What Did NHA Announce at Aarogya Manthan 2026, and What Does It Change for a PM-JAY Claim?

2. Auto-adjudication: what a machine already decides

PM-JAY handles nearly 50,000 claims a day across more than 1,900 treatment packages ^[4]. Every one is a hospital asking to be paid for treating a patient who paid nothing at the door. Checking them by hand is slow, and the wait falls on hospitals.

In April NHA said 15-20% of claims were auto-adjudicated ^[4]. At the event, BusinessLine reports, Barnwal put it at about 30%, with more than 60 triggers flagging suspicious claims for a person to look at ^[2]. Both can be true. A claim that passes through automated checks is not necessarily a claim that a machine approved, and the second number may count the first kind. Until the framework is published, treat 15-20% as the figure for claims decided without a human.

The direction is not in doubt. Automation so far covers high-volume, low-value claims, and NHA wants it on "high value, complex claims too" ^[2]. The checks it runs are mostly unglamorous: a package billed for the wrong sex, a procedure that can only be done once claimed a second time ^[10]. In 2025-26 NHA penalised 2,842 hospitals ₹114.06 crore and stopped ₹678.47 crore of suspicious claims before payment, according to reports of its annual report ^[10].

For a patient, faster routine approvals are good news. The risk is the opposite case, a legitimate but unusual claim that trips a rule. What should matter most in the framework, when it appears, is how a hospital appeals a machine's flag and how long that takes. Nothing published so far says.

3. Thirteen crore patient records, opened carefully

Eight years of PM-JAY have produced more than 13 crore patient records ^[1,2]: who was admitted, where, for what, at what package rate. For anyone studying what hospital care in India costs, or which treatments fail, nothing else comes close.

NHA's design keeps the records inside its walls. Built with IISc Bengaluru, the environment lets approved entities either work with de-identified data or send their code in to run against it, with results checked before release ^[1]. That second model, where the researcher never holds the data, is the stricter of the two.

What is missing is everything a researcher would need to apply. There are no published eligibility rules, fees, review criteria or start date ^[3]. The government lists "responsible AI innovation" among the purposes, so health-tech companies are evidently in scope, not only universities ^[1]. How commercial access will be governed is the question we would most like answered.

4. Paying only for standard treatment

ICMR has written Standard Treatment Workflows for 157 conditions across 28 disciplines ^[5,1]: flowcharts of symptoms, tests and treatment, condition by condition. ICMR's Director General Rajiv Bahl argued at the event that tying them to money would give them force. Under his proposal, reported by BusinessLine, a hospital would be paid for treatment that follows the approved pathway and not for treatment that does not ^[2]. "This is a two-way street," he said.

That sounds radical. It is closer to a formalisation of what already happens. The Department of Health Research's annual report says NHA "is also using the STW as the clinical guidance document to prepare their reimbursement package", that ICMR draws its topics from NHA among others, and that the two renewed an agreement to rationalise PM-JAY reimbursement rates and implement the workflows ^[5,6]. The very first volume, in 2019, listed supporting PM-JAY as an objective ^[5,6]. Health technology assessments have already removed drugs that were not cost-effective from PM-JAY and changed cancer treatment protocols ^[5]. NHA's own hackathon this year made checking claims against treatment guidelines its first problem to solve ^[11].

The workflows themselves pull the other way. Volumes III and IV each open with the same sentence: "These broad guidelines are advisory", and "there may be variations in the management of an individual patient... as decided by the treating physician" ^[7,8]. A payment rule built on an advisory document has to decide what happens when the physician is right to depart from it. Our view is that a tied-payment rule without a fast route for justified exceptions would mostly hurt the sickest patients, whose cases fit flowcharts least well.

The four volumes are below, unchanged from ICMR's files, so you can see exactly what a hospital would be measured against.

• Volume I: 53 conditions (PDF, 16 MB)
• Volume II: tuberculosis, 18 conditions (PDF, 4.6 MB)
• Volume III: 51 conditions in 11 specialties (PDF, 25 MB)
• Volume IV: 32 conditions (PDF, 38 MB)
• Department of Health Research Annual Report 2025-26 (PDF, 18 MB), pages 10-11 and 237 for the PM-JAY link

5. If your parent is 70 or older

Every Indian aged 70 and above is eligible for PM-JAY, regardless of income ^[12]. More than 1.36 crore had an Ayushman Vay Vandana card by 21 September, against about 6 crore eligible ^[9]. That is roughly 23% ^[9].

Everything above applies to their claims: the automated checks, and any future link between payment and standard treatment. Older patients are also the ones most likely to arrive with several conditions at once, which is exactly where a single-condition flowchart is weakest. None of it changes the cover itself. The ₹5 lakh, how it is shared within a family, and how it sits alongside a private policy are explained in government health schemes for senior citizens.

6. Why private policyholders should pay attention

IRDAI's chairman, Ajay Seth, spoke at the same event. He asked for wider use in health insurance of ABHA as an identity, codified treatments, standard billing formats, and standard treatment protocols and guidelines ^[1]. With 13.25 crore admissions behind it ^[9], PM-JAY's rules are the obvious template for everyone else.

A private insurer that adopts standard protocols could use them the way Bahl described, to decide what it pays for. Arguments over whether treatment was necessary already sit behind many refused claims; see why senior health claims get rejected. A published protocol would at least put that argument in writing, where both sides can see it.

We will update this page when NHA publishes the auto-adjudication framework or opens the Secure Data Environment to applications, and we will host the framework here the day it appears.