For assessment leads
For whoever owns assessment in a residency program.
Verbal feedback already happens. Preceptors and trainees talk after almost every encounter. DebriefMD turns those conversations into the assessment record your program needs, without asking a preceptor to open an app, hold an account, or type a word.
The problem
Assessment has not caught up with how teaching happens.
Nobody in a residency program wants worse feedback. The system they have simply makes good feedback expensive to record. Four things follow from that, and program directors will recognise all four.
The friction lands on the preceptor, the person with the least time and the least to gain from the form.
A preceptor who has just given five minutes of excellent feedback is then asked to open a form, find the right trainee, pick the right EPA, and write it all again. Almost none of that step adds anything the conversation did not already contain.
Residents end up chasing preceptors.
When the form is a separate task done later, completion depends on someone remembering. That turns the trainee into a collections agent for their own assessment record.
Competency forms turn judgment into checkboxes.
A long grid of scales and required fields rewards finishing over describing. What gets recorded is a shape the form accepts, and the specific, useful observation that was said out loud rarely survives it.
Late forms carry worse feedback.
Residents batch their assessment requests and file late. Forms completed late carry lower-quality feedback, while the verbal feedback that prompted them held its value regardless of when it happened. What decays is the record, not the teaching.
None of this is a motivation problem. There is no capture mechanism at the moment the feedback happens, so everything downstream is reconstruction.
From the literature
This is a documented problem, not a hunch.
Documentation is a separate task from teaching. It happens later, from memory, and it repeats for every form, every resident, every block. That is the arithmetic behind forms being batched at the end of a rotation, thinned to a sentence, or never filed at all. Each of those has been measured.
Residents and faculty describe a burden of assessment under competency-based medical education that can undermine its benefits. Programs are actively adapting to reduce it.
A Canadian general internal medicine program read three years of its own completed assessments. Only 26% carried feedback specific enough to act on, and 11% named an actual improvement strategy. Residents and preceptors in the same study said the narrative comments were the part that mattered to them. It is the sharpest measure available of what the form does to what was said.
Residents interviewed about EPA timing described asking for an assessment as an imposition on staff, and admitted to batching their requests and submitting forms late. Forms completed late carried lower-quality feedback. Verbal feedback held its value regardless of when it happened.
Learners describe written and spoken feedback as amplifying each other rather than substituting for each other. The spoken version carries the nuance; the written version is what they return to later.
Early evidence on AI-drafted feedback
Drafting from what was said helps. A human still has to decide.
None of these studies used DebriefMD. They tested the step it automates, turning spoken or written feedback into a draft, and each found both a gain and a limit.
In a randomized study at Yale, instructors turned an ambient transcript of their feedback into a written note with a language model’s help. Those notes scored higher on a feedback-quality scale than notes written unaided (median 3 against 2, out of 5). The learners were first-year medical students, and a small share of the drafted feedback was mischaracterised (6.8%) or invented (1.7%).
At Prisma Health, an emergency medicine competence committee compared resident summaries a language model drafted from educator feedback with human-written ones. Members rated the drafted summaries higher for quality, accuracy and specificity, but not for usefulness: 71% rated the human content good or very good, and 29% rated the drafted content unfavourable. The drafted summaries ran about five times longer.
At Penn State Health Family Medicine, a language model assigned milestone levels from written faculty feedback for 24 residents. Its ratings tracked the competence committee’s closely (r = 0.82) but agreed with them only moderately, and least well for communication.
DebriefMD does not try to write down everything that was said. The resident reads the draft, edits it, and decides what stands before anything is exported. What the form receives is still a human’s decision.
What leadership can open
Give your competence committee the words that were actually said.
A checkbox grid records a shape the form accepts. DebriefMD keeps what the preceptor said, dated when they said it. The resident reviews and signs off before program leadership can open it. Leadership opens only what a resident has signed off and chosen to share with the program.
What you get
What a pilot measures.
Completion rate
Preceptors give feedback verbally, as they already do. The conversation and the record happen in the same few minutes, and finishing one does not depend on getting hold of anyone afterwards.
Narrative specificity
What a preceptor actually said is what gets written down, in their words rather than a sentence composed weeks later from memory. One debrief can fill as many EPAs as the conversation genuinely covered. The resident picks the forms they are working on, and each is filled only where the conversation supports it.
Earlier signal
A structured assessment can be ready minutes after a shift ends rather than at the end of a block. A competence committee reading current feedback is reading something it can still act on.
Getting started
A one-block pilot.
The pilot runs for about a month, over a cohort your program chooses. That gives a clean before-and-after window without committing the whole program to it.
Meet with us and scope the pilot
One conversation sets the shape: about a month, a focused cohort of residents and preceptors your program picks. We agree on what the pilot is meant to answer before anyone records anything. Pilots are paid engagements, and scoping covers what one would cost.
Drop into your existing forms
DebriefMD exports to PDF or Excel, or emails the PDF. Send us your form and we build it as a template. A resident can select several at once, so one debrief covers the EPAs it touched rather than one form per conversation. You decide which metrics matter, and we refine them with you as the block runs.
Run the block
Residents sign up, record, and export in the first week. No IT project, no new system for preceptors to learn, and no software for your program to maintain.
Read the block and decide
At the end of the block we share the readout and you decide whether to continue. Expanding means more residents, more preceptors, and more form templates: the same product, run wider.
Possible metrics
A handful of measures, chosen with you.
These are starting points rather than a fixed set. Programs tell us which questions the pilot has to answer, and we refine the measures together before the block starts.
Number of assessments
How many finalized assessments the cohort produced during the block, set against the equivalent block in your own records.
Time from encounter to documented feedback
Median hours between the conversation and a signed-off assessment being exported.
Words of narrative feedback per assessment
A simple proxy for how rich the documented feedback is, compared to your baseline.
Preceptor and resident satisfaction
Short end-of-block survey covering perceived burden, usefulness, and willingness to keep using it.
We share the readout with the program at the end of the block. You decide whether to scale.
Programs provide their own baselines. DebriefMD holds no history from before the pilot, so every before-and-after comparison is measured against figures your program supplies.
What we do not do
DebriefMD is one narrow thing.
It turns a conversation that already happened into a draft assessment the resident edits and signs off. That is the whole product. Four things it is not, said here rather than discovered in month three.
Not a curriculum system.
DebriefMD holds no objectives map, no logbook, and no record of what a rotation is supposed to teach. It does not know what your program's curriculum says, and it does not check anything against it.
Not a scheduling system.
It does not know who is on which service, it does not assign anyone to anything, and it sends no reminders about work that has not happened. A resident opens it because they just had a conversation worth keeping.
Not your evaluation system of record.
Your program's record of a trainee stays where it already is. DebriefMD produces a finished document (an Excel file and a PDF), and where that document goes afterwards is your program's decision, not ours.
Works alongside the system you already run.
DebriefMD does not connect to or replace One45, Elentra, MedSIS or any other assessment platform. It produces a reviewed PDF and an Excel file, and your program decides where they go.
Keep your assessment system. Stop losing the feedback.
DebriefMD is not a platform to migrate to. It produces the assessment at the moment of feedback, and once the resident signs it off it exports as a PDF or Excel file that can go wherever your program already keeps assessments. Start with one program and one block.
Buying across more than one program? See For PGME offices.