Confidentiality, and when it breaks

A patient tells a doctor something they have not told their family. They do it because they believe it stops there, and because medicine has spent well over a century persuading them it will. Confidentiality is the price of admission to an honest consultation.

It is also one of the few duties in medicine described as strong rather than absolute. Interviewers use it for exactly that reason: they want to know whether you understand the duty well enough to recognise when it yields, and whether you can set it aside carefully rather than casually.

What follows is the reasoning, the three grounds for disclosure, the four scenarios you are most likely to be handed, and the process that earns the mark in every one of them.

Why the duty exists at all

There are two reasons and you should be able to give both, because giving only one makes the answer thin.

The instrumental reason is about outcomes. If people think their disclosures will travel, they edit them or they stay away. Someone worried about a sexually transmitted infection, a drinking habit, a mental health crisis or their immigration status will not come through the door, and what a clinician most needs to know is often what a patient is least willing to say. A service without a credible promise of confidentiality ends up treating sicker people later.

The intrinsic reason is about the person. Information about your body, your mind and your history is part of you, and controlling who holds it is an exercise of autonomy — the pillar you met in the four pillars of medical ethics. Even where no practical harm follows, the patient has been wronged. That is why “no harm was done” is not a defence, and why the duty outlives the patient rather than ending at death.

Hold both together. Argue only instrumentally and you imply the duty could be dropped the moment it stopped being useful. Argue only intrinsically and you cannot explain why it ever yields.

Sharing inside the team is not a breach

The most common misunderstanding is that confidentiality means telling nobody. It does not. It means information is held by the people who need it in order to care for the patient, and goes no further without a reason.

Sharing within the direct care team runs on implied consent. A GP referring to a cardiologist, a nurse reading the notes, a pharmacist checking an interaction: none of these needs its own negotiation, provided the patient has access to information about how their data is used and has not objected. That clause matters. A patient can object, and if they do you must respect it and explain honestly what it may cost their care, which is the conversation set out in consent and capacity.

Two consequences are worth carrying into a station. Asking a senior colleague for advice is not a breach, it is expected, so never present “I would speak to my consultant” as a reluctant compromise. And anything beyond the direct care team — a researcher, an employer, an insurer, a relative who is merely worried — needs explicit consent or one of the grounds below.

The three grounds for disclosure

The GMC frames this simply, and you should borrow its structure because it is the structure you are marked against. Confidential information may be disclosed on one of three bases.

1. With the patient’s consent. This is the route you attempt first in almost every scenario, and the route candidates skip.
2. Where the law requires it. Some disclosures are compulsory, and your own view of the merits is not what decides it.
3. In the public interest, where the benefits of disclosing outweigh the harms, including the harm done to the confidential service itself.

The GMC’s own list runs slightly longer — it also covers disclosure that is of overall benefit to a patient who lacks the capacity to consent, and disclosure approved through a statutory process that sets aside the common law duty for research or public health purposes. Neither is likely to decide a station, but knowing that the three headline grounds are a simplification is worth a sentence if an interviewer pushes.

Required by law

Four examples cover almost everything an interviewer is likely to raise.

• Notifiable diseases. Registered doctors have a statutory duty to notify suspected cases of listed diseases — measles, tuberculosis, acute meningitis, cholera and others — to the proper officer at their local authority or health protection team. Consent is not required. Each UK nation keeps its own list; HIV is on none of them, which is why it is handled through consent and public-interest reasoning instead.
• A court order. A judge can order disclosure and you must comply. You may object to producing material you believe is irrelevant, but once ordered you cannot withhold it. A solicitor’s letter or a police request is not a court order.
• FGM mandatory reporting. In England and Wales, a regulated health professional who is told by a girl under 18 that female genital mutilation has been carried out on her, or who sees physical signs appearing to show it, must report it to the police, normally by the close of the next working day. The duty came into force in October 2015, it is personal to the individual professional, and it cannot be delegated or overridden by the family. It covers known cases in under-18s only: suspicion, future risk and adult cases go down the ordinary safeguarding route.
• Other statutory duties. Road traffic legislation, for instance, requires anyone, a doctor included, to give the police information identifying a driver alleged to have committed an offence. That is a duty to provide a name and address, not an invitation to open the clinical record.

In the public interest

This is the ground that carries most interview scenarios. The test is a balance: the benefit of disclosing, usually prevention of serious harm or death, against the patient’s interest in confidentiality and the wider interest in a service people are willing to use. The bar is serious harm, not any harm. Crimes against the person — abuse of a child, serious assault, rape, homicide — will normally clear it. Minor property offences will not.

The courts have accepted this reasoning for a long time. In W v Egdell, decided by the Court of Appeal in 1989 and reported in 1990, an independent psychiatrist instructed by a detained patient’s own solicitors passed his unfavourable report to the hospital, from where it reached the Home Office. He was held not to have acted unlawfully: the public interest in protecting others from serious harm outweighed the patient’s interest in privacy.

The four scenarios you will actually be given

The patient with epilepsy who keeps driving

This is the canonical case and the sequence is worth learning exactly, because it is one of the few places an interviewer can hear the difference between a candidate who knows the guidance and one who is improvising.

1. Explain that the condition affects fitness to drive and that the patient has a legal duty to inform the DVLA themselves, or the DVA in Northern Ireland. The obligation is theirs first, not yours.
2. Advise them to stop driving now, and record that you gave that advice.
3. If they will not accept it, try again rather than escalating. Offer a second opinion, give them time off the road to think, and with their agreement involve a relative who may be able to persuade them.
4. If you become aware they are still driving and cannot be persuaded to stop, ask yourself whether that leaves others exposed to a risk of death or serious harm. Where it does, tell the patient you will contact the DVLA, then disclose the relevant information in confidence to its medical adviser.
5. Write to the patient afterwards to confirm the disclosure has been made.

Two refinements will mark you out. If a patient cannot understand the advice at all, because of advancing dementia for instance, the persuasion stage falls away and you notify without delay. And you are not expected to quote seizure-free periods: those licensing standards sit in the DVLA’s guide for medical professionals and are revised over time. Saying you would check the current standard is stronger than a half-remembered number.

A threat to a named person

A patient tells you they intend to seriously harm someone specific. What matters is seriousness, specificity and imminence: a named target and a credible plan is a different object from generalised anger about an ex-partner. Explore it rather than reacting to the first sentence — what has happened, what they actually mean, whether they have any means. Where a real risk of serious harm to an identifiable person remains, the public-interest ground is engaged, and disclosure to the police, and where appropriate to the person at risk, can be justified. Note the wording: UK law does not impose the blanket duty to warn that candidates sometimes import from American cases, so the question is whether disclosure would be justified, not whether it is automatic. You would involve seniors and the mental health team, but seeking advice is not a substitute for acting while someone is in danger.

A communicable disease and a partner who does not know

A patient newly diagnosed with a serious communicable infection refuses to tell a regular partner. Start with why. Fear of violence, of abandonment, of being outed to a family: most of that is workable with support, and specialist health advisers carry out partner notification routinely, often without identifying the index patient. Treatment has also changed the arithmetic: the position accepted by UK and international HIV bodies is that a patient on effective therapy with a sustained undetectable viral load does not pass HIV on to sexual partners, so the risk you are weighing may not be the risk the scenario implies. Say that carefully — it is specific to HIV, and it depends on the viral load actually being suppressed.

If a genuine ongoing risk remains and the patient still refuses, GMC guidance accepts that you may inform someone you have reason to think is at risk of infection. You would tell the patient of your intention first, unless that would put someone in danger, and disclose only what they need in order to protect themselves. What you may not do is inform employers, insurers or relatives who are not themselves at risk.

A child or vulnerable adult at risk

Safeguarding is the exception where the threshold is deliberately lower. You do not need proof, or the parent’s agreement, or certainty about what happened. A reasonable concern that a child is suffering or is at risk of abuse or neglect is enough to share with the appropriate agency, and the child’s safety takes priority over confidentiality owed to an adult. The same logic covers adults at risk who cannot protect themselves. If you are unsure, take advice from the designated safeguarding professional, the Caldicott Guardian or your defence organisation — but never let uncertainty become delay while a child is at immediate risk.

The process that earns the mark, every time

The scenarios vary; the sequence does not. Run this reliably and you can hold a confidentiality station on a topic you have never seen.

1. Clarify before you act. What has actually been disclosed, to whom, and how certain are you? A surprising number of stations dissolve once you ask rather than assume.
2. Try to obtain consent. Explain what you would share, with whom, and why. Most disclosures that look forced turn into consented ones after five minutes of honest conversation.
3. Disclose the minimum necessary, to the person who needs it, and no wider. The DVLA’s medical adviser, not the patient’s employer.
4. Tell the patient what you are doing and why, beforehand if possible and afterwards if not, unless telling them would put someone at risk, prejudice an investigation, or is genuinely impracticable.
5. Take advice and document your reasoning: what you disclosed, to whom, and why you judged it justified.

Your own confidentiality: placements and social media

The duty starts before medical school does. Every UK medical school runs a fitness to practise process, and confidentiality is one of the reliable ways students meet it. Applicants get caught in three specific places.

• Work experience stories. Anonymise properly. “A woman in her seventies with advanced COPD” is fine. A first name, a date, a ward, a rare diagnosis and a small district hospital are not, because together they identify someone even though each piece looks harmless alone.
• Notes and photographs. Never photograph notes, screens, wristbands or whiteboards, and keep no identifiable details in a reflective diary. Write about what you learned rather than who you saw, which is the discipline described in turning work experience into insight and the safer habit besides.
• Social media. Anything posted about a placement is effectively public and effectively permanent, private groups included. A clinical anecdote, a complaint about a patient, or a photograph in scrubs attached to a patient story are all avoidable. The standards expected of doctors online sit in Good Medical Practice, and the same expectations apply to students from the first week.

It comes up directly in the room as well. When an interviewer asks you to describe a patient you saw on placement, part of what they are watching is how you refer to them. Detail that could identify the person costs you. Noticing that a consultant checked who else was within earshot before starting a difficult conversation gains you a great deal.

Confidentiality is the clearest example in medicine of a duty that is both genuinely strong and genuinely defeasible, which is why it rewards structure over instinct. The next stop on the path, answering a dilemma live, turns that structure into something you can run in eight minutes with a timer going.