ChatGPT for Doctors and Nurses in India: Safe Use Guide
AI for OPD notes, discharge summaries and TPA letters in India: the DPDP Act and Rules, medical ethics on confidentiality, masking UHID and Aadhaar, and review.
By Redbison Dev · · 9 min read

India edition. Written for Indian hospitals and clinics, with the DPDP Act 2023 and Rules 2025, the 2002 medical ethics regulations, UHID, ABHA and Aadhaar masking and mg/dL units in place of HIPAA. US edition
In a busy Indian OPD the paperwork that follows (discharge summaries, referral notes, TPA queries, instructions for the family) is mostly written after hours. A general AI assistant is good at exactly this kind of rewriting: dictated points into a structured discharge summary, a dense instruction sheet into plain English that a relative can read aloud, a nurse's shift notes into an SBAR handover. It is not good at facts it was not given, and it is not the place for a patient's UHID, Aadhaar number or mobile number unless your hospital has approved the tool for that. This guide covers where AI drafting helps in Indian practice, what the data protection law and the medical ethics regulations expect, how masking handles Indian identifiers, and a review routine. The AI Prompts for Doctors (India) and AI Prompts for Nurses (India) kits contain ready-made versions of the prompts.
Where drafting helps in Indian hospitals and clinics
| Task | What the model contributes | What stays with you |
|---|---|---|
| Discharge summary from case notes | Headings, chronology, a clean medicines table | Diagnosis wording, the medicines and the follow-up date |
| Cashless pre-authorisation or TPA query reply | A complete, ordered letter: diagnosis, line of treatment, expected stay, estimated cost in rupees | Every figure, each tied to the record |
| Referral note to a specialist | A short question first, relevant history only | Urgency and the clinical question |
| Instructions for the patient's family | Plain English at a set reading level | Warning signs, when to return, the 108 or 112 advice |
| Nursing handover | SBAR order, abnormal values first | Escalation decisions |
| Teaching material and MCQs | Varied questions in a given exam style | Checking each answer against current guidance |
Where patient instructions need to be in Hindi, Tamil, Marathi or another language, the India setting in the kits asks the model to draft in plain English first and mark the text [TRANSLATE AND CHECK WITH A FLUENT COLLEAGUE]. Machine translation of dosing instructions is a place where a small error becomes a serious one.
What Indian law and ethics expect
Data protection. The Digital Personal Data Protection Act 2023 governs digital personal data, and the Government notified the DPDP Rules 2025 on 14 November 2025. According to the Government's backgrounder on the Rules, they give an eighteen-month period for phased compliance, require each data fiduciary to issue a clear, separate consent notice explaining the purpose, require affected individuals to be told of a personal data breach without delay, and set penalties of up to ₹250 crore for failing to maintain reasonable security safeguards. Data fiduciaries must also answer access and correction requests within ninety days.
Medical ethics. The Indian Medical Council (Professional Conduct, Etiquette and Ethics) Regulations 2002, published by the National Medical Commission, say that confidences patients share with a physician should never be revealed unless the law requires it (2.2), and that a registered practitioner shall not disclose the secrets of a patient learnt in the course of practice except in listed circumstances (7.14). They also ask physicians to prescribe by generic name as far as possible (1.5), to keep indoor patients' records for three years (1.3.1), and to issue records within 72 hours of a request (1.3.2). An AI draft that goes into the record is part of that record.
Your hospital's policy. Your hospital may have its own rules on which software may hold patient data. Ask the medical records department or IT before you start.
Masking Indian identifiers: a worked example
The kits' masking helper runs in your browser and replaces common identifiers with tokens. A made-up admission note:
Patient: Lakshmi Narayanan, UHID 20231187, ABHA number 91-2345-6789-1234, Aadhaar number 2345 6789 0124, mobile +91 98765 43210, Indiranagar, PIN 560038. Admitted 05/11/2026.
becomes:
Patient: [NAME-1], UHID [RECORD-ID-1], ABHA number [RECORD-ID-2], Aadhaar number [NATIONAL-ID-1], mobile [PHONE-1], Indiranagar, PIN [POSTCODE-1]. Admitted [DATE-1].
Everything labelled is caught. Two limits are worth knowing:
- Unlabelled numbers. An Aadhaar number written without the word "Aadhaar" is masked only when its check digit is valid, so a mistyped number slips through. A six-digit PIN code without the word "PIN" is not masked at all, because six-digit numbers are everywhere in clinical notes (lab values, bill numbers).
- Context. "Indiranagar" stayed in the text. A locality plus an unusual occupation or a rare diagnosis can identify a patient in a small town without any number at all.
Run the helper, then read the text yourself before you paste it anywhere.
An India-specific prompt
With the region set to India, the kit adds constraints such as:
- Write for Indian practice: British spelling, dates as 6 November 2026 or DD/MM/YYYY, blood glucose in mg/dL unless the source uses mmol/L, and money in rupees with lakh and crore.
- Write drug names as generic names where the source gives them; the medical ethics regulations the National Medical Commission applies ask doctors to prescribe by generic name.
- Name ICMR, national programme or specialty society guidance only if it appears in the material; otherwise write [CHECK INDIAN GUIDELINE: topic].
- Show ABHA numbers, UHIDs, Aadhaar numbers and mobile numbers only as placeholders such as [ABHA NUMBER].
The units line catches a common error. If your laboratory reports glucose in mg/dL and a model defaults to mmol/L, a silent conversion of "FBS 126" can produce a number nobody recognises. Tell it to keep units as written.
Nursing uses
Nurses in Indian hospitals can use the same approach for handover, care plans, incident reports and patient teaching. The India setting uses hospital terms (ward sister or nursing in-charge, duty doctor, OPD, IPD, casualty) and the 108 and 112 emergency numbers, and it tells the model to follow a named hospital or NABH standard only if you paste it in, otherwise writing [CHECK HOSPITAL POLICY]. Do not use a model for drug calculations or infusion rates; use your hospital's method and a second check. For nurses preparing for NCLEX-RN or the UK NMC test, the kit's practice-question template asks the model to follow the named exam's style without claiming to match the real paper.
Review before you sign
- Facts against the case sheet. Every finding, value, drug and date in the draft must be in your notes.
- Numbers and units. mg/dL against mmol/L, mg against mcg, and rupee amounts in TPA letters.
- Generic names and doses exactly as prescribed.
- Guideline claims checked in the primary source; models invent citations.
- Family instructions read aloud as the family would hear them, with the right emergency numbers.
Frequently asked questions
Can doctors in India use ChatGPT for patient notes?
Only with identifiers removed, or through a tool the hospital has approved for patient data. Patient data is digital personal data under the DPDP Act 2023, and the medical ethics regulations bar a doctor from revealing patients' confidences except as the law requires.
When do the DPDP Rules 2025 apply to hospitals?
The Rules were notified on 14 November 2025 with an eighteen-month period for phased compliance, which runs to May 2027. Check with your hospital's data protection contact which duties already apply to it.
Does the masking helper catch an Aadhaar number?
Yes when it is labelled "Aadhaar", and also without the label when the twelve digits have a valid check digit. A mistyped number without a label may slip through, so read the text before pasting.
Can AI write a TPA pre-authorisation letter?
It can draft the letter's structure from your notes, but every figure (diagnosis, line of treatment, expected stay, estimated cost in rupees) must come from the record and be checked before it goes to the insurer or TPA.
Should AI convert glucose values to mmol/L?
No. Keep units as written in the source. A silent conversion is an easy way to introduce an error into a summary.
Sources and further reading
- DPDP Rules, 2025 notified: backgrounder (17 November 2025), Press Information Bureau, Government of India.
- Code of Medical Ethics Regulations, 2002, National Medical Commission. Clauses 1.3, 1.5, 2.2 and 7.14.
- Data protection framework, Ministry of Electronics and Information Technology.
- SBAR Tool: Situation-Background-Assessment-Recommendation, Institute for Healthcare Improvement.
- Ethics and governance of artificial intelligence for health: guidance on large multi-modal models, World Health Organization.
Open the AI Prompts for Doctors
More AI for Professions
Every tool in the AI for Professions section, free and in your browser.