What ABDM integration involves
ABDM integration connects your HMIS, LIS or EMR to the National Health Authority’s network in milestones. M1 creates and verifies patients’ ABHA numbers and addresses. M2 makes the facility a Health Information Provider: it links each visit as a care context and sends encrypted FHIR R4 records when the patient consents. M3 makes it a Health Information User that requests records with consent. New work uses ABDM’s v3 APIs and must pass functional testing and a WASA security audit before NHA issues production keys.
Most ABDM projects stall between a sandbox demo and production keys
Creating a test ABHA in the ABDM sandbox can be done in a day or two. Getting production credentials takes months, and the work in between is what most estimates leave out: consent rules applied to every record that leaves the building, FHIR bundles that pass validation, encryption both sides agree on, callbacks that survive retries, NHA’s functional tests and a web application security audit. Since October 2026 NHA has also warned that idle sandbox accounts can lose access, so a stalled project now has a clock on it.
We work with two kinds of buyer: HMIS, LIS and EMR vendors who need ABDM inside their product, and hospitals, nursing homes and labs whose software vendor has not delivered it. Either way, the sandbox account, the production credentials, the Health Facility Registry entries and the auditor’s contract stay in your organisation’s name. We write the code in your repository and work inside your sandbox account. We are not an NHA testing agency or a security auditor on CERT-In’s list, and that is deliberate: both roles have to be independent of the people who built the software.
ABDM work sits inside our FHIR interoperability and HL7 integration practice. If your records start life as HL7 v2 messages, our page on HL7 to FHIR conversion covers the mapping; for India’s data-protection duties, see our DPDP Rules checklist.
What is ABDM? The building blocks your software talks to
The Ayushman Bharat Digital Mission (ABDM), launched in September 2021 and run by the National Health Authority (NHA), is India’s federated network for health records. NHA runs the registries and the switch in the middle; the records stay with the hospital, lab or clinic that created them and move only when the patient consents. A PIB backgrounder of 6 July 2026 counts over 104 crore health records linked to over 93 crore ABHA accounts. For a software team, ABDM is a handful of components to integrate with:
| Building block | What it is | What your HMIS does with it |
|---|---|---|
| ABHA number | A 14-digit health ID issued after KYC, one per person | Creates one at the registration desk or verifies an existing one (M1) |
| ABHA address | The patient’s handle on a consent manager, such as name@abdm (name@sbx in the sandbox) | Captures and verifies it, then links the patient’s records to it |
| Health Facility Registry (HFR) | The national directory of hospitals, clinics, labs and pharmacies (facility.abdm.gov.in) | Each customer facility registers and, in production, links its profile to your bridge ID so the network can reach it |
| Healthcare Professionals Registry (HPR) | The directory of verified doctors, nurses and other professionals (hpr.abdm.gov.in) | Doctors at participating facilities enrol; M4 brings registry look-ups and enrolment into your software |
| Consent manager (HIE-CM) | Holds the links between ABHA addresses and care contexts, and every consent the patient grants or revokes | Receives your new-record notices; sends you consent notifications and artefacts |
| Gateway | NHA’s hosted switch for discovery, linking, consent and data-flow messages | Your gateway client calls it; it calls back your bridge URL |
| PHR app | Where patients approve consents and see records: Aarogya Setu 2.0 since 29 June 2026 | Test scripts and desk instructions must name the right app; ABDM advised the separate ABHA App would close in production from 31 October 2026 (Elets eHealth, 27 October 2026) |
ABDM milestones M1, M2, M3 and M4 explained
The sandbox groups integration work into milestones. The names below are the ones on the sandbox’s own registration and exit forms, read on 7 November 2026.
| Milestone | Name in the sandbox | Role | What you build |
|---|---|---|---|
| M1 | ABHA Creation/Verification | Registration desk | ABHA creation through Aadhaar OTP and other modes, verification of an existing ABHA, ABHA address capture, the ABHA card and QR code, usually Scan & Share intake |
| M2 | Building Health Information Provider (HIP) | Record source | Care-context linking (started by the hospital or by the patient), consent notifications, FHIR bundles, encryption, data push and transfer status |
| M3 | Building Health Information User (HIU) | Record consumer | Consent requests, status tracking, consent artefacts, data requests, decryption and a viewer for clinicians |
| M4 | National Healthcare Provider Registry (HPR/HFR) | Registry integration | Searching and registering facilities and professionals from your own software |
Which milestones you need depends on who you are. NHA’s guidelines for HIPs and HIUs say an HMIS or LIS used by facilities must comply with at least M2. The incentive scheme goes further for hospitals and clinics: under DHIS Corrigendum 7 their transactions count only on software integrated up to M3, while labs, diagnostic centres and pharmacies qualify at M2. In practice we plan M1 to M3 for an HMIS, M1 and M2 for an LIS or radiology system, and add M4 only when the product onboards facilities or doctors itself.
Scan & Share is the registration-desk feature most hospitals notice first. The patient scans the facility’s QR code in a PHR app, the gateway sends the ABHA profile to the HIP through a profile-share callback, and the HMIS issues a queue token without retyping demographics. PIB counted over 23.21 crore ABHA-linked tokens issued by 18 June 2026. We build it alongside M1.
ABDM v3 APIs: what changed and what to migrate
All new ABDM work should target the v3 APIs. In early November 2026 the sandbox home page told integrators still on the V1 or V2 ABHA APIs for M1 to migrate “to avoid deactivation of your credentials”, and DHIS Corrigendum 7 stopped incentives from July 2026 for any software not on v3 for all of M1, M2 and M3. The v3 conventions we build to:
- Gateway calls sit under
/api/hiecmon the sandbox and production hosts, starting with a token from/gateway/v3/sessionsusing your client ID and secret. - Every request carries
REQUEST-ID(a fresh UUID),TIMESTAMP(UTC),X-CM-ID(sbxin the sandbox,abdmin production) and, for facility calls,X-HIP-IDorX-HIU-ID. - Your callbacks live under
/api/v3/on the bridge URL you register with the gateway. - ABHA enrolment and login move to
/abha/api/v3/, for exampleenrollment/request/otpandenrollment/enrol/byAadhaar, with the Aadhaar number and OTP RSA-encrypted with the ABHA service’s public key before sending. - In DHIS Corrigendum 7’s own summary, v3 adds standardised time formats, better validation and the ability to state the category of health record being created.
| Flow | Your system calls | The gateway calls you at |
|---|---|---|
| Hospital-initiated linking | /v3/token/generate-token, then /hip/v3/link/carecontext | /api/v3/hip/token/on-generate-token, /api/v3/link/on_carecontext |
| New-record notice | /hip/v3/link/context/notify, or the SMS notice when only a mobile number is known | /api/v3/links/context/on-notify |
| Patient-initiated linking | The matching on-discover, on-init and on-confirm responses | /api/v3/hip/patient/care-context/discover, /api/v3/hip/link/care-context/init and /confirm |
| Consent at the HIP | /consent/v3/request/hip/on-notify | /api/v3/consent/request/hip/notify |
| Sending records (HIP) | /data-flow/v3/health-information/hip/on-request, a push to the HIU’s data URL, then /data-flow/v3/health-information/notify | /api/v3/hip/health-information/request |
| Consent at the HIU | /consent/v3/request/init, /consent/v3/request/status, /consent/v3/fetch | /api/v3/hiu/consent/request/on-init, /notify, /api/v3/hiu/consent/on-fetch |
| Fetching records (HIU) | /data-flow/v3/health-information/request | /api/v3/hiu/health-information/on-request, then the HIP’s push to your data URL |
Paths as configured in NHA’s open-source ABDM Wrapper (v3 profile). Check them against the current sandbox documentation before you code; ABDM revises them.
If your product went live on v0.5 or V1/V2, treat migration as a small project of its own: new paths and headers, new callback routes, a regression run against the current test cases, and a check that your WASA report is still valid, because DHIS pays a software company only for transactions after a valid WASA report has been submitted.
Consent artefacts, encryption, callbacks and idempotency
Consent artefacts. A consent request names the purpose, the patient, the requester, the HIU, the record types wanted and a permission block: access mode (view, store, query or stream), date range, a dataEraseAt date and frequency. When the patient approves in their PHR app, the consent manager notifies the HIP with an artefact listing the care contexts covered. The HIP must store it, share exactly the record types and dates it allows, and stop on expiry or revocation. The HIU must delete fetched data by dataEraseAt. Expect testers to probe exactly these rules.
Encryption. Records never travel in clear between HIP and HIU. ABDM uses the Fidelius scheme: each side makes a fresh key pair on Curve25519 and a 32-byte random nonce; the sender derives a shared secret by ECDH, combines the two nonces, derives an AES key with HKDF and encrypts each bundle with AES-GCM. The HIU sends its public key and nonce with the data request, and the HIP returns its own with the pushed data. Key-encoding mismatches between Java, .NET, Node.js and PHP libraries are a common first bug.
Callbacks. ABDM is asynchronous. Most calls are acknowledged at once and the real answer arrives later at your bridge URL, which must be public HTTPS, answer fast and pass the work to a queue rather than calling the HMIS inline. A HIP also has to stay up: the guidelines say ABDM measures HIP uptime with heartbeats, and an on-premises HIP needs reliable connectivity and a static public IP.
Idempotency. Retries happen on both sides. Key every inbound message on its REQUEST-ID and its transaction or consent ID, store it before acting, and make handlers safe to run twice: a repeated consent notification must not create a second consent, and a repeated data request must not push the same bundle twice. Keep every request and response ID in the log; you will need them whenever you raise a ticket with NHA support.
FHIR R4 records: the NRCeS profiles ABDM expects
Records shared through ABDM are FHIR R4 document bundles that follow the NRCeS FHIR Implementation Guide for ABDM, version 6.5.0, the current published version on 7 November 2026 (FHIR 4.0.1). Each record is a document bundle whose first entry is a Composition profiled for one record type.
| Record type | NRCeS profile | Typical source in an HMIS | DHIS category (Corrigendum 7) |
|---|---|---|---|
| DiagnosticReport | DiagnosticReportRecord (with DiagnosticReportLab or DiagnosticReportImaging) | LIS result, radiology report | Category 1 |
| DischargeSummary | DischargeSummaryRecord | IPD discharge module | Category 1 |
| Prescription | PrescriptionRecord | OPD or discharge prescription | Category 2 |
| OPConsultation | OPConsultRecord | OPD consultation note | Category 2 |
| ImmunizationRecord | ImmunizationRecord | Vaccination module | Category 2 |
| WellnessRecord | WellnessRecord | Vitals and device readings | Category 2 |
| HealthDocumentRecord | HealthDocumentRecord | Scanned or uploaded documents | Category 2 |
| Invoice | InvoiceRecord (added in IG 6.0.0 in January 2024 to bring pharmacies on board) | Pharmacy or consultation bill | Category 2 |
- Reference resources inside a bundle with
urn:uuidfull URLs, as the guide has done since 6.0.0. - Use LOINC for lab tests and SNOMED CT where the guide binds it; a local test code alone does not meet the guide’s bindings.
- A PDF attachment is allowed to start with, but the HIP guidelines expect fully structured data within one year of certification. Plan the structured version from day one.
- Validate every bundle against the guide’s package in CI, not just in the sandbox. Elements marked Must Support have to be readable by your HIU even when other senders leave them out.
How an HMIS plugs into ABDM: a reference architecture
Whether you build into the HMIS, run NHA’s ABDM Wrapper (a Spring Boot service with MongoDB that covers the M2 and M3 workflows) or put an integration engine in front, the same ten components appear. We keep them in a separate ABDM service beside the HMIS, so the HMIS core changes as little as possible and the ABDM side can be released on its own schedule.
| Component | Job | Design notes |
|---|---|---|
| 1. Gateway client | Session tokens, v3 headers, retries with back-off, request logging | One code path for sandbox and production, switched by configuration |
| 2. Callback receiver (bridge) | Public HTTPS endpoint for every /api/v3 callback | Validate, store, acknowledge, queue; never block on the HMIS |
| 3. ABHA link store | ABHA number and address per HMIS patient, verification and KYC status | Your own patient ID stays primary; ABHA is voluntary for patients |
| 4. Care-context index | One entry per visit, admission or lab order, with a readable name and record types | The unit patients discover, link and consent to |
| 5. Consent store | Artefacts, status, expiry, revocations, erase dates | Every outbound record is checked against it |
| 6. FHIR bundle builder | Maps HMIS data to NRCeS profiles and validates the result | A template per record type, terminology mapping tables, validation in CI |
| 7. Document storage | Signed PDFs and images referenced or embedded in bundles | Retention set by NHA guidance and your written retention schedule |
| 8. Crypto service | Fidelius key pairs, encryption and decryption | Fresh keys per request; key material never logged |
| 9. Transfer workers | HIP pushes and transfer notices; HIU receipt and decryption | Idempotent and resumable, with status per care context |
| 10. Audit log and monitoring | Who asked for what, under which consent, and when it was sent | Feeds NHA queries, DHIS claims and DPDP accountability |
For labs with HL7 v2 feeds, the mapping belongs in an integration engine. A Mirth Connect or Open Integration Engine channel takes ORU^R01 results from the LIS, maps each OBX code to LOINC, builds the DiagnosticReportRecord bundle and hands it to the ABDM service with the care-context reference. The same channel keeps feeding hospitals over HL7 v2, so nothing upstream changes.
From sandbox to production: the ABDM exit process
- Register on the sandbox. The software owner applies at sandbox.abdm.gov.in with entity details (PAN, GSTIN or CIN) and a category such as HMIS or diagnostic lab. NHA’s Health Tech Committee reviews the application and issues sandbox credentials.
- Build and self-test. Each milestone is built against the sandbox gateway with test ABHA addresses ending in @sbx and the sandbox PHR app.
- Functional and non-functional testing. The exit form describes testing by an NHA team against published test cases; you upload the functional testing report and certificate.
- Security testing. A web or mobile application security audit by an agency empanelled by CERT-In or STQC, ending in a Safe-to-Host certificate, which the exit form calls the WASA certificate. STQC may also run validation tests on sample applications.
- Exit readiness form. The test report, the Safe-to-Host certificate, an application summary sheet and a signed undertaking.
- NHA approval. A final approval round by NHA’s internal team; production client ID and secret are then sent by email.
- Go-live per facility. Each customer facility registers on HFR and links its profile to your production bridge ID; its doctors enrol on HPR.
Two recent rules matter. NHA’s advisory of 14 October 2026 says integrators inactive for a year or more without milestone progress can be classified inactive and lose sandbox APIs, credentials and applications. And WASA has to stay current: under DHIS Corrigendum 7, a software company earns nothing for transactions made before a valid WASA report is submitted.
Radiatus does not certify software or perform WASA audits, and is empanelled for neither. We prepare the evidence, run dry runs against the test cases, fix findings and sit with your team through NHA’s questions; you contract the testing agency and the auditor directly.
Typical effort for ABDM integration, by milestone
Typical engineer-day ranges for a web HMIS with an API layer, confirmed after a specification review and then quoted as a fixed price per phase. They exclude NHA review time and the auditor’s own work.
| Work package | Typical effort | What moves the number |
|---|---|---|
| Foundations: gateway client, bridge, queue, audit log | 8–15 engineer-days | Whether the HMIS already has an API layer and a job queue |
| M1: ABHA creation and verification, address, card, Scan & Share intake | 15–30 engineer-days | Number of creation modes, changes to registration screens, card printing |
| M2: linking, consent, encryption, data flow, first two record types | 30–55 engineer-days | How cleanly the HMIS models visits and orders; data quality |
| Each further record type | 5–12 engineer-days | Structured or PDF-first; terminology mapping |
| M3: consent requests, data fetch, decryption, viewer | 20–40 engineer-days | How many record types the viewer renders, and where in the clinician’s workflow |
| Test dry runs and NHA testing support | 10–20 engineer-days | Rework found by the test cases |
| WASA remediation | 5–20 engineer-days | Your current security baseline |
| HL7 v2 lab feed to DiagnosticReportRecord | 10–20 engineer-days | Variety of LIS and analyser messages; LOINC mapping |
M1 to M3 with two record types usually adds up to about 90–180 engineer-days. Calendar time from sandbox access to production keys is typically four to seven months, and a fair share of it is waiting for sandbox approval, testing slots, the auditor and committee review. Booking the auditor early is the cheapest schedule saving.
Is the Digital Health Incentive Scheme still running?
DHIS, launched on 1 January 2023, pays facilities and digital solution companies for eligible ABDM transactions above 100 a month. The latest instrument we found is Corrigendum 7, dated 9 April 2026 (copy circulated by Rajasthan’s health department), covering April to September 2026 “subject to the availability of funds”. Its main terms:
- Only records linked to KYC-verified ABHA addresses count; Scan & Share and health-locker linking earn nothing.
- Diagnostic reports and discharge summaries (Category 1): ₹10 to the facility and ₹5 to the software company per eligible transaction. Other record types (Category 2): ₹5 and ₹2.50. The first 100 transactions each month are deducted.
- Consent-based sharing: ₹10 to the HIU and ₹5 to each record source per successful consent; auto-consent requests, and transactions where the HIP and HIU are the same, do not count.
- Each ABHA counts for at most one transaction a day and five a month per facility or software company.
- From July 2026, no incentive for a facility or software company unless the software is on v3 for M1, M2 and M3; software-company transactions before a valid WASA report are excluded.
- A ceiling of ₹5 crore per facility or software company.
PIB reported payouts by 18 June 2026 of over ₹107 crore to hospitals, ₹2.95 crore to diagnostics, labs and pharmacies, and over ₹26 crore to software companies. On 7 November 2026 we found no notice extending DHIS beyond September 2026, so check abdm.gov.in/dhis before incentive income goes into a business case. Build ABDM because customers and patients need it, and treat DHIS as a bonus.
ABDM consent and the DPDP Act are not the same thing
ABDM’s consent manager controls sharing over the network. It does not make your HMIS compliant with the Digital Personal Data Protection Act, 2023. The DPDP Rules were notified on 13 November 2025; the rule on registering DPDP consent managers applies from 13 November 2026 and most duties of data fiduciaries from 13 May 2027. A DPDP consent manager is a separate legal role from ABDM’s HIE-CM. Where the two meet in an ABDM project:
- Consent and notice at the desk. The HIP guidelines already require informed consent before assisted ABHA creation; DPDP adds the notice content.
- Data minimisation. Ask only for the record types and date range a clinician needs, set an erase date, and do not copy fetched records into the HMIS without a reason.
- Security and breaches. Encrypt linked records and consent artefacts at rest, keep access logs, and write the breach runbook before go-live.
- Children. The Rules’ Fourth Schedule lets clinical establishments and healthcare professionals process a child’s data without verifiable parental consent only as far as needed to provide health services.
- Retention. ABDM expects records to stay available for the long term; DPDP expects erasure once the purpose ends. Reconcile the two in a written retention schedule.
Common reasons ABDM integrations fail testing or stall
These are the problems we design against. Most follow directly from the guidelines, the consent rules and the v3 conventions above.
- Still calling V1/V2 ABHA or v0.5 gateway paths.
- Callbacks that time out because they query the HMIS before acknowledging.
- Care contexts that are too coarse (“all records”) or too fine (one per lab parameter), so patients cannot tell what they are consenting to.
- Sharing outside the consent: extra record types, dates outside the range, or data sent after revocation.
- Bundles that fail validation: missing mandatory elements, local codes where LOINC or SNOMED CT is bound, references that are not
urn:uuid. - Encryption that the other side cannot decrypt, usually key encoding or nonce handling.
- Retried messages processed twice.
- Registration screens that push patients into giving an ABHA, when the guidelines say it is voluntary and normal registration must carry on.
- No audit trail of who viewed which fetched record.
- Security findings: weak session handling, no rate limit on OTP requests, secrets in code or logs, outdated libraries.
- Facilities not linked to the production bridge ID in HFR, so nothing reaches patients after go-live.
- Server clocks drifting enough to fail timestamp checks.
ABDM integration services
Five engagements, each scoped after a specification review and priced per phase. Credentials, registrations and audit contracts stay in your name.
ABDM integration for HMIS, LIS and EMR vendors
The problem
Hospital customers want ABDM, your team is busy with the roadmap, and nobody has taken a product through sandbox exit before.
How we do it
We build M1 to M3 on the v3 APIs as a separate ABDM service beside your product, in your repository and your sandbox account: gateway client, bridge, consent store, FHIR bundle builder and the encryption, with automated tests mapped to NHA’s test cases. Your engineers review every merge, so they can run it after handover.
What you get
- M1, M2 and M3 on the v3 APIs
- FHIR bundles for your first record types
- Test suite mapped to NHA’s test cases
- Exit pack and handover runbook
ABDM for hospitals on an existing HMIS
The problem
Your HMIS vendor cannot say when ABDM will arrive, and the registration desk wants Scan & Share now.
How we do it
The hospital registers as the integrator and we build a sidecar ABDM service that reads from the HMIS through database views, its APIs or HL7 feeds and handles ABHA, linking and sharing on its behalf. The same approach works for open-source EMRs; see our OpenEMR services.
What you get
- Sidecar ABDM service beside your HMIS
- Scan & Share and ABHA at the desk
- Linked discharge summaries and reports
- A clean exit if your vendor ships ABDM later
FHIR bundle builder for the NRCeS profiles
The problem
You have the gateway calls working, but bundles fail validation or render badly at the other end.
How we do it
A bundle builder for the eight NRCeS record types with mapping tables for LOINC and SNOMED CT, validation against the guide’s package in your CI, and sample bundles from your own anonymised data for the testers.
What you get
- Builders for the record types you need
- Terminology mapping tables
- Validation in CI
- Sample bundles for testing
HL7 v2 lab feeds to ABDM through Mirth or OIE
The problem
Your LIS already sends HL7 v2 results to hospitals, and you do not want a second integration path for ABDM.
How we do it
A Mirth Connect or Open Integration Engine channel that takes the same ORU messages, maps results to LOINC, builds DiagnosticReportRecord bundles and passes them to the ABDM service with the right care context, while the existing HL7 feeds carry on unchanged.
What you get
- Engine channel with tests
- OBX to LOINC mapping
- DiagnosticReportRecord output
- Monitoring and alerting on the channel
v3 migration and sandbox-exit support
The problem
Your integration is on older APIs, failed functional testing, or has WASA findings nobody has time to fix.
How we do it
We move gateway and ABHA calls to v3, run the current test cases end to end, fix what fails, close WASA findings with your auditor’s retest in mind and assemble the exit readiness pack.
What you get
- v3 migration plan and code
- Test-case dry run report
- WASA findings closed
- Exit readiness pack
Where ABDM projects usually start
Illustrative starting points, not client case studies.
| Organisation | Starting point | Scope we would usually propose |
|---|---|---|
| HMIS vendor with dozens of hospital customers | Customers asking for ABDM and DHIS; no sandbox account yet | M1 to M3 on v3 in a separate ABDM service; discharge summaries and prescriptions first, other record types as later releases |
| Multi-specialty hospital on a vendor HMIS | The vendor has no ABDM date; the hospital wants Scan & Share and ABHA-linked discharge summaries | The hospital registers as the integrator; a sidecar ABDM service reads the HMIS through database views or HL7 feeds; M1 and M2 first, M3 next |
| Diagnostic lab chain | LIS with HL7 and ASTM analyser feeds; reports go out on paper and WhatsApp | M1 and M2 with DiagnosticReportRecord bundles fed by a Mirth or OIE channel from HL7 v2 results |
| Product live on older APIs | Running on V1/V2 or v0.5; DHIS payments stopped in July 2026 | v3 migration, regression against the current test cases and a WASA validity check |
| Telemedicine or PHR app | Wants to fetch patients’ earlier records with consent | M1 and M3 with a viewer for NRCeS bundles; M2 too if it issues e-prescriptions |
Illustrative plan: an HMIS vendor taking M1 to M3 live
Illustrative, not client work. Assumes a web HMIS with an API layer and one engineer from your side; NHA and auditor timings vary and often set the pace.
Specification review and sandbox access
Map the data model, registration flow and hosting, agree record types and order of work. You apply for sandbox access in your organisation’s name; we set up the gateway client and bridge when credentials arrive.
M1 on the v3 ABHA APIs
ABHA creation through Aadhaar OTP and mobile, verification of existing ABHAs, ABHA address capture, the card and QR at the desk, and Scan & Share intake.
M2: the HIP
Care-context model, hospital-initiated and patient-initiated linking, consent store, bundles for the first two record types, encryption, data push and transfer notices.
M3: the HIU
Consent requests from the clinician’s screen, artefact handling, data requests, decryption and a viewer for the NRCeS bundles.
Dry runs and NHA testing
We run the current test cases end to end and fix what fails, then support your team through NHA’s functional testing.
WASA and exit
Your chosen auditor tests the application; we fix findings and assemble the test report, Safe-to-Host certificate, summary sheet and undertaking.
Production and first facilities
Production credentials configured, first facilities linked to your bridge ID in HFR, monitoring and a runbook handed to support. Further record types follow as releases.
An ABDM-facing product we build ourselves
Proof of engineering, not of certification: it runs against the ABDM sandbox, and we say so.
Specifications and sources we build against
Primary sources, read on 7 November 2026. ABDM revises its APIs and rules often, so we recheck them at the start of every project.
NRCeS FHIR IG for ABDM 6.5.0
Profiles for the eight record types on FHIR R4 (4.0.1), with terminology bindings, examples and a downloadable package for validation.
ABDM sandbox v3 documentation
Current v3 flows, Postman collections and NHA’s test cases for M1, M2 and M3.
NHA ABDM Wrapper
Open-source Spring Boot service for HIP and HIU workflows, including the Fidelius encryption; a useful reference even when you build your own.
Guidelines for HIPs and HIUs
NHA’s obligations for providers, repository providers, users and PHR apps, version of 24 October 2022.
DHIS Corrigendum 7
Rates, record categories and the v3 and WASA conditions for April to September 2026.
Sandbox activity advisory
NHA’s 14 October 2026 notice on inactive integrator accounts losing access.
Tools we use alongside it
- HL7 FHIR Validator
- HAPI FHIR
- Mirth Connect / Open Integration Engine
- Postman
- BouncyCastle
- PostgreSQL
- Docker
- OpenTelemetry
Facts, versions and vendor details checked 7 November 2026.
Guides and related pages
Guides
Questions about ABDM integration
They are the stages of integrating software with the Ayushman Bharat Digital Mission. M1 covers ABHA creation and verification at registration. M2 makes the facility a Health Information Provider that links records to the ABHA and shares them when the patient consents. M3 makes it a Health Information User that requests and views records with consent. The sandbox also lists M4, for integration with the healthcare professional and facility registries.
For a web HMIS with an API layer, M1 to M3 with two record types typically takes about 90 to 180 engineer-days of build and test work. Calendar time from sandbox access to production keys is usually four to seven months, because sandbox approval, NHA testing, the security audit and committee review all take time outside your control.
NHA guidelines expect an HMIS or LIS used by facilities to comply with at least M2. For incentives, DHIS Corrigendum 7 counts hospital and clinic transactions only on software integrated up to M3, while labs, diagnostic centres and pharmacies qualify at M2. Most hospital HMIS vendors therefore build M1 to M3, and most lab systems build M1 and M2.
Version 3. In early November 2026 the sandbox asked anyone still using the V1 or V2 ABHA APIs for M1 to migrate to v3 to avoid deactivation of their credentials, and from July 2026 DHIS pays nothing unless the software is on v3 for M1, M2 and M3. Treat v0.5 and V1/V2 as legacy.
WASA is the web application security assessment required before sandbox exit. An agency empanelled by CERT-In or STQC tests the application and issues a Safe-to-Host certificate, which you upload with the exit form. Under DHIS Corrigendum 7, a valid WASA report is also needed for software companies to earn incentives. Radiatus does not perform WASA audits; we fix the findings before and after the test.
Records follow the NRCeS FHIR Implementation Guide for ABDM, version 6.5.0 on FHIR R4 as of November 2026. It defines document profiles for diagnostic reports, discharge summaries, prescriptions, OP consultation notes, immunisation records, wellness records, health documents and invoices, each built as a FHIR document bundle.
The latest notice we found is Corrigendum 7 of 9 April 2026, covering April to September 2026 subject to funds. As of 7 November 2026 we have not found an extension beyond September 2026. Check the DHIS page on abdm.gov.in before counting on incentive income.
Yes. The hospital can register as the integrator and run a separate ABDM service that reads from the HMIS through database views, APIs or HL7 feeds. That service goes through the same sandbox testing, security audit and NHA approval as any other software, and each facility still needs its own Health Facility Registry entry.
No. ABDM consent governs sharing records over the ABDM network. The Digital Personal Data Protection Act and Rules govern all your processing of personal data, with most duties applying from 13 May 2027. You need both: a proper notice and consent at registration, minimal data requests, security safeguards, breach handling and a retention schedule.
Not as HL7 v2 messages. ABDM exchanges FHIR R4 bundles, so lab results must be converted into a DiagnosticReportRecord bundle with LOINC-coded results. An integration engine such as Mirth Connect or the Open Integration Engine can build those bundles from the same HL7 v2 feed your LIS already sends to hospitals.
Planning ABDM for your HMIS, lab or hospital?
Tell us your product or HMIS, how it is hosted, which milestones you need and where you are in the sandbox. A senior engineer will reply with the scope we would propose, the order of milestones and a typical effort range, confirmed after a specification review.






Compliance-first delivery