Date of incident: Monday, 27 April 2026
Time: 10:13 BST – 11:48 BST (1 hour 35 minutes)
Status: Resolved
1. What happened?
On the morning of Monday, 27 April 2026, Accurx Scribe was unavailable for 1 hour, 35 minutes. During this window, users trying to start, continue, or complete a Scribe consultation saw a "Scribe authentication failed" error and could not use the feature. The disruption was caused by an issue with Tandem Health, the third-party provider that powers Scribe. No other Accurx products or features were affected.
2. What was impacted?
The disruption affected Accurx Scribe only, specifically, the ability to sign in to Scribe and carry out a consultation. While the incident was ongoing, users may have experienced any of the following:
A loading spinner icon was shown until a timeout after 30 seconds was reached, which then presented the "Scribe authentication failed" error message when opening Scribe.
A consultation that was already in progress would have seen the "Your internet connection seems to be slow, which may lead to delays" or "Reconnecting to Tandem" error messages, however the recording of the consultation would have continued.
Slow or unresponsive behaviour from Scribe in the moments before the full outage.
If a Scribe transcription was completed before the outage began at 10:13 BST, it remains safe and accessible.
3. What caused it?
The root cause of this incident was identified as a database issue at Tandem Health, our sub-processor for Accurx Scribe. A recent change on Tandem's side increased the amount of memory their database needed to handle each request. On Monday morning, when usage rose, a key database ran out of memory and became unresponsive.
Because Tandem's authentication service depends on that database, it began handing out invalid sign-in tokens to Accurx Scribe. Every Accurx Scribe session that was carried out during the outage received an invalid authentication token, which is why users saw the "Scribe authentication failed" error message.
In summary, a configuration change on our supplier's side, combined with a routine surge in Monday morning demand, overwhelmed the system that signs users in to Accurx Scribe.
4. What we did to fix it / What mitigations we have taken
As soon as the issue was reported, we:
Published an update on our public status page within 4 minutes of declaring the incident, and kept it updated as the situation evolved. We recommend that you subscribe to this service if you have not already.
Reviewed our own logs to confirm that the bad authentication tokens were coming from upstream (Tandem Health) and not from anything inside Accurx.
Engaged Tandem's engineering team directly through our shared engineering channel so we could coordinate live and share information on what users were seeing.
On Tandem's side, their engineers scaled up the affected database, reverted the change that had increased memory usage, and confirmed normal performance had returned. Service was confirmed restored at 11:48 BST, and the public status page was marked Resolved at 11:54 BST.
5. What's next?
As part of our incident response, we have liaised with our sub-processor, Tandem Health, to outline the preventative actions that will reduce the likelihood of similar incidents occurring in the future. These are:
Improving memory usage alerting. Tandem will lower the threshold at which they receive alerts when memory usage has grown. This will allow earlier detection of issues.
Remove inefficient database queries. Tandem will conduct an immediate review to identify and address slow or inefficient queries across its databases.
At Accurx, we are always working to improve our incident response process. In response to this incident, we will carry out the following changes:
Better in-product communication during Scribe disruption. Following feedback during this incident, we're reviewing how we surface live status information to clinicians inside the product, so users know what's going on.
Resilience review of our Scribe integration. Our engineering team is reviewing how Scribe behaves when Tandem is degraded, including how we communicate failures to the user and how we recover sessions, so that any future supplier issue is less disruptive to your day.
6. FAQs
How does this impact SLAs?
How does this impact SLAs?
Accurx upheld its incident response SLA for Severe (Severity 2) incidents. The incident was resolved within 1 hour 35 minutes, well within the 4-hour SLA.
Uptime for Accurx Scribe during April 2026 was 99.78%.
Why was no time to resolution provided in the status page?
Why was no time to resolution provided in the status page?
Our incident response teams work as swiftly as possible to resume services. As the nature of fixes changes between incidents, it’s incredibly difficult to provide an accurate estimated resolution time. Instead, we try to provide confidence through frequent updates, with transparent messaging and clearly defined impacts.
I was in the middle of a Scribe consultation when the outage started, what happened to it?
I was in the middle of a Scribe consultation when the outage started, what happened to it?
Users in an active consultation would have seen "Your internet connection seems to be slow, which may lead to delays" or "Reconnecting to Tandem…" error messages during the consultation, however recording would continue.
Clicking "Finish and create note" would have shown a Spinner icon “Sending audio" with a message. Trying to close the tab or navigate would show the warning message "Leave Site? Changes you have made may not be saved. Yes/No".
If a user maintain clicked no and kept the tab open, the note and transcript would have been created successfully after 11:48 BST.
What was the Clinical Risk?
What was the Clinical Risk?
This incident was assigned a clinical risk rating of 2 (Moderate). The full clinical risk assessment can be viewed below.
Has any patient or clinical data been lost?
Has any patient or clinical data been lost?
No. Our engineering team and Tandem (our Scribe provider) have both confirmed that no stored patient or clinical data was lost as a result of this incident. The disruption was limited to authentication, meaning Scribe couldn't sign users in, and didn't affect data already saved. The only thing that wouldn't have been captured is transcription work for consultations that were actively running during the 1-hour 35-minute window.
Were any other Accurx products affected?
Were any other Accurx products affected?
No. The disruption was limited to Accurx Scribe. Patient messaging, Patient Triage, Self-book/accuBook, Record View, the Inbox, and the rest of the Accurx toolbar and Accurx Web continued to work normally throughout. If you experienced an issue with another product on the morning of 27th of April, it wasn't related to this incident — please contact our support team and we'll investigate.
How will I know if Scribe is having problems again?
How will I know if Scribe is having problems again?
Our public status page is the fastest place to check — we publish updates there within minutes of declaring an incident, and we keep it updated throughout. We're also reviewing how we surface live status information inside the product itself, following feedback that some clinicians weren't aware the outage was happening until they tried to use Accurx Scribe.
7. Clinical Risk Assessment
7. Clinical Risk Assessment
This incident was assessed in line with DCB0129 clinical risk management.
Initial Risk Rating
Severity: Significant (as multiple users affected)
Likelihood: Medium
Clinical Risk Rating: 2
Affected Product/Module(s): Scribe
Clinical Impact Summary
Between 10:13am and 11:48am on Monday 27th April 2026, Accurx Scribe users experienced:
Authentication errors
Slow or failed page loads
Periods where core services were unavailable
At the peak of the incident, the platform was fully unavailable.
The primary impact was the inability of clinicians to use Accurx Scribe, forcing a temporary reversion to manual note-taking and causing minor disruption to workflows.
Clinical Risk:
A clinician relies on Accurx Scribe for a complex consultation and, due to the outage, is unable to generate a transcript. If they fail to take concurrent manual notes, important clinical details could be omitted from the patient record, potentially leading to misinformed follow-up care. Impact was limited to the duration of the outage.
No data was lost during the database failure; transcriptions resumed or were visible once service was restored, mitigating the risk of permanent record gaps.
During the outage, clinicians may not have had access to previous notes, relying on memory for the duration of the outage for consultations that had not been copied elsewhere already.
Existing Mitigations:
No data was lost during the database failure; transcriptions resumed or were visible once service was restored, mitigating the risk of permanent record gaps
The outage was limited to a maximum of 95 minutes and so access to previous notes etc. was temporary and with a low likelihood of meaning actions could not be carried out in the interim, or significant context for decisions was missing.
Authentication error messages clearly informed users that the service was unavailable, preventing clinicians from falsely believing a consultation was being recorded when it was not.
Clinicians are trained to maintain contemporaneous notes. In the event of a technical failure, the default workflow is a return to manual entry within the Electronic Patient Record (EPR).
Users would have noticed the issue when it occurred and hence had to shift to an alternative documentation method to maintain patient safety during the downtime.
Resolution and Next Steps
Immediate Fix & 'Made Safe' Actions: Tandem engineers scaled up primary database resources to 128GB to stabilize memory pressure and reverted backend changes that had increased resource consumption. Service was fully restored by 11:37am
Longer-Term Improvements & Learnings: Root Cause Analysis (RCA) to be carried out.
Residual Risk Rating
Severity: Significant
Likelihood: Low
Clinical Risk Rating: 2
If you have any remaining questions or concerns, please reach out to our support team at support@accurx.com. We sincerely apologise for the disruption and thank you for your patience as we work to make our systems more resilient.
Thanks,
The Accurx team