Burnout has an official definition, and almost nobody who uses the word has read it. It sits in the eleventh revision of the International Classification of Diseases as code QD85, and it describes a syndrome resulting from chronic workplace stress that has not been successfully managed, with three dimensions: exhaustion, mental distance or cynicism about the job, and reduced professional efficacy.

Then comes the sentence that does the damage. Burnout, the entry says, refers specifically to phenomena in the occupational context and should not be applied to describe experiences in other areas of life.

Not a warning. An instruction. The word belongs to paid employment, and the classification says so in as many words.

Two things everyone got wrong in 2019

When that entry was published, headlines reported that the WHO had classified burnout as a medical condition. The WHO issued a correction the same week. Burnout sits in the chapter on factors influencing health status and contact with health services, which is explicitly the section for things that are not diseases. It is a reason people come to services, not a diagnosis they receive.

The second error was calling it new. Burnout was already in ICD-10 as Z73.0. What changed in 2019 was that the definition became more specific, and the specificity is exactly what created the problem this piece is about.

The construct itself goes back to Christina Maslach and Susan Jackson in 1981, and the inventory they built was designed for human-services workers. It has never had a validated clinical cutoff, which is why every prevalence figure you have ever read is really a report about where the authors drew a line. The clearest demonstration of that is a review of burnout in medical students, which found prevalence estimates ranging from 7.0 percent to 75.2 percent across 58 studies, depending entirely on the instrument and the threshold chosen.

Four burnouts, one symposium

At this year’s European Health Psychology Society meeting in Pafos, there was a symposium on burnout in all its forms, and the forms were four: professional, parental, academic, and the burnout of informal caregivers. What struck me was not that researchers are studying all four. It is that only one of them is allowed to use the word officially.

The parental version has been the most thoroughly built out. Isabelle Roskam, Maria Elena Brianda and Moïra Mikolajczak developed the Parental Burnout Assessment in 2018, and it has four dimensions rather than three. Three of them map onto the job version: exhaustion in the parental role, emotional distancing from your children, a loss of the sense of being effective as a parent. The fourth has no occupational equivalent at all.

It is called contrast with one’s previous parental self. The measurable experience of no longer recognising the parent you used to be.

Nobody has ever needed a scale item for no longer recognising the employee you used to be. In the parental version, it is one of the four legs the construct stands on.

What 42 countries showed

The largest study of it covers 17,409 parents across 42 countries, and the finding that gets repeated is that the strongest country-level correlate of parental burnout was individualism, correlating with prevalence at 0.53 and with mean levels at 0.50, more strongly than economic inequality, family size or hours spent with children.

Three warnings come attached to that. It is an ecological correlation at country level, which says nothing about any individual parent in any of those countries. The samples were convenience samples of very unequal size, around 71 percent mothers. And the prevalence percentages shift depending on which cutoff is used, so a range is the honest way to quote them, or no number at all.

The mechanism work is more useful for a reader than the geography. A later study of 16,059 parents in 36 countries found that the strongest mediator was the discrepancy between the parent a person believes they are supposed to be and the parent they actually are.

Not the workload. The gap.

It is cross-sectional, so causal direction is not established, and it is still the sentence I expect most parents to recognise themselves in.

And parental burnout is not job burnout relocated. Mikolajczak and colleagues found in Child Abuse and Neglect that it has consequences job burnout does not, including escape ideation and effects on neglect and violence. That work is cross-sectional and self-reported, and the variance figures attached to it are routinely mangled in summaries into causal claims they cannot support.

The caregiving statistic that turned out to be backwards

The informal-caregiver strand carries the clearest correction in this whole area, and it matters because the frightening version is still quoted in caregiving literature.

In 1999, Richard Schulz and Scott Beach published a paper in JAMA reporting that caregivers had a 63 percent higher risk of dying over four years than non-caregivers. That figure has been in circulation ever since.

Two things get dropped from it. The 63 percent applied only to caregivers reporting strain, with a confidence interval running from exactly 1.00 to 2.65, which is as borderline as a result can be while still counting. Caregivers who were not reporting strain showed a relative risk of 1.08 with a confidence interval crossing one, meaning no detectable effect at all.

Then the population studies arrived. David Roth and colleagues propensity-matched 3,503 caregivers to 3,503 non-caregivers and found caregivers dying at a lower rate, with a hazard ratio of 0.823. A review in The Gerontologist reported that all five population-based studies conducted since 1999 pointed the same way, and the authors called for a more balanced narrative.

That reversal is not evidence that caregiving is good for you. The likeliest explanation is healthy-caregiver selection: people take on caregiving when they are well enough to, so the group is healthier before anyone starts counting. What the whole sequence establishes is narrower and more actionable. The harm attaches to strain, not to caring, and strain is modifiable in ways that the fact of caring is not.

The scale of the population involved is not small. The 2020 report from the National Alliance for Caregiving and AARP put the number of US adults providing unpaid care in the previous year at 53.0 million, or 21.3 percent of adults, with 23 percent saying caregiving had made their own health worse. The 2025 successor report gives a considerably higher figure, so it is worth naming which year any number comes from.

The concept that does not ask whether you were paid

If the word burnout is contractually restricted to employment, there is a physiological framework that has no such restriction, and it is the one I keep returning to.

Bruce McEwen and Eliot Stellar introduced allostatic load in 1993 as the cost of chronic exposure to fluctuating or heightened neural and neuroendocrine response. Allostasis is the achievement of stability through change: the body adjusts to demand. Allostatic load is the accumulated bill for having adjusted, repeatedly, for a long time.

It was operationalised by Teresa Seeman and colleagues as a ten-biomarker index in 1,189 adults aged 70 to 79, predicting mortality and cognitive and physical decline over seven years. That was a narrow cohort and does not generalise cleanly to midlife. A later meta-analysis put high allostatic load at a hazard ratio of 1.22 for all-cause mortality, with heterogeneity above 90 percent, which is high enough to treat the pooled figure with real caution.

The concept has its own measurement problem, which is exactly what the constructs debate exists for. A review of 26 studies found that all of them included a cardiovascular or metabolic marker while only about half included any measure of the HPA axis, meaning half of allostatic-load research contains no measurement of the stress hormones the idea was built around.

Still, the framework does the thing the burnout definition refuses to do. It does not ask who was paying you. A body under sustained demand accumulates cost whether the demand came from a manager, a dependent parent, a thesis committee or a four-year-old.

What the missing word costs

I am a researcher, not a clinician, and this is not clinical guidance. What I can say is that naming has consequences that are not merely semantic. A recognised occupational syndrome gets workplace policy, organisational interventions, occupational health referrals and research funding. The parent, the postgraduate and the daughter managing her father’s appointments get told they are tired.

If the burnout-versus-depression question comes up, and it always does, the honest position is that the overlap is large and the argument is about the remainder. Two figures sit at the edges of it: an exhaustion-depression correlation of .80 in one analysis after correcting for unreliability, and .52 observed in another. Most of the gap is a methodological choice about that correction, and declaring a winner would be overstepping what the evidence supports.

What is not in dispute is that four groups of people are showing up in four separate literatures with substantially the same complaint, and that one of them has a word and three do not. If any of this is landing closer to home than it is landing as an interesting argument, that is worth taking to a professional rather than to another article about it.