Health care has more burnout research than almost any other sector, and most of it arrives as a single averaged answer: this works, that does not. A systematic review published in Annals of internal medicine in 2026 did something more useful by refusing to average across professions.
The result is the most interesting thing in it. The same intervention helped one group and not another. [pubmed-burnout-jul08-2026-source]
What the reviewers did
They searched for randomised trials of any intervention intended to reduce burnout in health care professionals, compared against no active intervention, and found 99 trials covering 9,330 participants. Study selection and data extraction were each done independently by two people.
Crucially, results were analysed stratified by role rather than pooled into one figure, using the standard burnout components: emotional exhaustion, depersonalisation, and sense of personal accomplishment.
What they found
For physicians: professional coaching probably reduced emotional exhaustion and depersonalisation. Mindfulness-based interventions may not have worked at all, with no detectable effect on depersonalisation and a wide, inconclusive estimate for exhaustion.
For nurses and midwives: mindfulness-based interventions may have reduced emotional exhaustion substantially, by one of the larger effects in the review.
Across mixed groups of health professionals: mindfulness improved emotional exhaustion, depersonalisation and sense of personal accomplishment.
Across the board: the interventions that worked generally ran longer than four weeks.
Standardised mean differences reported by Collett and colleagues (2026), Annals of internal medicine, shown as magnitudes. The mindfulness estimate for physicians is omitted rather than plotted: its range crossed zero, so the review could not distinguish it from no effect. Most estimates were graded low certainty.
The finding inside the finding
Take those first two results together. Mindfulness training may do little for a doctor and a good deal for a nurse.
That is worth sitting with, because burnout interventions are usually procured for an entire organisation at once. A hospital that buys one mindfulness programme for all clinical staff is, on this evidence, buying something that may work for some of them and not others.
The review cannot explain why, and any explanation offered here would be speculation. But the practical implication does not require knowing the mechanism: match the intervention to the role, and be suspicious of anything sold as working for everyone.
The duration finding cuts the same way. Programmes lasting more than four weeks were the ones with effects. That rules out most of what organisations actually deliver, which is a wellbeing day, a lunchtime session, or a module in mandatory training.
What the whole review cannot address
Every intervention in all 99 trials was delivered to individuals.
That is not a criticism of the reviewers, who summarised the evidence that exists. It is a description of the evidence base, and it has a consequence: this literature can tell you how to help people withstand their working conditions, and it cannot tell you anything about changing those conditions.
The World Health Organization defines burnout as an occupational phenomenon resulting from chronic workplace stress that has not been successfully managed. The definition points at the workplace. The research points at the worker.
Nothing here tests staffing levels, rota design, administrative load, or how much control someone has over their own day, all of which are strongly associated with burnout and none of which an individual can fix by being coached.
The certainty problem
Most findings were graded low or very low certainty. Most trials were unblinded, and outcomes were self-reported.
That combination matters more than it might sound. Someone who has just completed an eight-week programme designed to help them, knows they received it, and is then asked to rate their exhaustion, has several reasons to report improvement beyond the programme working. It is very hard to blind a mindfulness course, so this is a structural limitation of the field rather than sloppiness.
Read the effect sizes as upper bounds rather than best estimates.
If you are burnt out at work
The transferable version of this review, for anyone outside healthcare:
- Match the intervention to the job. What helps someone with high autonomy and overwhelming demands differs from what helps someone with no control over their schedule.
- Anything shorter than a month is unlikely to do much. A one-day workshop is not a dose.
- The individual route has a ceiling. If the cause is the workload, coping better with the workload is a holding action.
Our burnout guide covers the three components, how burnout differs from ordinary stress, and why recovery takes more than rest.
The source
These findings are drawn from “Efficacy of Individual-Level Interventions to Mitigate the Risk for Burnout Among Health Care Professionals: A Systematic Review and Meta-analysis of Randomized Controlled Trials” (Collett G, Gupta J, Eltayeb A, et al., 2026), published in Annals of internal medicine. Read the full study on PubMed.