwritten. Denials stay denials; home and EMS medications stay separate.

The workflow

From voice to a chart a reviewer has accepted

Five steps from the first word on scene to a QA reviewer's decision — with the clinician in control at the only place that matters, and a written reason whenever a chart goes back.

1Capture

Record the call as you run it

Start the blue microphone on scene, or upload audio afterward. Pause and continue as the call demands — finishing hands the recording straight to transcription.

PCR 2026-0184

Draft

Capture · recording

Source inputs

Capture

Suggestions stay pending until reviewed
Recording00:42

Tap to pause

When you stop — no live transcript on screen

  1. Uploading
  2. Transcribing
  3. Identifying speakers
  4. Extracting

The device sends the recording once, at the end.

Finish recording
Paste transcript…
Process transcript

2Understand

Every voice, kept apart

The recording is transcribed and split by voice. Name the speakers before the call and every turn carries their name — so history is credited to whoever said it, and a speaker label is never inferred from the audio.

PCR 2026-0184

Draft

Transcript · diarized

Speakers were named by the crew before the call. An unnamed voice stays Speaker A, Speaker B — a label is never inferred from the audio.

MedicFamilyPartner
Medic0:12

Any blood thinners? When was he last normal?

Family0:19

Around 10:30 last night. He takes Eliquis.

Partner0:31

BP 184 over 98, pulse 86 irregular, sat 97 room air, GCS 13.

Medic0:44

Right facial droop, arm drift, slurred speech. Cincinnati positive. Stroke alert, priority one.

3Extract

Fields, with their evidence

Clinical details become suggested PCR fields — each carrying the exact phrase, speaker, and confidence behind it. Denials stay denials; home and EMS medications stay separate.

PCR 2026-0184

Draft

Extraction · source-backed

Suggested fields · each with its source

Primary impression

High

Possible stroke (CVA)

“facial droop, arm drift, slurred speech… Cincinnati positive”

Medic · 0:44

Initial vitals

High

184/98 · 86 irreg · 97% RA · GCS 13

“BP 184 over 98, pulse 86 irregular…”

Partner · 0:31

Home medication

Medium

Eliquis (anticoagulant)

“He takes Eliquis”

Family · 0:19

4Review

You approve every value

Suggestions are adjudicated one at a time inside the real form — accept, edit, reject, or skip. Nothing enters the chart without the clinician's decision.

PCR 2026-0184

Draft

Review · in the form

Complaint & impression

Chief complaint
Facial droop, slurred speech
Primary impression
—

4 of 12

Primary impression

High

Possible stroke (CVA)

“facial droop, arm drift, slurred speech… Cincinnati positive”

Medic · 0:44

AcceptAEditERejectRSkipS
BackExitEsc

5Hand off

A second pair of eyes, on the record

The finished chart reaches a QA queue with its next step named. A reviewer accepts it or sends it back with a written reason the crew can act on, and every transition is versioned — so how the chart got here is part of the chart.

PCR 2026-0184

Draft

Worklist · waiting on QA

QA worklist · oldest first

Age

Incident 2026-0184

Medic 12 · stroke alert

Waiting on QA14m

Incident 2026-0183

Medic 7 · chest pain

Waiting on QA1h

Incident 2026-0181

Medic 3 · refusal

Sent back to crew3h

Reviewer decision

Sent back to crew — with a reason

Last known well is in the narrative but not on the timeline. Please add the time and resubmit — everything else looks good.

R. Alvarez, QA · 08:41 · version 3

Every transition is versioned, and every accepted field keeps its evidence and its audit entry.

Capabilities

Built for documentation that holds up

Fast where it can be, careful where it must be. Responda takes the typing off the crew without taking accountability away from QA, billing, or a courtroom — and says plainly where it stops.

Voice capture on scene

Record the call as you run it, or upload audio afterward. Multi-speaker audio is split by voice, and speakers the crew named before the call carry their name on every turn — so a statement stays attributable to its source.

A speaker label is never inferred from the audio. Voices nobody named stay Speaker A and Speaker B.

Clinician-approved autofill

Every suggested field carries its transcript evidence, speaker, and confidence. Denials stay denials, home and EMS medications stay separate, and the narrative is drafted in your agency's SOAP, CHART, DCHART, or SOAPIER format.

Approval gating and audit trail

Suggestions stay pending until a clinician accepts, edits, or rejects them one at a time, and existing values are never silently overwritten. Every decision is logged per field, and complaint-specific checklists surface missing or high-risk documentation before the chart is handed off.

Compliance record and state rules

NEMSIS charts serialize against the v3.5 document tree and validate against the national XSD plus stacked national and state Schematron, with Maryland and North Carolina rule sets in the pilot. kiirabikaart records build against the standard's published card structure and validate against its national rule set.

Responda does not submit to any state. A clean validation run is a mapping preview, not acceptance — and state rules only apply once an administrator has configured the agency's destinations.

Billing worksheet and claim lifecycle

Level of service, condition codes, payer coverages, loaded mileage, and a face sheet — on a claim lifecycle that runs separately from the clinical one.

A biller can never edit clinical documentation. Level of service and medical necessity are selected by a human, never inferred.

Security and access

Roles and per-status permissions, agency-scoped queries, MFA support, an append-only audit trail, and a versioned snapshot on every transition.

Responda does not claim HIPAA compliance.

System of record

Built to be audited

A pilot decision is a records decision. These are properties of the record itself — how it is read, how it is retired, who can sign in, and where a time may come from.

  • A read leaves a trail

    Opening a chart is an audited event, not just editing one. Print, the face sheet, the billing queue and the claim workspace write read events too — the surfaces where patient identity is seen at volume.

  • Voided, never deleted

    There is no destructive path in the product — not for a submitted chart, and not for an empty draft. A record can stop being usable; it cannot stop having existed.

  • Invitation-gated sign-in

    Accounts are provisioned by an administrator and sign in with organizational Google or Microsoft identities. There is no self-serve signup.

  • No time is taken from the transcript

    A time is never taken from the digits a model heard — a misheard “ten-oh-five” must not become a fact in a legal record. Where no honest time exists, the field stays blank for the clinician to stamp.

Request access

Bring Responda to your agency

Access is granted to approved agency accounts, so there is no self-serve demo of a live patient-record product. The screens above mirror the real interface; a request gets you a walkthrough of it with your crew, QA and billing leads in the room.

  • For the people who decide: chiefs, QA leads, medical directors, billing managers
  • We review every request by hand and reply with a walkthrough, not an instant login
  • Your details are used only to reply to you — no sales sequence, no newsletter
  • Already approved? Sign in with your agency Google or Microsoft work account — access is invitation-gated
Sign in instead

We only use this to review your request and reply.