Cath Lab Documentation and Audit: Where the Work Goes and What Software Can Help (2026)

Who this is for, and a disclosure

This guide is for cath lab managers, clinical leads, quality leads and the IT and information-governance staff who support them. Disclosure: Zapx Labs makes Cath Copilot, described near the end. It is a prototype running on synthetic data only, offered to departments as a pilot. It is not a medical device, it is not in clinical use, and it recommends, drafts and routes; it never diagnoses, orders or decides. Regulatory and audit facts below come from the official pages linked at the end, as we read them on 8 November 2026.

The records one case creates

Each record below has its own owner, its own format and often its own system. The same facts — the patient's allergy status, the access site, contrast volume, dose, the devices implanted — appear in several of them.

• Safety checks: NatSSIPs 2, published by the Centre for Perioperative Care in January 2023, covers all invasive procedures including those outside the operating department, so it applies in the cath lab. Providers write local procedures from its high-level standards.
• The procedure report: what was done, with what devices, and what was found, usually written or signed by the operator after the case.
• Radiation dose: under IR(ME)R 2017, the employer's procedures must cover assessing patient dose, using and reviewing diagnostic reference levels, recording a clinical evaluation of each exposure including factors relevant to patient dose where appropriate, and clinical audit.
• The national audit: NICOR runs the National Audit of Percutaneous Coronary Interventions (NAPCI), which collects data on PCI activity in NHS hospitals and a selection of private hospitals across the UK. Operator- and centre-level results are published through the BCIS Clinical Outcomes Publication programme. NICOR lists the current dataset as version 5.6.6.
• Handover: the patient moves to recovery, CCU or a ward, and the receiving nurse needs the access site, medicines given, complications and the plan.
• Implants: devices used need to be traceable to the patient for recall.

Where the work goes

The effort is rarely in typing one record. It is in assembling the same facts from the EPR, RIS, PACS, the physiology system, inventory and ePMA, and noticing when they don't match.

Readiness gaps can be found late. An unrecorded allergy status or a missing last anticoagulant dose may only surface at sign-in, when it is most disruptive.

Values disagree between systems. The injector might record 145 mL of contrast while the nursing record says 140 mL. Someone has to notice, decide which is right, and make the note and audit agree.

Audit data is compiled after the fact. NICOR reports data completeness as part of NAPCI outcomes, so missing fields show up in national reporting, not just locally.

Handover means retelling the case. The receiving team needs the access site, medicines given, complications and the plan, and what they are told has to match the record.

What to ask of any cath lab software

Vendor haemodynamic and reporting systems, cardiovascular information systems and general AI scribes each cover part of this. Whatever you are evaluating, these questions separate tools that reduce work from tools that add a new place to check.

• Where does each value come from? Every field in a brief or note should show its source record.
• Which checks are rules and which are AI? Missing-item checks should be deterministic and predictable, not generated.
• What happens when two sources disagree? The mismatch should be surfaced and confirmed by a person before anything is signed.
• Can it sign something it cannot support? Drafts should cite each statement, and unsupported statements should block sign-off.
• Where does the AI model run? If it calls a cloud service, patient data leaves your network, which your information-governance team will need to assess.
• Who can see what? Access should follow the user's role, with emergency access needing a reason and leaving an audit trail.
• How does it connect? Look for FHIR, HL7 v2, DICOMweb and, for devices, IEEE 11073 SDC, rather than screen-scraping.
• What is its clinical safety position? NHS England's DCB0129 sets clinical risk management duties for manufacturers of health IT, and DCB0160 sets them for the organisations deploying it.

How Cath Copilot approaches it

Cath Copilot is a local-first, chat-first desktop copilot for the cardiac catheterisation lab. Staff sign in with their hospital account, and their role decides what they see: seven roles see the same patient differently, a nurse sees planned access but not order details, a radiographer cannot open the medication chart, and break-glass access needs a reason, is time-limited and is audited.

The pre-case brief carries a source on every field, and missing items such as an unrecorded allergy status are flagged by deterministic local rules, not AI, for someone to resolve, assign or escalate. In the room, events from the physiology system and inventory arrive on a timeline for validation, with haemodynamics shown for display only. Reconciliation catches values where two sources disagree, such as contrast volume, and the mismatch must be confirmed before the note can be signed.

The copilot drafts the procedure note, the national audit return and the handover to CCU, recovery or the ward, with a citation on every sentence; signing locks when a statement has no support. Audit returns are pre-filled with code suggestions and their evidence, and only the quality lead can export. In the prototype the audit return is not mapped to the current NAPCI dataset and does not submit to NICOR. Actions arrive as cards a person confirms, and an append-only, hash-chained audit log records sign-ins, queries, drafts and denials.

The AI model runs locally through Ollama, so nothing leaves the hospital network. It runs on one PC, as a lab host, or against a hospital server, and connects to FHIR R4, DICOMweb PACS, HL7 v2 over MLLP with TLS, and IEEE 11073 SDC devices.

What a pilot means

Cath Copilot has only been run on synthetic data. Any use with real patient data would need a clinical safety and information-governance process first. There is no public download and no published price; departments interested in a pilot can book a demo, on synthetic data, by emailing [email protected].