Peer outcome benchmarks for specialty units, computed inside each hospital
TL;DR Normwell lets cardiac surgery, dialysis and ICU units in India see how their risk-adjusted mortality, readmission and infection rates compare with peer units. The statistics are computed on each hospital's own server and only aggregates leave. Hospital quality teams pay, because they collect indicators today with nothing to compare them against.
Indian hospitals measure outcomes but rarely know whether those outcomes are any good.
An audit of six hospitals preparing for NABH accreditation found quality indicators "collected but not standardized", clinical audits incomplete, and infection surveillance started but never analysed[1]. In paediatric cardiac care, a 2026 letter notes that only 20–30 of roughly 40–90 tertiary centres report patient-level data to multicentre studies. A pooled analysis of 30,587 patients put operative mortality at 5.63%, well above Western databases, and the author says the lack of national data blocks benchmarking[2].
Why not just pool the data? Hospitals will not hand raw records to a rival-run registry, so the comparison never happens.
What it does, in regulatory terms - quality-improvement analytics on aggregate statistics only. It gives no patient-level advice, screening or diagnosis, so medical-device rules should not apply. It still processes personal health data on site under the DPDP Act and Rules[3].
A small node sits inside each member hospital and reads a fixed outcome dataset for one specialty - case mix, procedure, 30-day death, readmission and infection. It computes counts, rates and risk-model terms locally. Only those aggregates, with small cells suppressed, travel to a coordinator that fits a shared risk model and returns each unit's observed-to-expected ratio against the peer group. Specialty societies set the definitions.
The seed pitched federated learning. Peer comparison does not need a trained model, only well-defined statistics, so I've reframed it as federated analytics. That is simpler to build and easier to explain to a hospital's legal team.
The user is the unit head and the hospital's quality manager. The payer is a hospital group's quality or clinical-excellence function, on an annual per-unit subscription, since NABH-accredited hospitals already have to monitor quality indicators[1]. A second route is a specialty society running a benchmarking programme for its members. No Indian price exists to anchor on, though US benchmarking programmes are sold to hospitals as paid participation[4][5].
The DPDP Rules were notified in November 2025, with eighteen months of phased compliance and penalties of up to ₹250 crore for weak security safeguards[3]. That makes pooling raw records harder to approve and on-site computation easier to defend. Federated setups have also run in India before: a hospital chain, a chipmaker and an imaging-AI platform linked hospital clusters in three cities in 2022[6].
There is no public figure for what Indian hospitals spend on outcome benchmarking. The bottom-up count is narrow - tens of tertiary cardiac centres[2] plus the larger dialysis and ICU chains. Abroad, one large US benchmarking database covers 1,600+ organisations and 250+ health systems[5], which shows hospitals pay for peer comparison once trust exists. Treat that as proof of behaviour, not your market size.
ACS NSQIP in the US runs a risk-adjusted surgical registry built on clinical records, with 30-day follow-up[4]. The STS National Database publishes cardiac surgery outcomes voluntarily and warns against naive hospital-to-hospital comparisons[7]. A global research-data network runs a federated network across 14,200+ clinical sites in 20+ countries, aimed at research rather than quality[8]. In India, the cardiothoracic surgeons' association has started a national surgical database[2]. The gap is a neutral, record-free benchmark that Indian hospital groups will actually join.
| Difficulty | Hard (52/100) |
| Starting capital | ₹1.6–3 Cr |
| Time to launch | 12–18 months |
The software is modest. The hard part is getting clean, comparable data out of messy hospital systems, agreeing definitions with clinicians and building a risk model surgeons accept as fair. Selling is slow because no hospital wants to join before its peers do.
No licence to operate, but data protection is the heavy part.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't trust AI-coded data without human checks, because bad coding makes a good unit look bad. And keep outputs aggregate, since patient-level alerts could bring in device rules.
Want to work out the AI layer for Normwell? Talk to Yash →
Here's what can go south -
I think this suits a founder with clinical-quality or registry experience and the patience for consortium selling. If you're hoping to sell AI then it's the wrong idea for you. Prove one thing first - that five units in one specialty will sign a shared data definition and let a node run on their servers.
Plan for roughly ₹1.6–3 Cr, much of it the 12–18 months of data work, clinician meetings and sales before a peer group is big enough to sell.
No sector licence, but you must comply with the DPDP Act as a processor and get a regulatory opinion before adding any patient-level feature.
Yes, and models like ACS NSQIP already exist in the US. Check local health data law and whether a national registry already covers your specialty.
The free Normwell validation kit has the line-by-line capital breakdown, the effort scoring, the data-protection and accreditation checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for quality heads and unit chiefs, and 5 assumptions to test. Download the kit →
Thinking about building Normwell or something close to it? The make-or-break work is recruiting the first peer group and agreeing definitions that clinicians trust. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.