Aged Care Nutrition and Swallowing Assessments: Use case and AI Confidentiality

When a resident loses weight or coughs at meals, the nurse screens her nutrition and the speech pathologist assesses her swallowing. The result is a short form with a score, a texture level for food and another for drinks, and a list of risks, and it goes to the kitchen, the care staff and the family. It is tempting to have ChatGPT or Claude turn it into a handover sheet or a kitchen card. The form holds her weights, her risk score, how she swallows, her medicines and what she likes to eat. This page marks every item in a realistic assessment and shows what the Quality Standards bear on, and how the same form is turned into a kitchen card on Ermos.

The nutrition and swallowing assessment

Completed when screening shows a risk, or the resident coughs at meals or has lost weight, by the nurse and the visiting speech pathologist, and reviewed after a few weeks.

Kurrajong Gardens Aged Care
Nutrition screening (Monday 2 November 2026) and swallowing assessment (Tuesday 3 November 2026)

Resident
Height and weight
MUST score
Swallowing observed
Recommended diet
Medicines
Risks
What she likes
People and review
Level key

An invented assessment. Hover or tap any highlighted item for the analysis; the full list is below.

Nutrition and swallowing pre-filler

An Agent+ that gathers what is already recorded for one resident: weights over time, the screening score, the current tablets and what she likes to eat. It pre-fills your assessment form and flags anything missing, so the nurse and the visiting speech pathologist can complete the assessment.

What it does

  1. Reads the weight record, screening result and medication list
  2. Pre-fills your assessment form from the file
  3. Flags changes in weight and missing entries

What stays with your people. The nurse and the speech pathologist complete the assessment and set the diet and fluid levels. The agent prepares the form; it does not assess.

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 service

  1. 1
    The resident

    A named woman, her room and her date of birth. Section 168 of the Aged Care Act 2024 requires the provider to protect personal information about the people it delivers services to. Privacy Act guide.

  2. 2
    Height and weights

    Measurements three months apart, which show a loss. Health information is sensitive information under the Privacy Act. Action 5.5.5 of the strengthened Aged Care Quality Standards asks for regular malnutrition screening and a response to the risk of malnutrition and unplanned weight loss. Patient data guide.

  3. 3
    The MUST score and its arithmetic

    The BAPEN tool scores BMI over 20 as 0, weight loss under 5% as 0 and 5 to 10% as 1, and acute disease as 0 or 2. A total of 0 is low risk, 1 medium and 2 or more high. Here the BMI is 52.0 divided by 1.58 squared, or 20.8, and the loss is 4.0 of 56.0 kg, or 7.1%, so the total is 1. A tool can misread a cut-off or a percentage, and the score decides the response. BAPEN says it is for adults. Patient data guide.

  4. 4
    How she swallows

    Clinical observations by a speech pathologist: coughing, a wet voice, pocketing. Action 5.5.2 of the Quality Standards asks the provider to support safe chewing and swallowing when eating, drinking and taking oral medicines. These are health findings and a safety record. Patient data guide.

  5. 5
    The IDDSI levels

    Level 5 Minced & Moist for food and Level 2 Mildly Thick for drinks. The International Dysphagia Diet Standardisation Initiative framework has eight levels, 0 to 7, and says its testing methods decide whether a product meets a level. A wrong level is a choking or aspiration risk, and the kitchen works from it. Action 6.3.1 asks that menus, including for texture modified diets, be developed with the input of chefs, cooks and an Accredited Practising Dietitian. Patient data guide.

  6. 6
    Tablets, crushing and two named medicines

    Action 5.3.1(c) asks for safe administration, including assessing swallowing and deciding whether crushing is suitable. An anticoagulant and an antipsychotic in a swallowing note. A tool should not decide that a tablet can be crushed. Patient data guide.

  7. 7
    The risks

    Aspiration pneumonia, dehydration and wound healing, with the link to a psychotropic that she is on under a consent record. The risk list is a clinical judgement and links this record to her wound and to her restrictive practice consent. Patient data guide.

  8. 8
    What she likes to eat, and a family wish

    Her food history and a family’s request. Action 6.2.1 asks the provider to assess what the person likes to eat and drink and when. Preferences, a loss and a dignity issue, all in one line. Privacy Act guide.

  9. 9
    The people and the review dates

    A named contractor, a named nurse, her guardian and the dates of the reviews. Standard 6 on food and nutrition applies only to residential care homes.

  10. 10
    The level key

    A generic summary of the published IDDSI framework. It names no one and is the same on every card in the country.

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 service 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 resident’s 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 aged care providers.

Common questions

Do ChatGPT and Claude use the information in a nutrition and swallowing assessment?
Text pasted from the assessment 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 assessment carries Agnes Thornbury's weights and MUST score, her IDDSI levels, the names of apixaban and risperidone, and her love of shortbread, which is no longer safe.
Is it safe to paste a nutrition and swallowing assessment into ChatGPT or Claude?
No, because the assessment records a named woman's weight loss, swallowing difficulty and medicines. 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. Leaving out her name still leaves a woman in Room 14B with a 4 kg loss in three months, minced and moist food and an anticoagulant, whom her family and the wing's staff would know.
Can ChatGPT or Claude be trusted to draft a nutrition and swallowing assessment?
Not without a qualified person checking every item. Language models can produce fluent text with details that are not in the source, wrong figures or dates, or references that look real but are not. Here the most harmful errors would be a misread MUST cut-off or percentage, a wrong IDDSI level for food or drinks, and any suggestion that apixaban or risperidone can be crushed. The speech pathologist and the registered nurse have to check each one.
How can an aged care service use AI for a nutrition and swallowing assessment 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 summary is drafted from the recorded measurements and the service's own assessment template. The speech pathologist and the registered nurse check the score, the IDDSI levels and the medicine instructions, and sign it.
What are the IDDSI levels?
The International Dysphagia Diet Standardisation Initiative framework has eight levels, 0 to 7. Drinks run from Level 0 Thin through Level 1 Slightly Thick, Level 2 Mildly Thick and Level 3 Moderately Thick to Level 4 Extremely Thick. Foods run from Level 3 Liquidised and Level 4 Pureed through Level 5 Minced and Moist and Level 6 Soft and Bite-Sized to Level 7 Regular. IDDSI says its testing methods decide whether a product meets a level.
How is malnutrition risk scored with MUST?
The BAPEN Malnutrition Universal Screening Tool adds three scores. A BMI over 20 scores 0, 18.5 to 20 scores 1 and under 18.5 scores 2. Unplanned weight loss in the past three to six months under 5 per cent scores 0, 5 to 10 per cent scores 1 and over 10 per cent scores 2. Acute disease with no intake for more than five days scores 2. A total of 0 is low risk, 1 medium and 2 or more high. It is for adults.
What do the Quality Standards say about swallowing and nutrition?
In the strengthened Aged Care Quality Standards, action 5.5.2 asks providers to support safe chewing and swallowing when the person is eating, drinking and taking oral medicines, and action 5.5.5 asks for regular malnutrition screening and a response to the risk of malnutrition and unplanned weight loss. Standard 6, on food and nutrition, applies only to residential care homes and covers assessing nutritional needs and menus for texture modified diets.

See Dominion answering from your own documents.

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

Book a demo