How to plan a healthcare MVP that can actually ship
Healthcare MVPs die in a predictable way. The first version tries to be a platform: every specialty, every EHR, a patient app, a billing module and an AI feature, all before a single clinician has used it on a real afternoon. The build runs long, the integration is still theoretical, and the thing you meant to learn never gets learned. A better MVP is smaller and ruder. It proves one workflow for one kind of user, and it is honest about the system that workflow has to touch.
Write the MVP as a sentence before you write a backlog. 'A clinician in this setting can finish this task, and the result shows up here.' For an ambient scribe, a workable sentence is: a doctor in an outpatient visit can leave with a draft note, review it, and file the accepted text into eClinicalWorks. That sentence has a user, a task and a destination. If you cannot say the destination, you do not have a product yet. You have a notebook.
Discovery is the work of making that sentence true. Sit with the people who do the job. Time the steps they already take. Collect the artifacts: the note template, the fields they retype, the point where they swear at the software. You are looking for effort you can remove, not for a feature list. Two or three observed sessions will beat a month of hypothetical user stories. Bring a clinical reviewer into that room. A product that saves clicks and adds risk will not be adopted, and the reviewer is how you notice.
Decide what the first version will not do, and write it down where investors and clinicians can both see it. Not every specialty. Not every EHR vendor. Not automated filing. Not a patient portal. Scope that is only in someone's head will leak back into the sprint the moment a demo approaches. The MVP plan is a boundary. Unicodez usually starts client work with a short discovery sprint and then a fixed-scope MVP for that reason. The same team should build it. Handing a prototype to a different group is how the boundary gets lost.
If the workflow ends in an EHR, the integration is inside the MVP, not a phase two you will 'get to'. We have done this with eClinicalWorks, through developer APIs the client provided, and we do not describe that as a universal EHR layer. Your plan should name the system, the access the client can actually give you, and the one write you will attempt. The public description of those APIs lives on eCW Connect, and the data model they refer to is FHIR R4. Use those documents to ask better questions. Do not paste them into a proposal and call the integration designed.
Put clinician review in the first clickable prototype, even if the prototype is only screens. A founder needs to see the moment a person accepts or throws away the model's draft. So does an investor. So does the clinic's privacy officer. HHS explains the Privacy and Security Rules that sit behind that moment in its HIPAA professional guidance. The prototype needs a screen where nothing reaches the chart until a person says so, and a one-page note of what data you will hold. If your company will see protected health information, the work sits under the client's BAA and HIPAA requirements.
Make the prototype something a user can click, with the real words of the workflow, empty states included. A pitch deck is not a prototype. The test is whether a clinician can narrate the visit using your screens without you narrating for them. Do this before you freeze the engineering scope. Changes are cheap while the prototype is still a sequence of screens and expensive once an EHR write is in progress.
The plan you take out of discovery should fit on a few pages. The user and the task. What is in and what is out. The system the data must reach, and the access the client will provide by a named date. The review step. The success check: not a vanity metric, but something like 'reviewers file the draft without retyping it' or 'the front desk stops keeping a side spreadsheet for this one job'. How you will know the MVP failed, which is just as useful. A fixed scope is allowed to be small. It is not allowed to be vague.
Sequence the build so a weekly demo is always a working slice. First the path with fake data and the review screen. Then the real read from the system, once credentials exist. Then the single write. Then the fixes the pilot group asks for. Resist the second specialty until the first one has been used on purpose. A pilot of one site and a handful of clinicians will teach you more than a soft launch to everyone.
Costs and calendar time vary with access, note complexity and how ready the client is. Anyone who quotes a universal 'healthcare MVP in six weeks' without those facts is guessing. What you can insist on is the shape: discovery, a clickable prototype, a fixed MVP, and the same team afterwards if you continue. That is the plan we use with founders and with clinics.
The products worth shipping are the ones a specific person will use next Tuesday. Name that person, name the system, and leave the platform for the version that has earned it.