Healthcare Clinical Incident Reports: Use case and AI Confidentiality

After a patient falls in the clinic, the practice manager writes an incident report, and it is one of the most candid documents the practice produces. It names the patient, describes the injury and the medicines they take, names the staff who saw it, records the apology and sets out a first theory of what went wrong. Managers paste it into ChatGPT or Claude to tidy the timeline or draft the review summary. This page marks every item in a routine clinical incident report that the Privacy Act, the Civil Liability Act 2002 (NSW), the Australian Open Disclosure Framework and the Health Practitioner Regulation National Law bear on, and shows how the same report is handled on Ermos inside the practice.

The clinical incident report

Written by the practice manager within a day or two of a clinical incident, with the clinician who attended, and kept on the incident register. It is the internal record that the review, the open disclosure meetings and any insurer notice are built from.

Kestrel Lane Medical Practice
Clinical incident report (internal) · Ref KLMP-INC-2026-014 · Prepared Thursday 29 October 2026, 9.40am

Person affected.

What happened.

Seen by.

Injury and immediate actions.

Disclosure.

Contributing factors (preliminary and not findings).

Reporting decisions.

Actions and learning.

An invented incident report. The Medicare number, policy reference and insurer are placeholders or fictional. Hover or tap any highlighted item for the analysis; the full list is below.

Clinical incident report drafter

An Agent+ that drafts the clinical incident report in your practice's own format. It works from the clinician's account, the record entries and any witness notes, builds the timeline, and lists missing details such as the notification decision and the insurer notice.

What it does

  1. Reads the clinician's account and the record entries
  2. Drafts the report in your incident template
  3. Lists missing details and open questions

What stays with your people. The practice manager and the attending clinician review the report, decide on any notification and enter it on the register. The agent drafts only.

An Agent+ is built once and anyone in the practice can run it. It works from your own documents and runs on Ermos Dominion, the AI unit in your office, or on Ermos Edge, your own instance in an Australian data centre.

What leaves the practice

  1. 1
    The patient’s identity

    Name, date of birth, home address and Medicare number tie the whole report to one person. Information about an injury of an identifiable person is health information (Privacy Act s 6FA(a)(i)), and health information is sensitive information under s 6(1). Privacy Act guide.

  2. 2
    The account of how the patient was hurt

    A dated, timed description of a fall on the premises, including a wet floor and a sign in the wrong place. Health information about the patient, and the practice’s own account of an event that may later be examined by an insurer or a court. Patient data guide.

  3. 3
    Witnesses and a contractor

    Named staff and a named cleaning contractor. Personal information about colleagues and a contractor, written down at a time when a review has not yet decided what each person did or did not do.

  4. 4
    Injury, medication and the emergency response

    A laceration, a reported fracture, a blood thinner and the times of the Triple Zero call and the ambulance. Health information, and the medicine explains why the head strike was treated as high risk. Patient data guide.

  5. 5
    The apology and the open disclosure arrangements

    Records that a doctor said sorry, what she explained and that a meeting is booked. The Australian Open Disclosure Framework (June 2026) says an initial discussion should include acknowledging that care did not go as expected with the known facts, an apology or expression of regret, steps to prevent recurrence and contact details for support, and that the discussion is documented. In NSW, Civil Liability Act 2002 s 69 provides that an apology is not an express or implied admission of fault or liability and evidence of it is not admissible in civil proceedings as evidence of fault or liability. Part 10 does not apply to liability excluded by s 3B or to defamation (s 67(2)) and deals with the effect of an apology and says nothing about who may receive a copy of this report. Patient data guide.

  6. 6
    The first theory of cause

    A list of contributing factors the report itself says are not findings. The Framework says discussions should not speculate on causes or apportion blame and should not pre-empt the results of reviews. A draft theory pasted into an outside tool sits on someone else’s servers before the practice has decided what it thinks. Patient data guide.

  7. 7
    The decision whether to notify Ahpra

    Section 140 of the Health Practitioner Regulation National Law describes four kinds of notifiable conduct: practising while intoxicated, sexual misconduct, placing the public at risk of substantial harm because of an impairment, and a significant departure from accepted professional standards that places the public at risk of harm. Practitioners who form a reasonable belief must notify Ahpra as soon as practicable (s 141(2)), and an employer who reasonably believes a practitioner it employs has behaved in a way that constitutes notifiable conduct must notify too (s 142(1)). Ahpra says a reasonable belief generally needs direct knowledge rather than suspicion or rumour. A patient fall is not one of the four kinds. The report records the practice’s reasoning, and the decision is a person’s.

  8. 8
    The insurer notice and policy reference

    Names the practice’s insurer and a policy reference and shows the practice treated the event as potentially serious. What a policy requires and when is set by its terms, which this page does not describe.

  9. 9
    Corrective actions and learning

    Fixes to cleaning, signs and reception practice, with owners and dates. Workplace safety measures that identify no patient, though together they describe the weakness the incident exposed.

  10. 10
    Author and reviewer

    Ties the report to a named practice manager and doctor, and records when it was written. It identifies who to ask, and who is responsible for the content.

What happens in Frontier Models (ChatGPT, Claude)

  • The whole document is transmitted to the provider’s servers. On personal plans the user cannot choose where it is stored or processed, and for the major providers that has generally meant outside Australia, under that country’s legal process. Some business plans offer storage in Australia, but the provider still receives the content.
  • On consumer tiers, prompts can be retained and used to train future models.
  • The provider’s staff and systems can reach the content under the provider’s own policies.
  • Once sent, the practice no longer controls it. Retention and deletion happen on the provider’s terms.

The same document on Ermos

Ermos Dominion

The document is drafted on a unit in your office, from your own templates and the patient record. Every marked item stays on the unit, which keeps working with the internet switched off. Nothing in it is transmitted to an AI provider, so there is no disclosure to analyse: the confidentiality work is the same as for your file server.

Ermos Edge

Your documents, the index and your access boundaries stay on your own instance in an Australian data centre. When a staff member asks for this draft, only their question and the excerpts they are permitted to see go to a named Australian provider, who is contractually barred from keeping the content or training any model on it. Nothing is processed outside Australia.

Frontier Models (ChatGPT, Claude)Ermos EdgeErmos Dominion
Where the document is processedProvider servers, generally offshore on personal plansYour instance in an Australian data centre; answers generated in AustraliaThe unit in your office
Who can read itProvider staff and systems, under policyYour staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothingYour staff only
Is it keptOn the provider’s termsKept on your instance; the provider is barred from retaining itStays on the unit
Can it train a modelOn consumer tiers, yesNo, by contractOnly your own unit learns from it, on your hardware

Part of Use cases in healthcare practices. See how this works in practice: Clinical documentation on Ermos.

Common questions

Do ChatGPT and Claude use the information in a clinical incident report?
Everything pasted or uploaded is sent to the AI company's servers and processed there to produce a reply. As at October 2026, whether a conversation on a personal plan (ChatGPT Free, Plus or Pro; Claude Free, Pro or Max) can be used to train the models depends on a setting each user controls. Business plans (ChatGPT Business and Enterprise, Claude Team and Enterprise) and the developer APIs sit under commercial terms, and OpenAI and Anthropic each say they do not train on that content by default (OpenAI, Anthropic). On either product, the thumbs up and thumbs down buttons can send a whole conversation back for training. This report carries Mrs Paterson's name and Medicare number, her blood thinner and fractured wrist, and a list of contributing factors that the report itself says are not findings.
Can we use ChatGPT or Claude for a clinical incident report if we remove the names?
Not safely, because the account still points to one patient. Removing names rarely makes a document anonymous. Information stays personal information while the person is reasonably identifiable, and the OAIC warns that information can be at risk of re-identification even when it has been de-identified or anonymised. A woman of 78 who fell in a corridor mopped at 10.05am, was taken to Royal North Shore Hospital and takes apixaban would be recognised by the staff, the ambulance crew and her daughter. The injury is health information under section 6FA(a)(i) of the Privacy Act. The Office of the Australian Information Commissioner (OAIC) recommends in its guidance on commercially available AI products that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools, because of the significant and complex privacy risks.
Does deleting the chat remove a clinical incident report from ChatGPT or Claude?
Not completely. Deleting a chat removes it from the user's history, but OpenAI and Anthropic each say they can keep a deleted chat on their systems for up to 30 days, and longer where safety or legal obligations apply (OpenAI, Anthropic). That matters here because the report holds the apology, a first theory of cause and the insurer notice before the review has started. Deleting the chat would not show the practice that those lines are gone.
How can a medical practice use AI for a clinical incident report without sending it to ChatGPT or Claude?
Ermos is built so the document stays under your control. On the Dominion AI unit it is drafted inside your office and none of it is sent to an AI provider. On Ermos Edge your documents and index stay on your own instance in an Australian data centre, and only the question and the excerpts the user is permitted to see go to a named Australian provider that is contractually barred from keeping them or training any model on them. The report is drafted from the practice's incident register and its own report template. The practice manager checks every fact, and the doctor reviews it before it is finished.
Is a clinical incident report health information?
Where it records a patient's injury, yes. Section 6FA(a)(i) of the Privacy Act counts information or an opinion about the injury of an individual as health information if it is also personal information, and section 6(1) lists health information as sensitive information. A report that names a patient, the injury and the medicines they take therefore needs the same care as the clinical record.
What does open disclosure involve after a patient is hurt in the clinic?
The Australian Open Disclosure Framework (June 2026) says an initial discussion should happen as soon as possible, even if not all the facts are known. It should acknowledge that care did not go as expected, include an apology using the words I am sorry, listen to the patient, describe the steps being taken, give support contacts and be documented.
Must a practice notify Ahpra after a patient falls in the clinic?
Not because of the fall alone. Ahpra's mandatory notifications concern four kinds of notifiable conduct by a practitioner: intoxication, sexual misconduct, impairment placing the public at risk of substantial harm, and a significant departure from accepted professional standards. A notification follows a reasonable belief that one has occurred, and Ahpra says that generally needs direct knowledge rather than suspicion or rumour. A practice should take advice where the facts are unclear.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo