Skip to main content

ABDM Gateway Timeouts: Continue OPD and Reconcile Exchange Requests

An ABDM exchange timeout means the clinic has not received a confirmed result; it does not prove that the remote action never happened. Keep appropriate local registration and care workflows usable, show the missing information clearly and assign someone to reconcile the exchange later. Retry only under the supported integration rules, recheck current consent and distinguish a logical request from its repeated attempts. A recovered connection does not mean every record has arrived or been clinically reviewed.

By Bractis · Updated · 10 min read

For Indian clinic managers and doctors planning how an external health-exchange interruption should affect a busy OPD.

Identify what has actually failed

Separate the clinic’s internet connection, the local application, identity services, consent handling and an external provider’s response. An exchange panel that is slow does not necessarily mean the consultation record cannot be saved. Equally, a functioning internet connection does not prove the external record is available.

Use a status explanation tied to the task: local draft saved, external request pending, permission unavailable or report not received. Avoid a single red ‘system down’ message when only one dependency has failed. Staff need to know which work can proceed and which result cannot yet be relied upon.

Do not infer national outage from one failed request. Confirm the scope through the provider’s support and monitoring process. The clinic can record the observed failure time and affected task without making an unsupported public statement about the entire ABDM ecosystem.

MoHFW describes ABDM record exchange as consent-based. An outage does not remove that requirement or justify bypassing the supported access process. Any alternative source of information still needs appropriate identity, authority and clinical review.

Source: MoHFW: ABHA-linked records and consent-based exchange, 22 May 2026

Continue the appropriate local workflow and expose the information gap

Keep a returning patient on the correct local chart. A failed external lookup should not force creation of another patient file. Use the clinic’s ordinary registration and identity procedure, and tell the clinician which outside information remains unavailable.

The clinician decides whether available history, patient-held documents and the clinical assessment are sufficient for the intended care or whether another action is needed. Record the limitation and the resulting decision. Reception should not label a missing report as normal or tell the patient it has been reviewed because a request was sent.

If the patient supplies a paper or PDF copy, retain its source, date and verification context through the normal import process. It is a separate evidence route, not proof that the failed exchange succeeded. Later arrival of the same report needs comparison so one investigation does not become two clinical events.

Explain the fallback in plain language: today’s local record can continue where supported, the external report is still unavailable and a named person will handle the agreed follow-up. Avoid promising a recovery time that the clinic or provider has not established.

Distinguish a known failure from an unknown remote outcome

A request may be rejected before it is applied, accepted with a delayed response or processed while the acknowledgement is lost. These cases require different recovery decisions. Keep a stable reference for the logical task so the next operator can investigate it without issuing a new task blindly.

The table is an original operational state model. Map it to the deployed API and provider documentation; the labels below are not asserted to be official ABDM status codes. The important feature is that each state explains both the evidence and the next responsible action.

On a narrow screen, scroll the table horizontally to read all columns.

Exchange interruption states and their next action
Observed stateWhat is knownNext action
Not submittedNo request was sent from the local workflow.Assess current need and permission before submitting.
Explicitly rejectedA specific request was refused with a known reason.Resolve the reason; do not repeat unchanged input indefinitely.
Timed out or acknowledgement missingThe local system lacks a confirmed outcome.Reconcile by request reference before creating duplicate work.
Partial receiptSome identified records arrived; others remain missing.Preserve received items and target the unresolved part.
Late response after a status changeData arrived after consent, identity or clinical context changed.Restrict routine use pending applicable reconciliation and review.
Transport recoveredConnectivity works again.Check request completion and clinical review separately.

Use controlled retries rather than repeated clicks

RFC 9110 distinguishes operations that can safely be repeated from non-idempotent requests that should not be automatically retried without additional knowledge. It is general HTTP guidance, not an ABDM endpoint specification. The practical question for the provider is whether repeating this particular operation can create a second effect.

Ask how the integration identifies one logical request across attempts, detects an already applied result and handles duplicate or out-of-order responses. Do not prescribe a made-up retry interval or assume every endpoint accepts an idempotency key. Use the actual current contract and supported recovery mechanism.

Bound automatic recovery so an unavailable service does not receive an uncontrolled burst of repeated attempts. Staff should see whether a retry is scheduled, exhausted or awaiting a decision. A user clicking Try again should not unknowingly multiply the same work in several queues.

Reassess current permission and identity before any later attempt. The patient may have withdrawn sharing consent, the clinic may have corrected a link or the original request may no longer be needed. An old authorised intention is not enough to justify every future retry.

Keep sensitive response bodies and credentials out of ordinary error reports. A useful incident reference can connect the provider’s restricted diagnostics to the clinic’s task without exposing a patient record in a public support message.

Source: IETF RFC 9110: idempotent methods and automatic retries

Source: OWASP: Logging Cheat Sheet

Recover with a worklist that belongs to someone

At restoration, retrieve the unresolved logical requests, not merely the number of failed network calls. Assign each request to an operator who can confirm its outcome and each received clinical record to the clinician responsible for review. A technically complete transfer can still leave a clinically important result unattended.

Compare late arrivals with patient-supplied copies and earlier imports using source, identity, document date and version. Similar filenames are weak duplicate evidence. Preserve a corrected report as a distinct version, and make the clinician aware of a meaningful change rather than discarding it as another copy.

Close an item only against its defined endpoint. A cancelled request needs a cancellation outcome; a required report needs receipt and the agreed review or follow-up. If the clinician no longer needs the original retrieval, record that decision instead of keeping a retry running indefinitely.

Retain the distinction between outage recovery and ordinary record retention. An incident log should not become a permanent archive of copied clinical payloads. Keep only the diagnostic and accountability evidence required by the applicable policy.

Reconcile requests, attempts, documents and review separately

In a fictional outage, ten logical retrieval requests generate 16 technical attempts. After recovery, six requests are fulfilled, two are cancelled after current need or permission changes and two remain unresolved. Accounted-for request outcomes total ten, but only six are fulfilled. The 16 attempts are a technical workload measure, not 16 patients or 16 reports.

The six fulfilled requests return eight documents. The clinician has reviewed five of those documents, leaving three for review. Request fulfilment is 6/10, or 60%, while document review is 5/8, or 62.5%, among received documents. These denominators answer different questions and should not be blended into one reassuring completion percentage.

Add a late duplicate response to the exercise. The system should recognise the already handled result while retaining relevant event evidence. Add a corrected report as a separate case; it must not disappear merely because the source and date resemble an earlier file.

Use the worklist to identify the next action for each unresolved item. If staff cannot name who owns the three unreviewed documents, connectivity recovery has not restored the full care workflow.

Test failure and recovery before depending on exchange

Bractis’s local product catalog marks ABDM/FHIR integration Planned. This guide defines evaluation scenarios; it does not claim production gateway availability or a released offline capability. Confirm the exact environment, supported functions and limitations in the proposed configuration.

In an authorised test setup with fictional data, simulate a delayed response, an explicit rejection, partial receipt, a consent change and a duplicate completion. Keep the local consultation open during the exercise and inspect whether its draft and patient identity remain intact. Restore the dependency and follow the worklist to its actual endpoints.

Do not disconnect a live clinic’s gateway or alter real consent merely to create a dramatic demonstration. The provider should offer controlled failure simulation or explain what can be verified through test evidence. Record untested cases honestly.

After a real interruption, review which information staff lacked, whether repeated clicks increased the backlog and whether patients received the agreed follow-up. Use those observations to improve the local procedure and product requirements, without treating one successful drill as a service-level guarantee.

Source: Bractis: ABDM and interoperability scope

Fictional clinic example: the report arrives twice after the visit

A request times out during consultation. The doctor uses an attributable patient-held report and records the remaining information limitation. After connectivity recovers, the exchange returns the same report twice. The responsible operator relates both responses to the original request, avoids creating duplicate clinical events and routes any genuinely new or corrected information for review. The clinician’s earlier decision remains recorded in its original context. Transport success does not retrospectively mean the report was available during the visit.

Exchange Outage and Recovery Worksheet

Use these prompts in your clinic’s own notes. Record the result, responsible person and next action for each item.

Failure scope
Distinguish local saving, identity, consent and external retrieval, and write the status explanation staff and patients should receive.
Retry evidence
Ask how one logical request survives repeated attempts and how the provider detects an already applied result.
Recovery ownership
Assign request reconciliation and clinical document review separately, including late and partial arrivals.
Completion measures
Report logical requests, technical attempts, received documents and reviewed documents with their own denominators.

Questions doctors ask

Does an ABDM timeout mean the request failed remotely?

Not necessarily. It means the local system lacks a confirmed response within its waiting period. The remote action may be pending or already applied. Reconcile the request reference using the supported mechanism before creating duplicate work.

Can the clinic continue OPD while an external exchange is unavailable?

Use the appropriate local workflow where supported and make the missing information visible. The clinician decides whether available evidence is sufficient or another action is needed. A gateway problem should not silently create a new patient chart or a false reviewed-result status.

Should reception keep clicking retry until the report appears?

No. Follow the supported retry and reconciliation process with a stable task reference, bounded attempts and current permission checks. Repeated clicks can duplicate work or obscure what already happened.

Is the workflow complete when the gateway connection returns?

No. Reconcile request outcomes, confirm which documents arrived and assign any necessary clinical review and patient follow-up. Connectivity, receipt and clinical completion are separate milestones.

Sources and scope

Updated on . The state model, worklist and calculations are original operational analysis. Primary references support narrow ecosystem and HTTP distinctions; no current API conformance, production integration or recovery-time guarantee is asserted. No independent clinical or legal review is claimed.

Try the workflow
with Bractis.

Simulate an exchange timeout, a late duplicate response and a consent change using fictional data, then follow the recovery worklist through clinical review.

Request a clinic demo