Healthcare Software Development

Healthcare software in India is a different discipline from healthcare software anywhere else, and the difference is not the medicine. It is that the identity layer is national, the consent layer is now statutory, and the economics of a clinic are decided by a number most software does not even measure.

ABDM and ABHA: identity you do not own

The Ayushman Bharat Digital Mission gives a patient an ABHA number — a portable health identity that is theirs, not your hospital's. Linking your records to it means your facility must register in the ABDM ecosystem, implement the Health Information Provider flows, and — this is the part teams underestimate — honour a consent request from a Health Information User you have never heard of, correctly, on demand.

The mental shift is real. Your patient record is no longer a row in your database that you control. It is a record the patient can grant a different hospital access to, for a defined purpose, for a defined window, and revoke afterwards. Software written on the assumption that the hospital owns the data cannot be retrofitted to this cleanly. We build the consent artefact as a first-class entity from the first sprint, and every read of clinical data is checked against a live consent grant, not against a role.

The DPDP Act made consent a legal artefact, not a checkbox

Under the Digital Personal Data Protection Act, health data is sensitive, consent must be specific, informed and revocable, and you must be able to demonstrate — not assert — what a patient agreed to and when. A checkbox in a form, with no record of what the notice said at the moment it was ticked, does not do that.

So we version the consent notice itself, store the exact text the patient was shown alongside the grant, timestamp it, record the purpose, honour revocation by actually cutting access rather than setting a flag, and log every access to sensitive data in an append-only audit trail that an engineer cannot quietly edit. Purpose limitation is enforced in the data layer. This is more work. It is also the difference between a compliance story and a compliance claim.

No-shows are the number that decides whether a clinic is profitable

A specialist consultation slot that goes unfilled is revenue that cannot be recovered — the hour is gone. Indian outpatient no-show rates commonly sit between 20 and 30 percent, and most hospital software does not measure it at all, let alone act on it.

What actually works, in the order it works: a WhatsApp reminder 24 hours out with a one-tap confirm or reschedule, because SMS is ignored and calls do not scale. A second nudge on the morning of. A waitlist that automatically offers a released slot to the next patient the moment a cancellation lands. Deposit-backed booking for the appointment types where no-shows concentrate. And a risk score per patient built from their own history, so the front desk knows which slots to overbook. We have taken no-shows from 27 percent to 11 percent doing exactly this, and nothing on that list is clever — it is just measured, and then acted on.

HL7 v2 and FHIR, in the real world

Your radiology system speaks HL7 v2 with pipes and carets, because it was installed in 2016 and nobody is replacing it. ABDM speaks FHIR R4. The lab machine speaks something the vendor invented. We build an integration layer that terminates all three, maps to a canonical internal model, and does not let the ugliest of the three dictate the shape of your entire domain. Every inbound message is stored raw before it is parsed, because the day a mapping is wrong you will want the original.