EMR implementation is a sequencing problem, not a software problem. A two-doctor clinic can be off paper in about two weeks. A 50-bed hospital needs six to ten. The order that works is the same at both sizes: master data first, then a pilot department, then a capped parallel-running period, then everything else.
Key takeaways
- Two doctors: roughly two weeks. Fifty beds: six to ten. Installation is one day of that; the rest is masters, migration and training.
- Build master data before you migrate anything. Services, tariffs, doctors, roles, and the ABDM facility and practitioner IDs if you’re joining.
- Migrate the active panel and about twelve months of history. Not the archive. Keep the paper registers.
- Parallel running needs an end date before it starts. Two weeks is usually enough.
- Month 1 to 3 breaks at the front desk and the billing tally, almost never at the doctor’s screen.
Key facts
- Two-doctor clinic, master data to retired register
- ~2weeks
- 50-bed hospital, master data to stabilisation
- 6-10weeks
- Time you get to produce records on request, IMC 2002 reg 1.3.2
- 72hours
What implementation actually is, once the sales language comes off
Installing the software takes an afternoon. For a cloud system it takes a login. What takes weeks is everything the software cannot do for you: deciding what your service list is called, agreeing on the tariff for a dressing, cleaning a patient list with four spellings of the same surname, and getting a receptionist to trust a screen over a register she has run for eleven years.
That’s the whole job. Cloud or an on-premise box decides who patches the server and where the data sits. It changes almost nothing about the list above. Our cloud versus on-premise breakdown covers that decision separately; this piece assumes you’ve made it.
A realistic timeline: two doctors versus fifty beds
Here’s how we’d sequence it at both ends of the range. The phases don’t change with size. The duration and the number of people in each one do.
| Phase | What it covers | 2-doctor clinic | 50-bed hospital |
|---|---|---|---|
| Master data | Services, tariffs, doctors, users, roles, departments, beds | Days 1 to 3 | Weeks 1 to 2 |
| Data migration | Active panel, last 12 months, outstanding balances | Days 3 to 5 | Weeks 2 to 3 |
| Core training | Front desk first, then billing, then doctors | Day 5 | Week 3, repeated per shift |
| Pilot department | One OPD or one specialty, real patients | Days 6 to 8 | Week 4 |
| Parallel running | Both systems, capped, daily tally | Days 8 to 12 | Weeks 5 to 6 |
| Full go-live | Registers retired, one fallback day only | Day 12 | Week 7 |
| Stabilisation | Daily fix list, weekly review, master data cleanup | Weeks 3 to 6 | Weeks 8 to 12 |
Add two weeks to the hospital column if you’re also doing IPD, pharmacy and insurance claims in the same rollout. We’d argue you shouldn’t, at least not in the same month.
The hospital timeline stretches for one reason: shift work. A clinic trains two doctors and one receptionist once. A 50-bed hospital trains three shifts of front desk, a billing team, nursing on two floors, and visiting consultants who come twice a week and were not there on training day.
Master data first, because everything downstream inherits its mistakes
This is the least interesting part of the project and the one that decides whether it works. Get it wrong and you’ll be correcting bills in month three.
What has to exist before a single patient is entered: the service list with the exact names your staff already say out loud, the tariff for each service including any package and insurance rate, every doctor with their registration number, user accounts with roles that actually restrict what a receptionist can see, departments, and bed numbers if you run IPD.
Name one owner for this. One person, named, whose job it is to approve a new service or tariff, before and after go-live. The commonest failure we see is that master data has an owner during the project and nobody after it.
If you plan to join ABDM, register here rather than after go-live. Ayushman Bharat Digital Mission is run by the National Health Authority and participation is voluntary; nothing forces a private clinic to integrate. Facility and healthcare-professional registration now both run through ABDM’s National Healthcare Providers Registry, and the older facility and professional registry addresses redirect there. Those IDs sit in the same doctor and facility records that appointments, prescriptions and reports all reference, so doing it later means touching every one of them again. Our ABHA linking walkthrough covers the day-to-day flow.
One more decision worth making now: coding. The MoHFW’s Electronic Health Record Standards for India, 2016 name SNOMED CT as the primary clinical terminology, LOINC for tests, measurements and observations, and the WHO ICD family for disease classification. You don’t need to code everything on day one. You do need masters that can hold those codes, because retrofitting them into three years of free text is not a project anyone finishes.
Data migration: the decision is what to leave behind
Almost every clinic starts by asking how to get ten years of records into the new system. Wrong question. The right one is how little you can carry and still run Monday’s OPD.
| What you’re leaving | Carry it in | Leave it where it is |
|---|---|---|
| Paper OPD register | Active panel demographics, phone, allergies, current medicines | Old consult notes; keep the register itself |
| Excel patient list | The full file, after de-duplicating on phone number | Free-text notes columns that no one will search |
| Old on-prem system, clean export | Demographics, appointments, last 12 months of visits, outstanding balances | Deep clinical history unless it exports structured |
| Old on-prem, no usable export | Rekey the active panel only, over two weeks | Everything else; keep the old machine read-only |
| Lab and imaging reports | Scan and attach for active patients only | The archive shelf |
| Outstanding dues | All of it, reconciled to the cash book on one date | Settled ledgers older than the current financial year |
De-duplicate on phone number before the import runs. Cleaning it up afterwards costs far more. Two records for one patient is the single most expensive mistake in this table, because it splits the clinical history and nobody notices for months.
Two rules sit behind that table. First, the paper doesn’t get destroyed. Under regulation 1.3.1 of the Indian Medical Council (Professional Conduct) Regulations 2002 a physician maintains indoor-patient records for three years from the start of treatment, and under 1.3.2 those records must be produced within 72 hours of a request. A migration that quietly leaves a gap is a problem you discover at the worst possible moment.
Second, more is not safer. The Digital Personal Data Protection Act 2023 ties consent to a specified purpose and to the personal data necessary for that purpose, and requires erasure once the purpose is no longer served, except where another law in force requires you to keep it. Record-retention rules like the one above are that exception, so erasure is a decision you make record by record, not a switch you throw. Bulk-loading everything you happen to hold because it might be useful someday is the opposite of that posture. Our DPDP guide for clinics has the longer version, and our security page sets out how we handle it on our side.
Parallel running, and how to stop it
Run both systems for two weeks. Front desk enters in the EMR and writes the register. Billing raises the bill in the EMR and tallies it against the cash book at close. It’s genuinely annoying, which is the point: parallel running that isn’t annoying doesn’t end.
Set the end date before you start. Then pick a measurable signal for hitting it. Ours is simple: three consecutive days where the day’s collection total comes out of the EMR and matches the cash book without anyone reconstructing anything. Hit that, stop the paper the next morning.
Clinics that skip the end date are still double-entering in month four. At that point the staff have concluded the EMR is extra work rather than the work, and you have lost the rollout without anyone announcing it.
Training, and the front-desk problem
Picture a 40-bed hospital in Nashik. Two receptionists, three shifts between them and a relief, a billing clerk, six resident doctors, and visiting consultants across ortho, gynae and general medicine. The doctors will be fine. They spend two minutes a visit in the system and their alternative was already dictation or a hurried scrawl.
The front desk is where this succeeds or fails. Their resistance is arithmetic. The senior receptionist registers a walk-in in about fifteen seconds with a pen. The new screen takes her ninety on day one and there are nine people waiting. Telling her the system is better is useless. Getting her to sixty seconds by day four is not.
So: train the front desk first, longest, and on real patients rather than dummy data. Name one super-user per shift. Train billing separately, because their failure mode involves money. Give doctors twenty focused minutes, not a workshop. And book fewer patients for week one, because week one is slower for everyone.
Go-live day
Pick a Monday. We think that beats a quiet Saturday. Call it a preference, but the reasoning is practical: a Monday gives you a full week of vendor support, your own staff at full strength, and four working days to fix whatever surfaces before the weekend. Going live on a Saturday means finding the problem on Saturday evening and living with it until Monday.
What we’d want in place before the first patient: a printed one-page cheat sheet at each counter, the vendor’s support person on a number that actually answers, a lighter appointment list than usual, the paper register on the counter as a fallback for exactly one day, and one person walking the floor whose only job is to answer questions.
Then keep a written fix list. Not a WhatsApp group, a list. Every issue, who raised it, what happened. We’d budget for a few dozen items in the first fortnight at hospital scale, mostly tariff corrections and missing masters. The ones nobody writes down are the ones still broken in month three.
What actually goes wrong in month 1 to 3
The failures are boring and they repeat.
Tariffs turn out wrong at billing, usually a verbal discount the system doesn’t know about. Duplicate patients pile up because the front desk searches by name instead of phone. Doctors quietly go back to paper for “quick” cases, so the share of visits with a note in the system stalls well short of all of them. The month-end report doesn’t tally with the cash book and nobody can say why. And the champion who drove the project gets busy, master data ownership evaporates, and new services start getting added by whoever is at the counter.
None of that is a software defect. All of it is cheap to fix in the first ninety days and expensive after.
How to tell whether it worked
Four numbers, checked at day 30, 60 and 90. What share of visits have a note in the system, which should be near 100% by day 60 and tells you whether the clinical side actually adopted it. Whether the EMR’s day total matches the cash book three days running. How many duplicates the patient master has picked up, checked by counting repeated phone numbers. And how long a registration takes now against week one.
Add one more if documentation time was the reason you started: how much of the charting is still happening after the last patient leaves. That’s the number clinicians care about, and it’s the one a record system alone rarely moves.
When you should not implement now
Being straight about this matters more than a feature list.
Don’t start if you’re weeks away from a NABH assessment or any audit. NABH, run under the Quality Council of India, publishes hospital standards and separate entry-level standards for small healthcare organisations, and preparing for an assessment on a system your staff learned last week is a bad trade. Finish one, then start the other.
Don’t start in your busiest season, don’t start without a named master data owner, and don’t start if the person championing it is leaving in two months. A rollout eats slack you may not have. Those are the three cheapest ways to find out.
And don’t replace a working system purely on price. Migration has a real cost in staff hours that no quote shows you; our EHR cost guide breaks down where the money actually goes. If your current EMR stores records fine and the pain is the hour of typing after the last patient, you have a documentation problem, not a records problem, and swapping the record system won’t touch it.
Where we fit, and where we don’t
If the documentation hour is the real bottleneck, the cheaper move is to add a layer rather than replace the system. That’s a two-week change instead of a two-month one, and our small-clinic scribe rollout guide walks through it.
AI Scribe by Patient Square is an ambient AI medical scribe that listens during the visit and hands back a structured SOAP note, ICD-10 suggestions, and a prescription draft — ready to review and sign about two minutes after the visit.
That module runs alongside whatever you implement. There’s no HL7 or FHIR interface between us and your EMR; every plan includes “EHR-ready export (PDF · HL7 · FHIR)”, which is an export, not a certified interface. If you’d rather have the record system bundled, our Copilot tier includes the AI Copilot EHR. India pricing per clinician per month on annual billing, ex-GST: ₹1,599 for Assist, ₹2,399 for Copilot, ₹3,999 for Autopilot. Add 18% GST and the Assist invoice is about ₹1,887. The pricing page has what sits in each tier.
Where we’re the wrong answer: if live ABHA linking has to work on go-live day, we can’t do that today. ABDM integration is on our roadmap and not shipped, and we’re not ABDM-certified. eka.care already offers live ABHA linkage, and if that’s a hard requirement for your facility, they’re the better fit. We’d rather say so than let you find out in week six.
The honest verdict on the whole exercise: EMR implementation succeeds or fails on master data, a hard stop to parallel running, and the front desk. Not on the feature list you compared. If you want to sanity-check your own shortlist first, our EMR buyer’s guide for India covers what to ask vendors before any of this starts.
If documentation is the piece you want to test on a real OPD day before committing to anything larger, compare the options and bring a typical Monday list.