Ask a nursing aide what wears her down and she may not say the shouting, or the hitting, or the night a resident with dementia tried to walk out into the dark. In a survey of staff across one Hungarian county, the single most heavily rated difficulty was quieter than any of that: the demand for constant readiness and attention, which scored a mean of 3.60 on a five point scale, ahead of the physical strain of caring for residents confined to bed. It is the part of the job that does not switch off, the low hum of vigilance that follows a person out of the building at the end of a shift. The finding comes from a cross-sectional questionnaire study by Hedvig Kiss of Semmelweis University and colleagues at Debrecen and Nyiregyhaza, published in BMC Nursing on 30 March 2026.

The numbers behind it are unusually complete. Researchers approached the heads of 97 residential social institutions in Szabolcs-Szatmar-Bereg county in northeastern Hungary; 78 agreed, and the survey ran in 77 of them between 1 June and 31 July 2024. Against a nominal workforce of 1,454 professional staff, 1,428 people answered, a response rate of 98.21 percent, with 1,254 forming the final sample and 1,207 entering the regression models. Nine in ten respondents were women (93.3 percent), the largest age band was 41 to 50 (36.8 percent), and roughly three quarters held vocational social nurse or caregiver qualifications, with 11.3 percent holding a degree. Burnout was captured not by a standard clinical instrument but by a four level severity scale (0 to 3) built from what respondents said about themselves and about colleagues. Some signal of burnout was reported by 46.4 percent; 19.3 percent sat in the two highest levels.

Within individuals, the pattern looked the way the Job Demands-Resources model predicts. In adjusted ordinal models, higher subjective workload was associated with greater burnout severity (odds ratios of roughly 1.46 to 1.53, p less than 0.001), as was running into care difficulties more often (roughly 1.21 to 1.31, p less than 0.01). The sharpest demand-side finding was about resources rather than load: staff who said they had no possibility to cope had close to three times the odds of higher burnout severity (roughly 2.86 to 2.91, p less than 0.01). Having previously learned something about burnout was associated with much lower severity (odds ratios around 0.06 to 0.08, p less than 0.001), a figure so large that it should be read with care, since this is a single snapshot in time and the authors themselves note that cross-sectional data cannot establish which came first. Notably, how often people managed to rest or relax was not a significant predictor at all (OR 1.09, 95 percent CI 0.95 to 1.24, p = 0.212), which is an uncomfortable result for anyone who believes the answer to this is a better weekend.

Then comes the part that reframes the story. Because staff were nested inside institutions, the team fitted a multilevel cumulative link mixed model with a random intercept for each facility. It fitted the data significantly better than the single level model (AIC of about 1,994 against about 2,042, a difference of 48.165), and the intraclass correlation coefficient came out at 0.206: roughly 21 percent of the variance in burnout severity was attributable to differences between institutions. Two aides doing the same work, with the same caseload and the same training, in two different buildings in the same county, were not facing the same odds. Burnout, on this evidence, is partly a property of the place.

It matters that the study stops exactly there. The random intercept measures how much facilities differed; it does not say why. This survey did not test staffing ratios, access to backup on a bad shift, or the quality of management support, so any account that names those as the cause is going beyond what was measured. What the authors do propose as concrete, countable targets are institutional training coverage, with an emphasis on burnout and on the behavioural and psychological symptoms of dementia, and shrinking the share of staff who answer that they have no way to cope. Those are facility-level levers, not personal ones, and that is the point: resilience workshops aimed at the individual are being asked to carry weight that a fifth of the variance says they cannot.

For a family choosing or already living with a residential placement, none of this is a verdict on any particular home, and nothing here identifies good care or bad care in a building you can name. But it does change the shape of a reasonable question. Asking whether the staff on your relative's unit have had dementia-specific and burnout-specific training, and whether anyone on a hard shift has somewhere to turn, is not a soft question about morale. In this county, in this one summer, it was the kind of question that separated one institution's numbers from another's.