AI for Clinicians

Home / Oncology

Use caseIntermediate5 min read

Drafting Referral Letters to Palliative Care and Support Services

How to use an AI tool to draft structured, sensitive referral letters from oncology to palliative care, psychology, social work, and allied health services.

The problem

Oncology involves a dense web of referrals. At any point in a patient's journey, you may be referring to palliative care, psychology, social work, dietetics, physiotherapy, or specialist nursing services. Each referral needs to clearly communicate the patient's current situation, the reason for referral, and what you are asking the receiving service to provide.

Writing these referrals well takes time you do not always have. Written poorly — too brief, too jargon-heavy, or missing key context — they delay care and generate unnecessary back-and-forth. Written well, they smooth the patient's transition and allow the receiving team to come prepared.

How AI helps

An AI tool can take a brief set of clinical bullet points and produce a structured, clearly written referral letter. You provide the clinical facts. The tool provides the structure and the prose. You check the accuracy and adjust for the specific situation.

Referrals to services like palliative care and psychology often require a more careful tone than standard clinical referrals — they need to communicate the complexity of the patient's situation without reducing it to a clinical problem to be solved. The tool can produce a first draft of that tone, which you then refine.

A real example

Dr Marcus is a consultant medical oncologist. He wants to refer a patient in his mid-fifties with advanced pancreatic cancer to the specialist palliative care team. The patient has physical symptom needs but also significant psychological distress around his diagnosis. His wife is his main carer. Dr Marcus wants the palliative care team to know all of this before they meet him.

He opens an AI tool and types:

Try it yourself
Draft a referral letter from a consultant medical oncologist to a specialist 
palliative care team.

Clinical context:
- Patient: man, mid-fifties, advanced pancreatic cancer
- Current oncology status: no further systemic treatment planned
- Primary referral reason: pain management and symptom control
- Secondary concerns: significant psychological distress around diagnosis; 
  wife is primary carer, also finding situation very difficult
- What I am asking for: initial assessment and ongoing symptom management; 
  also whether psychological support can be offered to patient and carer

Format: professional letter with headings. 
Tone: respectful and humanising — this patient is a person in a difficult 
situation, not a case to be managed.
Do not include specific drug names or clinical thresholds.

The tool produces a letter. Dr Marcus reads it. He adjusts one sentence that refers to the wife's distress in a way that feels slightly clinical. He adds a note about the patient's preferences for home-based care where possible. He checks that the primary and secondary reasons for referral are both clearly stated. The letter is ready in seven minutes.

Try it yourself

Prompt

Things to watch for

Clinical facts must be verified. The tool only knows what you tell it. If your bullet points are incomplete, the letter will be incomplete. Read the draft to check that the clinical picture is accurate and that nothing important has been lost in the translation to prose.

Tone requires careful calibration for sensitive referrals. Referral letters to palliative care or psychology must communicate complexity without being either cold or overly dramatic. Read the draft for emotional register as carefully as for clinical accuracy. Adjust any phrasing that sounds like a bureaucratic checklist.

It cannot include local context. The receiving service may have specific information it wants in every referral — urgency categories, specific symptom scores, or questions it routinely asks. You know your local referral requirements. The tool does not.

The patient's preferences should be reflected. If the patient has expressed specific preferences — about where care is delivered, about what information has or has not been shared with the family — these should be in the referral. You add these yourself; the tool cannot infer them.

Remember: AI is a helpful assistant, not a clinician. You make the call.

Was this lesson helpful?

Related lessons