In eight European countries, obesity added hospital nights. Underweight added far more

European health ministries have spent the past two years arguing about who pays for obesity drugs. A study published this week reframes the question from the other end of the ledger: what obesity already costs European hospitals in bed-nights, and how much of that a small shift in prevalence would recover [s1].

The analysis used wave 3 of the European Health Interview Survey, collected in 2019, covering adults aged 18 and over in eight countries — Belgium, the Netherlands, Germany, Luxembourg, Spain, Italy, Portugal and Denmark [s1]. The outcome was the number of hospital nights a respondent reported in the preceding 12 months. The authors fitted negative binomial regression models with zero-inflated specifications, which is the standard way to handle a variable where most people report zero and a minority report a great many [s1].

What it found

Compared with people of normal weight, respondents with obesity reported more hospital nights, with an incidence rate ratio of 1.19 [s1]. Pre-obesity — the BMI band between normal weight and obesity — showed no such association at all [s1].

The larger signal was at the other end of the BMI distribution. Underweight respondents had an incidence rate ratio of 1.92, close to double the hospital nights of the normal-weight group [s1]. That is a much stronger association than the one for obesity, and it is worth stating plainly because it is the sort of finding that gets lost when a paper is summarised by its title.

It is also the finding most vulnerable to reverse causation. In a cross-sectional survey, low body weight can just as easily be a consequence of serious illness as a cause of hospital admission — cancer, advanced organ disease and frailty all reduce weight. The study design cannot separate those directions. The same caution applies, more weakly, to the obesity estimate.

Other associations in the model behaved as expected. Hospital nights rose markedly with age [s1]. Hypertension, diabetes and depressive symptoms each substantially raised hospital use [s1]. Higher education and higher income were each protective, reducing stays by 7% and 8% respectively [s1]. Physical activity was associated with a 28% reduction — the single largest protective coefficient in the analysis [s1].

Geography mattered too. Southern and Northern European respondents reported lower hospitalisation rates than those in Western Europe [s1]. The paper does not resolve whether that reflects differences in disease burden, in admission thresholds, or in how each health system uses inpatient beds.

The €4.27 billion figure

The headline economic result comes from a simulation rather than an observation. The authors modelled what would happen if obesity prevalence fell by one percentage point across the eight countries, and estimated annual savings of €4.27 billion — equivalent to 1.71% of the combined hospital budgets in those countries [s1].

That is a real number attached to a hypothetical. Nothing in the study demonstrates that any particular intervention can move population obesity prevalence by a percentage point, or that hospital budgets would actually shrink if it did. Hospital costs are famously sticky; freed capacity tends to be absorbed by other demand rather than returned to treasuries. The estimate is best read as a measure of how much inpatient activity is statistically associated with obesity, not as a forecast of cash savings.

Where it sits alongside smoking

The comparison is instructive because a second team used the same survey wave, published a week earlier, to do the equivalent exercise for tobacco [s2]. That analysis covered all EU member states except France, Germany, Ireland and Sweden, which were excluded for data access and data quality reasons [s2]. It found smoking positively associated with both inpatient and daycare hospitalisation, with smoking-attributable inpatient costs ranging from about €5 billion in Spain down to €28 million in Latvia, and daycare costs from €550 million in the Netherlands down to €3 million in Malta and Slovenia [s2].

The smoking authors were explicit that their totals understate the true burden, because they modelled only two categories of service [s2]. The same structural limitation applies to the obesity paper, which counted hospital nights and nothing else — no outpatient care, no primary care, no medication, no lost productivity.

What this does and does not establish

The European Health Interview Survey is self-reported. Height, weight, hospital nights and smoking status all come from respondents rather than from clinical records, and all four are known to be reported with systematic bias — weight downward, height upward. The 2019 collection also predates both the pandemic and the arrival of GLP-1 receptor agonists at scale, so it describes a European hospital system that no longer quite exists.

What the analysis does provide is a like-for-like comparison across countries using one harmonised instrument, which is otherwise very hard to assemble in Europe. National hospital datasets are not built to be compared. Both papers make the same argument from that vantage point: the cross-national evidence quantifying these burdens has been sparse, and prevention is being discussed without a shared denominator [s1][s2].

What to watch is whether EHIS wave 4 shows the associations moving, and in which direction. If the obesity coefficient holds while prevalence rises, the modelled savings grow. If incretin therapies compress the gap, that will show up in hospital nights before it shows up anywhere else.

Sources

• [s1] Estimating the hospitalisation burden of obesity in Europe: evidence from the European health interview survey (EHIS) wave 3 — BMC Public Health, 10 March 2026. https://doi.org/10.1186/s12889-026-26705-w
• [s2] Smoking-attributable hospitalisation costs in Europe: an econometric analysis of the European Health Interview Survey, 2019 — BMJ Public Health, 3 March 2026. https://doi.org/10.1136/bmjph-2025-003101