Carer Burnout and Your Own Health: What the Evidence Really Shows

Sustained caring reliably damages mood, sleep and money, and the population evidence backs that up. What it does not do, on the best available evidence, is shorten your life, and the statistic most often used to claim it does has been badly misused for twenty-five years. Both halves of that sentence matter. Carers are routinely frightened by a number that does not say what they are told it says, while the real and measurable harms get less attention than they should.

This is the article I would have wanted at the point where I had stopped going to my own appointments and had started treating being tired as a personality trait. It sits under caring for a parent at home.

What does caring actually do to a carer’s health?

It raises mental distress consistently, worsens self-rated health modestly, and hits sleep hardest of all. The best population comparison comes from the US behavioural risk factor surveillance system: among adults across 35 states and Puerto Rico, 20.5 percent of carers reported frequent mental distress, meaning 14 or more bad mental health days in the past month, in 2021 to 2022, against 13.6 percent of non-carers. Carers also rated their own health fair or poor more often, 19.6 percent against 16.9 percent (CDC, MMWR 2024, 92,461 carers against 353,242 non-carers)¹. Note the movement: carer mental distress rose from 17.2 percent in 2015 to 2016, while self-rated health barely changed¹.

Sleep is the most reliably damaged and the least often asked about. Some 36.7 percent of US carers aged 45 and over got insufficient sleep, under 7 hours in 24, and 40.7 percent had two or more chronic diseases, rising to 53.4 percent among those aged 65 and over (CDC with the National Association of Chronic Disease Directors, 2018, using BRFSS data from 2015 to 2017)². Sleeping when you are on call covers what actually recovers some of it.

Intensity, not the label, drives most of this. Caregiving in the US 2025 found 20 percent of carers in fair or poor health overall, 38 percent reporting high emotional stress, 19 percent high physical strain and 24 percent feeling alone, with 44 percent in high-intensity caring situations³. The intensity split on health itself comes from the 2020 edition of the same survey, which found 27 percent of those in high-intensity situations in fair or poor health against 16 percent in medium to low intensity³.

In the United Kingdom, among carers responding to Carers UK’s survey run in June to August 2025, 35 percent rated their mental health bad or very bad, rising to 54 percent among carers aged 35 and under; 30 percent rated their physical health bad or very bad, up from 27 percent; 74 percent had felt stressed or anxious; 40 percent felt depressed; 43 percent felt lonely; 59 percent had lost touch with family and friends; and 65 percent found it hard to get a good night’s sleep (State of Caring 2025, 10,539 respondents)⁴. Carers UK states plainly that this is not a representative survey, and the sample was 82 percent female, 90 percent White British and 61 percent caring 50 or more hours a week, so read it as carers who responded rather than as carers in general.

Does caring shorten your life? The split verdict

No good evidence says it does, and here are both primary results in full so you can see why the popular version is wrong.

The 1999 study. Schulz R and Beach SR, “Caregiving as a risk factor for mortality: the Caregiver Health Effects Study”, JAMA 1999; 282(23): 2215 to 2219. A prospective population-based cohort nested in the Cardiovascular Health Study, four US communities, 1993 to 1998, average 4.5 years of follow-up, 392 caregivers and 427 non-caregivers aged 66 to 96, all living with their spouses. Adjusted against a reference group whose spouse was not disabled, the results were: caregiving with mental or emotional strain, relative risk 1.63, 95 percent confidence interval 1.00 to 2.65; caregiving with no strain reported, RR 1.08, CI 0.61 to 1.90; disabled spouse but not providing care, RR 1.37, CI 0.73 to 2.58⁵.

Five things every secondary source drops. The confidence interval’s lower bound is exactly 1.00, which is the value meaning no effect, so the result sits at the very edge of significance. It applies only to the strained subgroup, since unstrained carers showed no elevation. The population is narrow: older co-resident spousal carers aged 66 to 96 drawn from a cardiovascular cohort in four communities, not adult children, not non-resident carers, not younger carers. Strain was a single self-reported item, and 56 percent of this sample reported some or a lot of it, far above later population samples. And it is observational, so it cannot establish that caring caused the deaths⁵⁶.

The 2013 study, which points the other way. Roth DL, Haley WE, Hovater M, Perkins M, Wadley VG and Judd S, “Family caregiving and all-cause mortality: findings from a population-based propensity-matched analysis”, American Journal of Epidemiology 2013; 178(10): 1571 to 1578. It took 3,503 family caregivers in the REGARDS study and matched each one individually to a non-caregiver on 15 demographic, health-history and health-behaviour covariates, with a median 6.1 years of follow-up from 2003 to 2012. 264 caregivers, 7.5 percent, died against 315 matched non-caregivers, 9.0 percent: hazard ratio 0.823, 95 percent CI 0.699 to 0.969, an 18 percent reduced rate of death. Adult-child carers did better still, 3.0 percent against 5.2 percent. No subgroup showed increased mortality, including the strained-spouse subgroup that resembled the group at elevated risk in 1999⁷.

Five further population studies found reduced mortality, collected in a 2015 reappraisal that names the 1999 paper as the landmark result “often a centerpiece of advocacy”: O’Reilly 2008 RR 0.74, Brown 2009 HR 0.79, Fredman 2010 HR 0.74, Ramsay 2013 HR 0.83 and Roth 2013 HR 0.82⁶. The proposed explanation is the healthy-caregiver selection effect: healthier people are more likely to take on caring and more likely to stay in it, while people in poor health either never start or drop out, so comparing carers with the general population compares a positively selected group with an unselected one⁶.

And the later syntheses do not agree with each other, which is the part nobody reports. A meta-analysis of twelve longitudinal population studies found combined mortality 16 percent lower among carers, but the effect was not significant among the US studies alone and it disappeared where caregiving was defined precisely as help with activities of daily living rather than broadly. Its authors concluded their findings “provided more support for stress theory than the healthy caregiver hypothesis”, which cuts against a simple reading that caring is protective (Mehri and colleagues, Journal of Applied Gerontology 2021)⁸. A systematic review the following year found five longitudinal mortality studies with hazard ratios of 0.62 to 0.92, all significant, and attributed them to a healthy-carer bias in longitudinal observational studies. But its morbidity findings run the other way: a large longitudinal German health-insurance study found significantly raised incidence of severe stress, adjustment disorders, depression, spine disease and pain conditions among carers (Janson and colleagues, IJERPH 2022)⁹.

So the house position, and it is the honest one: consistent evidence of worse mental health and worse morbidity among strained carers, and no good evidence that caring shortens your life. Anyone who tells you caring raises your risk of death by a specific percentage is quoting one subgroup of one 1999 study, without its confidence interval, and without the twenty-five years of work that followed it.

Why the distinction matters practically

Because it changes what you do. If caring were simply lethal, the rational response would be to stop, which most carers cannot and do not want to do. What the evidence actually identifies as the harmful variable is strain, not caring, and strain is the thing that can be reduced.

In the 1999 study it was strained carers who showed the elevated estimate and unstrained carers who did not⁵. In the morbidity data it is stress-related and pain conditions that rise⁹. In the survey data the differences track intensity: 54 percent of carers aged 35 and under rated their mental health bad or very bad against 35 percent overall⁴, and high-intensity US carers were substantially more likely to report fair or poor health³. The target is the strain, and it is reachable.

What actually reduces the strain?

Skills training, in a group, has the best support, and it is still a modest result. NICE guideline NG97 recommends offering carers a psychoeducation and skills training intervention covering education about the condition and its progression, personalised strategies and skill-building, understanding and responding to behaviour change, adapting communication styles, advice on the carer’s own physical and mental health, planning enjoyable and meaningful activities, information about relevant services, and planning for the future, and adds that “carer interventions are likely to be most effective when provided as group sessions”¹⁰.

Beneath that, the evidence thins quickly, and NICE says so more frankly than most:

• Psychosocial and psychoeducational support is worded as “consider” rather than “offer”, because “overall, the quality of the quantitative evidence was low”, though “the evidence for group interventions was more convincing than for one-to-one interventions”¹¹.
• Psychotherapy and counselling: “The evidence on the effectiveness of psychotherapy and counselling was conflicting and was limited to cognitive behavioural therapy for carers of people living with dementia. There was not enough evidence to recommend these interventions for carers generally.”¹¹
• Advocacy: “There was no evidence in this area.”¹¹
• Mindfulness-based stress reduction, against an active control, improved depression, standardised mean difference −0.63, CI −0.98 to −0.28, and anxiety, mean difference −7.50, CI −13.11 to −1.89, both at low quality, from five trials of 201 carers. Burden did not improve and may have worsened, SMD 0.24, CI −0.11 to 0.58 (Cochrane CD012791, August 2018)¹².
• Telephone counselling improved depression at moderate quality, SMD 0.32, CI 0.01 to 0.63, but not burden, and Cochrane notes that all quantitative studies had a high risk of bias (CD009126, September 2014, 9 trials)¹³.
• Respite has no demonstrated benefit at all: four trials, 753 participants, very low quality evidence, and the conclusion that “current evidence does not demonstrate any benefits or adverse effects from the use of respite care for people with dementia or their caregivers”¹⁴. NICE’s own review of carers’ breaks found “a lack of evidence on effectiveness or cost effectiveness to support a recommendation about the circumstances in which carers’ breaks should or shouldn’t be offered”¹¹.

Arrange the respite regardless. Absence of demonstrated benefit is not the same as demonstrated absence, the Cochrane authors say so themselves¹⁴, and 65 percent of carers responding to Carers UK’s 2024 survey felt overwhelmed because they had not been able to take a break¹⁵. The point of knowing the evidence is not to talk you out of a break. It is so that when the break does not fix how you feel, you do not add that to the list of things you have failed at. Getting help in maps what exists, and respite care explained covers the forms it takes.

The six things that made the difference for me

None of this is a trial result. It is four years of what held up, offered as one household’s experience rather than as instruction.

1. Telling my own GP I was a carer and asking for it to be recorded. It changed what was offered, and it took ninety seconds. The equivalent step in the system is the carer’s assessment, and 77 percent of carers responding to Carers UK’s 2024 survey had not had one in the previous 12 months¹⁵. Carers skip their own appointments is about why almost nobody does this.
2. One fixed hour a week that belonged to nobody else. Not a hobby, not self-care. An hour with a boundary around it.
3. The folder. Being able to say “the referral was made on the fourteenth, by the ward sister” instead of “I think somebody referred her” removed an enormous amount of low-grade rage. Keeping a care file.
4. Getting the sibling conversation onto a shared document instead of into a phone call. See sharing care with siblings.
5. Accepting paid help a year later than I should have. It did not solve anything; it made the thing survivable, which is different.
6. Saying the resentment out loud. Guilt, resentment and anger exists because that feeling is the ordinary experience of the job, not evidence that you are doing it wrong.

What burnout looks like before it is a crisis

Persistent exhaustion that sleeping does not touch. Irritability out of proportion to the trigger. Dread in the hallway before you go into the room. Loss of interest in things you used to want. Withdrawal from friends, which 59 percent of Carers UK’s 2025 respondents reported as lost touch with family and friends⁴. Headaches, stomach trouble, a chest that feels tight. Going through the tasks correctly and feeling nothing.

The tell that matters most is when you start describing your own health in the past tense, or when the phrase “once this is over” begins doing all the work in your sentences. That is the point to ask for a carer’s assessment in your own right, rather than the point at which you have collapsed and somebody else has to.

This is a call today, for you and for them

For you. If your own mood, your drinking, or thoughts of harming yourself or the person you care for are the reason you are reading this, that is a call to your own doctor or an urgent mental health line where you live, today. Not after the next respite booking, not once things settle. Being the carer makes it more urgent, not less. Say the words “I am an unpaid carer and I am not coping”, because in several systems that sentence changes what is offered.

For them. Ring the same day if the person you look after is more confused than they were yesterday or the confusion comes and goes, has fallen, has stopped eating or drinking, has a new fever, or has new or worsening pain, including agitation in somebody who cannot tell you what hurts. In the United Kingdom, their GP surgery or NHS 111 out of hours. In the United States, their doctor’s office or the nurse line their plan runs. In Australia, their GP or healthdirect on 1800 022 222. In Canada, their family doctor or the provincial health line, 811 in most provinces. In Ireland, their GP or the out of hours GP service. For changed breathing, chest pain, one-sided weakness, changed speech, or bleeding that will not stop, use the emergency number: 999 or 112, 911, or 000.

Exhaustion is a reason to ring sooner, not a reason to wait and see. The most dangerous sentence in this job is “I will give it until the morning”.