Temperature at admission and mortality in older adults with infection: Limited prognostic value in non-sepsis cases

Temperature abnormalities are established prognostic markers in sepsis, but their predictive value in older adults with infection without sepsis remains unclear.

To examine the association between arrival temperature and 28-day all-cause mortality among emergency department patients aged ≥65 years with infection, stratified by sepsis status.

We performed a post-hoc analysis of a prospective cohort from a Danish emergency department. Sepsis was defined as infection plus an acute ≥2 point increase in the Sequential Organ Failure Assessment (SOFA) score. Propensity score matching balanced covariates between groups. The restricted cubic spline regression modelled non-linear temperature-mortality associations. Sensitivity analyses excluded patients with comorbidities that affected baseline SOFA and used the National Early Warning Score (NEWS2) to classify likely sepsis.

Among 1431 patients (median age 78.9 years; 49.3% male), 545 (38.1%) met sepsis criteria. In non-sepsis patients, mortality was stable across 36-41 °C with no statistically significant association (global spline p = 0.320), although a modest increase was observed below 36 °C. In sepsis, hypothermia (<36 °C) was associated with higher mortality, whereas fever (>38 °C) was protective. At temperature extremes (<36 °C and > 39 °C), confidence intervals widened substantially due to sparse data. Findings were consistent across sensitivity analyses.

In older adults with infection but without sepsis, arrival temperature did not predict 28-day mortality. In sepsis, hypothermia was associated with higher mortality, and fever with lower mortality. The prognostic interpretation of very low or very high temperatures remains uncertain, suggesting clinical caution in these temperature ranges.