What a NABH assessor asks to see on screen, and which of it your system has to produce unaided

NABH assessor evidence software requirements, read from the other side of the table: what gets asked for on screen, and what your system should produce without help.

Ask a quality head in any 120-bed hospital what the six weeks before an assessment look like and you will get the same answer. The committee calendar is reconstructed. Indicator sheets are rebuilt from registers that were never designed to be summed. Somebody goes through two years of email looking for the corrective action that was agreed after an incident, and finds the agreement but not the verification. The file is assembled, the assessment happens, the certificate arrives, and the hospital is no better at knowing its own infection rate on an ordinary Tuesday than it was before.

So the practical question behind NABH assessor evidence software requirements is not which standard applies. It is this: of everything an assessor can reasonably ask to see on screen, how much does your system produce unaided, with timestamps it did not get to choose after the fact? Answer that honestly and you will usually find the standard was never the problem. NABH publishes its standards openly, and the 6th edition hospital accreditation standard is a free download that sets out the objective elements and how they are scored. Every quality head we meet has read it. The gap is almost never knowledge.

Why the evidence gets rebuilt instead of accumulated

Three mechanisms do most of the damage, and none of them is laziness.

The first is that the evidence an assessor wants is longitudinal and the systems holding it are not. An objective element assessed on implementation wants to see something happening month after month, with a trend and a response to the trend. A hospital whose indicators are computed once a year, in March, for a file, has no trend. It has a number. The number is usually defensible and almost never useful.

The second is that the record is split. Admission sits in one system, pharmacy in another, the OT register in a book, the infection control log in a spreadsheet on one laptop, biomedical maintenance with the vendor. Any evidence that crosses two of those has to be assembled by a person. Assembly by a person is reconstruction, and reconstruction is what an experienced assessor is trained to notice.

The third is that quality work has no write-time enforcement. Nothing stops a discharge summary going out without the prescriber's statutory identity on it. Nothing stops a record request from drifting past the statutory window. The control exists as a circular, the circular exists as a PDF, and compliance is therefore a matter of memory under load.

Five things worth doing that cost nothing

• Pick five indicators and compute them monthly. Not fifteen. Five that your committee will actually discuss, computed on the same definition every month, including the months that look bad. A twelve-point trend with one honest dip is stronger evidence of a functioning system than a flawless annual figure.
• Put CAPA in one dated register with three columns that are usually missing: who owns it, when verification happened, and what the verification found. Most hospitals can show the corrective action. Far fewer can show the check that it worked.
• Make the record-release clock visible. One list, requests with dates, counted in hours remaining rather than days elapsed. If nobody can see the clock, nobody can escalate before it lapses.
• Stop back-dating anything. A minute written late and marked late is evidence. A minute written late and dated early is a finding waiting to be made.
• Name a data owner per chapter. Not a committee. One person who can say where each number comes from and reproduce it in front of someone else.

None of that requires a procurement cycle. It requires deciding that the evidence is for you rather than for the assessor.

Where a system should be carrying this

What software can legitimately change is the share of evidence that is a by-product of ordinary work rather than an artefact produced for an occasion. On our compliance page we hold ourselves to two words with fixed meanings: supported means implemented and demonstrable, and by design means the platform produces the required behaviour or evidence as a by-product of normal operation. We publish those definitions because the words are otherwise free.

In Ospia HOS, NABH evidence is produced that way. CAPA, a live risk register, HAI surveillance and antimicrobial stewardship run as part of daily operations, and the indicator packs sit on the same record rather than on an export of it. One patient record across 31 domains means cross-domain evidence does not need assembly. NMC rules are enforced at the point of writing: generic-name prescribing, prescriber statutory identity on every document, serialized certificates, and the 72-hour record-release clock escalated before it lapses rather than audited after. ABDM and ABHA are supported, and every use of personal data traces to recorded consent held in a consent ledger that is consulted before use, not reconciled afterwards.

Alongside that sits a fleet of watchdog agent definitions in the codebase, 17 of them, including quality, infection control, compliance and biomedical watches. They are governed the way you would govern a new hire: a defined scope, permissions, an escalation path, and an autonomy level the hospital sets per task from L0, where a person does the work and the system records it, to L4, where the system acts inside boundaries the hospital wrote and every run still leaves a replayable ledger entry and a reversal path. Every AI response carries its recommendation, the reasoning, the evidence, a confidence, alternatives considered, policy references and the expected outcome. For an assessment conversation that distinction matters: you can show not only what was done but why, months later. For a medical director weighing a multi-speciality deployment or reading how we compare with Napier, that replayability is the part worth testing hardest in a demonstration.

What we will not tell you is how an assessment went. Hospitals run on Ospia HOS today, and the clinical core is live, but we publish no case studies and no measured outcomes until a hospital has completed hypercare and agreed to be named. Accreditation is awarded to a hospital on the strength of what it does, not to its software. Any vendor who implies otherwise is selling you a certificate they cannot issue.

If you want to see which of your evidence a system can produce unaided, the useful test is a live one. Ask for the executive demonstration and bring three objective elements you expect to be hard.