Records in October: check corrections and understand finding changes

After a FHIR correction, the next check shows the result. The October update covers three ways to inspect data and use the results in Records:

• Follow corrections: identify resolved, new and changed findings directly on the resource.
• Explore documents: inspect synthetic SYNDERAI examples in the public demo.
• Return dataset-quality findings: share affected resources and indicators as CSV ZIP or Excel tables.

Watch the recorded correction workflow — recorded on 22 October 2026, with English captions and a transcript.

One correction, two requirements

The example uses a synthetic oxygen-saturation Observation checked against two profiles:

1. Before: the laboratory category is present. The oxygen-saturation profile requires the missing vital-signs category.
2. After the correction: a reviewed Claude Code correction adds vital-signs and preserves laboratory. After the source write and readback, Records independently revalidates the resource. The missing-category finding is resolved.
3. Spot a new gap: a deliberate test then removes laboratory, producing a new laboratory-profile finding. Both categories are subsequently restored and checked again.

The measurement remains unchanged. A separate finding about an unresolved extension definition stays open. The example does not establish complete conformance or clinical correctness. The AI-assisted workflow guide explains the process in more detail.

Resolved, new and changed findings

Records shows the changes in the Validation pane beside the affected field:

• Resolved: the finding is no longer present. It remains inspectable without contributing to the current severity total.
• New: the next check reports an additional finding.
• Changed: the before-and-after explanation can be compared.

Findings that remain open stay visible alongside these changes.

What the comparison uses

• Resource: the last two complete checks of the same current resource. This comparison clears when the browser reloads.
• Sandbox: the temporary draft and its first complete check.
• Recorded runs: an explicitly selected baseline and another run. Fewer occurrences do not necessarily mean a resolved issue type.

Incomplete checks or changed validation inputs can prevent a comparison. Temporary resource comparisons do not retain immutable profile and terminology snapshots. Not every finding change can therefore be attributed to the correction alone.

More October updates

Explore SYNDERAI documents

Explore synthetic documents from SYNDERAI Release 5 in the public demo. Start in Bundle, then open the embedded resources:

• Medplum: laboratory reports, patient summaries and discharge reports.
• Health Samurai: laboratory reports and patient summaries. Discharge reports are excluded after an R4 status value was rejected there.

Embedded resources are not necessarily searchable individually.

Return dataset-quality findings

The MII-oriented workflow delivers findings and affected resources as CSV ZIP or Excel return tables, based on a selected population and versioned quality programs. Each indicator includes its numerator, denominator and not-evaluable cases; a zero denominator is not a passing result.

The checks include independently defined indicators and selected CQL-derived functions for individual resources. The scope summary documents the function comparison and synthetic workflow example separately. Whole Measures, full original CQL execution and original R execution require separate evidence. The return-table format is provisional and has not been accepted by customers.

The research-data workflow shows how teams can use the results to agree a correction and a comparable follow-up run.

Try it yourself

Start with a synthetic Patient: validate it, correct one value and check again. The sandbox works with a temporary copy and does not change a source server.

For integration, see the developer guide. Request an evaluation to agree a scope for a staged workspace.