How to Help Someone With Depression

If they are not safe, or you cannot tell: in the UK, 999, or 111 and choose the mental health option. In the US, 911, or call or text 988. Stay with them if you safely can.

If they are withdrawn but talking: name one change you have noticed, ask one open question, make one specific offer. Do not tell them to cheer up. If hopelessness comes into it, ask them directly about suicide, which is covered next and does not do the harm people fear.

If they have stopped replying: send something short that does not require an answer. "Thinking of you, no need to reply." "I'll be at the shop at five, want anything left on the step?" If the silence frightens you, call a line for yourself. Samaritans will take a call about someone else, and 988 exists for people worried about a loved one as much as for the person in crisis.

If you have been doing this for months and you are exhausted and resentful: that is documented in the research on carers, in almost every study that asks. It is not a private disgrace and it is not evidence that you love them insufficiently.

This is the strongest finding in the whole of carer guidance, and the one most people refuse to believe.

Dazzi and colleagues at King's College London reviewed 13 studies covering adolescents and adults, in general and high-risk populations. Not one found a statistically significant increase in suicidal thinking after someone was asked about suicide. Several found reductions. Blades and colleagues then pooled 18 studies and found a small decrease in ideation, with adolescents showing about twice the drop adults did, and lower odds of subsequent suicidal behaviour. DeCou and Schumann, working with 13 prospective studies, found no harmful effect including in higher-risk groups. Polihronis and colleagues reviewed 17 studies and meta-analysed 8, and found no significant harm to later suicidal behaviour, self-harm or distress.

Four independent teams, different methods, same answer. Asking does not plant the idea.

Two honest limits on that. All of it was measured in research and clinical settings rather than in kitchens, so it establishes that the question is safe rather than that your asking tonight will reduce risk. And no trial has compared the wordings against each other, so "are you thinking about suicide", "have you thought about ending your life" and "are you saying you want to die" are all recommended on consensus rather than evidence of superiority.

What the guidance agrees on is that the question should be plain. Ask it as a question, not as a plea with the answer built in. "You're not thinking of doing anything stupid, are you?" is easy to say no to, which is why trainers warn against it.

Do not promise secrecy you may have to break. Do not react with visible relief if they say no, because relief teaches them that the honest answer was unwelcome.

If suicidal thoughts are present, the thing with evidence behind it is a written safety plan the person makes themselves: the warning signs they notice first, what they can do alone, the people and places that distract them, who to tell, which professionals and lines to call, and a step on making their surroundings safer.

Stanley and Brown's version was tested across nine Veterans Health Administration emergency departments, comparing 1,186 patients who got the plan plus structured follow-up against 454 who got usual care. Over six months, suicidal behaviour occurred in 3.03% of the safety-plan group against 5.29% of the usual-care group, about half as likely, and the odds of attending at least one outpatient mental health appointment more than doubled. It is a cohort comparison rather than a randomised trial, in a mostly male veteran population, so the size of the effect elsewhere is not established.

Your role is to be one of the named contacts, to help them find the plan when they need it, and to help with the step about their surroundings.

On making surroundings safer: the guidance from Mind, from SAMHSA and inside the safety-plan format itself is to reduce access to the things the person could use, particularly anything they have named themselves, for an agreed period. How to do that in a particular house is a question for their clinician or a crisis line. This page will not go further than that, deliberately.

None of what follows has been tested as wording. It comes from what charities recommend and from what patients in qualitative studies say landed well, which is a real source of evidence and a weaker one than a trial.

Sentences that recur across Mind, Beyond Blue and Samaritans:

"I'm sorry you're feeling like this, and I'm here for you." "You're important to me." "You haven't seemed yourself lately, is everything OK?" "I can see this is a really hard time." "I'm not sure what to do, but I'm sure we can figure it out together." "What can I do to help you today?"

Then the one structural point everybody makes and nobody has tested: a specific offer beats an open one. "Would you like me to come over and keep you company?" and "I'll bring dinner Thursday" ask almost nothing of them. "Let me know if you need anything" asks them to work out what they need, decide it is reasonable, and then request it, which is three tasks that depression is particularly good at blocking.

Samaritans add one warning about a sentence supporters reach for under pressure. Avoid "I'll be so upset if you die." It moves the subject to your distress and adds guilt to a person who is already carrying too much of it.

Listen before you solve. That instruction is unanimous across the guidance and it is consensus rather than a tested comparison. The reason patients give for it is worth more than the instruction: advice arrives as evidence you have not understood, or as one more task on a list they cannot start. A useful distinction the guidance rarely draws is between advice about their life, which lands badly, and doing a concrete thing, which does not. Driving them to an appointment is not advice.

This is where supporters freeze, and the instruction is to not argue. Do not start listing reasons to live, because it turns into a debate they can only win by getting worse. Rethink's wording is to encourage focusing on getting through today rather than the future, and to ask about their reasons for living and then actually listen to the answer. Samaritans put it as validating rather than rebutting, on the grounds that many people have spent a long time having their feelings dismissed.

A sequence that compiles those sources, which is not a tested protocol: "That sounds unbearable. I'm glad you told me." Then the direct question from Part 2. Then, if the answer is yes, "Thank you for telling me. I want you to stay safe. Can we get help together right now?"

Fernandez y Garcia and colleagues ran 15 focus groups with 116 patients and asked how the things family and friends said had actually landed. Four categories came back: feeling labelled, judged, lectured, and rejected. One patient described being told it was all in her head and to get over it as belittling, and said it made her feel she was the one trying to draw attention to herself. Another quoted being asked why she was so miserable when she had two children and a nice husband, and answered, I just am.

The mechanism patients name is consistent. "Snap out of it", "think positively", "you have so much to be grateful for", "everyone gets sad", "other people have it far worse" all imply the symptoms are under voluntary control. Worthlessness and guilt are already two of the nine diagnostic criteria. Those sentences add to the symptom.

The one that deserves separating out is exercise, because the sources contradict each other in public.

The same category error runs through most of the advice aimed at supporters. Walks, gratitude lists, getting out of the house and social prescribing are written for low mood and then handed to people who cannot get out of bed. For someone who has stopped rising, the things that matter tonight are safety, contact, food and a route to a clinician.

Progress checks are worth naming too. "Are you feeling better yet?" turns recovery into something they owe you.

You cannot make an adult with capacity accept treatment. Every organisation says this and none of them says what to do next with any confidence.

Nobody has tested raising it again against leaving it alone. There is no trial of persistence versus backing off in depression, and the motivational interviewing evidence that exists is from family interventions in substance use, which is a different problem with different dynamics. Anyone who tells you the research says keep pushing, or says back off, is extrapolating.

What the reasons-for-refusal literature does show is that refusal is usually not stubbornness. People say they see no need, that they want to handle it alone, that they are already dealing with other illness, or that they are frightened of services or have been treated badly by them before.

The practical synthesis the better charity pages arrive at, untested as a package: stop arguing about the label, find a problem they will accept help for instead, offer to go with them to that appointment, keep making contact, and set your own limit. Someone who will not see a doctor about depression may well see one about not sleeping.

If safety is the concern, refusal does not leave you with nothing. You can call a crisis line about someone else. Samaritans will speak to you, and will contact the person on your behalf if you ask, naming you or keeping you anonymous, though they will not tell you afterwards what was said or whether the person accepted anything. NHS 111's mental health option takes calls about another person. So does 988.

Almost all the guidance is written for a generic friend or family member. The research is not.

The largest and most studied group, and the one where the loss is hardest to name, because it is the loss of the person's company rather than the addition of chores. A meta-ethnography of 15 studies by Priestley and McPherson describes partners moving through making sense of the depression, changed family dynamics, and then either coping or not. Partners describe becoming more carer than companion, conversations that only ever orbit the illness, and intimacy going. Spousal concordance is real, with one partner's depression associated with roughly a 1.6 to 2-fold increase in the other's risk, though how much of that is contagion rather than shared circumstances and assortative mating is argued about.

Couple therapy is the intervention with an evidence base here, and it is worth being precise about what it is good for. The 2018 Cochrane review covered 14 studies and 651 participants. Against individual psychotherapy it found no difference in depressive symptoms, at end of treatment or at six months. It did reduce relationship distress more than individual therapy did. Quality was low to very low throughout. So it is a reasonable thing to ask a clinician for when the relationship is part of the problem, and it is not a better depression treatment.

The confidentiality wall is the specific problem. Once they are under a team you may get nothing without their consent, while remaining the person who gets the call at 2am. NICE's answer is that confidentiality should be negotiated explicitly rather than left to collapse into silence, which is worth asking for by name at an appointment they agree to take you to.

A different legal and clinical position. You can usually take them into the GP and CAMHS pathway, the school is often already involved, safety planning should be done with you and you should have a copy, and limiting access at home is treated in youth guidance as a parental job rather than an intrusion.

This is the position guidance almost never addresses, and where the observational evidence is strongest. Weissman's 30-year follow-up of offspring of depressed parents found roughly three times the risk of major depression, higher still for those whose parent's depression began before puberty. The parentification literature describes children organising themselves around the parent's wellbeing and hiding their own distress so as not to add to it. Van Parys interviewed young adults looking back who described having had no space to think about themselves.

Advice pages tell supporters to keep supporting. If the supporter is a child, the useful instruction is the opposite: an adult should be taking that role back. In the UK, young carers' services exist and are almost never mentioned on depression family pages.

No legal standing, often no information, and the easiest position to be shut out of. Everything above still applies, including calling a line if you are frightened. Friends burn out when they become the only person holding it, which is an argument for recruiting others rather than for stepping back.

You are not their clinician and should not try to be. The employment guidance in both countries is consistent: have the conversation privately, listen, know where the employee assistance programme and occupational health are, and consider adjustments to hours, workload or time for appointments. In the UK depression can count as a disability under the Equality Act, which brings a duty to make reasonable adjustments; in the US the ADA analysis is fact-specific. You cannot require a diagnosis and you cannot compel treatment.

Here the evidence is unusually clear and unusually unflattering to the things charities promote.

Morgan, Ross and Reavley pooled 18 controlled trials and 5,936 participants. The course reliably improves knowledge, recognition of mental illness, confidence and stated intention to help, with effect sizes between about 0.2 and 0.75. Then it stops. The amount of help provided at follow-up improved slightly, at d=0.23, and the quality of the help was unclear. Intentions decayed from 0.75 immediately after training to 0.26 past six months. A later systematic review found mixed effects on whether trainees used the skills at all and no demonstrated benefit to the person being helped. The 2023 Cochrane review, covering 21 studies and 22,604 participants, concluded the training may have little or no effect on the mental health of the people around trainees at six to twelve months, with the evidence very uncertain.

Take the course to become literate. Do not take it expecting to change what happens to the person you are worried about.

Structured sessions for relatives: what depression is, communication, problem-solving, planning for relapse. The evidence splits in an odd place.

For the patient, it can work well. A Japanese trial of four sessions for families of patients in remission found relapse at nine months of 8% against 50%, a risk ratio of 0.17, with a number needed to treat of about 2.4. Small, and one setting. A 2022 meta-analysis of five studies covering 301 patients and families put patient symptom improvement at a standardised mean difference of about 0.52, graded very low certainty.

For you, it does not. The same meta-analysis found no significant effect on family functioning or on family members' distress. When a trial was designed with the relative's own distress as the primary outcome, it came out null: 49 families of patients with chronic depression, K6 distress at 16 weeks, mean difference 1.17, confidence interval crossing zero. A benefit at eight weeks had disappeared by sixteen, and the authors suggested that ending the programme may itself have raised anxiety.

A course that teaches you what to expect is worth having. It has not been shown to make you less distressed, and being told otherwise and then not improving is its own small injury.

A 2026 meta-analysis pooling 11 studies and 1,366 caregivers put mean Zarit Burden Interview scores at 38.15, rising to 42.40 for carers of people with severe depression. That is in the range usually reported for schizophrenia caregiving, which is not where most people would guess it sits.

Figures for depression and anxiety among carers themselves are high and badly specified. The best recent pooling, 52 studies and 10,232 people, reports depression at around 44% and anxiety at around 67%, but that is caregivers of people with mental illness generally rather than depression specifically, and the heterogeneity is enormous. Treat it as an indication that this is common, not as a rate for your situation.

The qualitative work is more use than the numbers. What carers name as worst: the unpredictability, the person still looking fine to everyone else, feeling personally responsible for whether they live, and the job never being finished.

Guidance is timid here. Mind says there are always limits. Beyond Blue tells supporters to decide what they can provide and for how long. SAMHSA has started saying that 988 is for carers too.

The line that is defensible, though it is reasoning rather than a trial result: if you are the only person holding safety, the problem is not your stamina, it is that safety needs more than one holder. Recruiting a clinician, a crisis team or another relative is not withdrawal. Stepping back from being the treatment is different from stepping back from the relationship, and patients are clear in the qualitative literature that the second one wounds. They describe a supporter drifting away as confirmation they were a burden all along.

1. Ask directly whether they are thinking about suicide. Four independent reviews found it does not increase risk, and some found small reductions. This is the best-evidenced thing in this article and the most widely disbelieved.

2. If there are suicidal thoughts, help them build a written safety plan and be one of the names on it. Never accept a promise instead. In the only head-to-head trial, a plan beat a contract by a wide margin.

3. Make specific offers rather than open ones. "Let me know if you need anything" asks them to do three things depression blocks.

4. Listen before suggesting. Advice about their life lands as evidence you have not understood. Doing a concrete task does not.

5. Do not tell anyone to snap out of it, think positively, or count their blessings. Patients report those as labelling, judging and lecturing, and they add shame to a symptom that is already shame.

6. Nobody has tested whether to raise it again after a refusal. Anyone telling you the evidence says push, or says wait, is guessing.

7. Mental Health First Aid improves what you know and not what the person receives. Family psychoeducation may help them and has repeatedly failed to help you.

8. Carer burden in depression scores in the same range as schizophrenia caregiving. Resentment and exhaustion are ordinary. You need other people holding this with you, and that is different from leaving.