Association between early ICU fever burden and 28-day mortality in patients with acute ischemic stroke: a retrospective cohort study based on the MIMIC-IV database.
Ma C, Zheng X, et al. • Frontiers in neurology • 2026
In ICU patients with acute ischaemic stroke, greater fever burden during the first 72 h after ICU admission was associated with higher 28-day all-cause mortality, with adjusted HRs of 1.404 and 1.948 for low and high fever burden groups respectively compared with no fever burden.
Key Findings
Results
Higher fever burden during the first 72 hours of ICU admission was independently associated with increased 28-day all-cause mortality in acute ischemic stroke patients.
In the fully adjusted multiply imputed Cox model, low fever burden was associated with an adjusted HR of 1.404 (95% CI 1.126–1.751) compared with no fever burden.
High fever burden was associated with an adjusted HR of 1.948 (95% CI 1.564–2.427) compared with no fever burden.
A dose-response relationship was observed, with progressively higher hazard ratios from no to low to high fever burden groups.
Fever burden was calculated as the area under the temperature-time curve above 37.9°C during the first 72 h after ICU admission, standardized to 24 h.
Results
The 28-day all-cause mortality rates increased substantially across fever burden groups.
28-day mortality was 15.3% in the no fever burden group (n=2,229).
28-day mortality was 28.7% in the low fever burden group (n=394).
28-day mortality was 36.6% in the high fever burden group (n=393).
Total cohort included 3,016 adult ICU patients with acute ischemic stroke from the MIMIC-IV database.
Results
The association between fever burden and 28-day mortality was robust across multiple sensitivity analyses.
Findings were consistent across complete-case analysis, inverse probability of treatment weighting (IPTW) analysis, alternative exposure windows, and landmark analyses.
Results were robust to alternative temperature thresholds, adjustment for temperature measurement density, and exclusion of patients with sepsis.
These sensitivity analyses were performed to assess the robustness of the primary findings.
Methods
Fever burden, defined as the area under the temperature-time curve above a threshold, was used to capture cumulative thermal exposure rather than single-point temperature measurements.
The threshold temperature was set at 37.9°C.
Fever burden was calculated over the first 72 hours after ICU admission and standardized to 24 hours.
Patients were classified into three groups: no fever burden, low fever burden, and high fever burden.
The authors noted that 'single temperature values may not adequately capture the cumulative effect of sustained thermal exposure.'
Results
The majority of included acute ischemic stroke ICU patients had no fever burden during the first 72 hours.
Of 3,016 total patients, 2,229 (73.9%) had no fever burden.
394 patients (13.1%) had low fever burden and 393 patients (13.0%) had high fever burden.
The study included adult patients with acute ischaemic stroke admitted to the ICU for the first time, drawn from the MIMIC-IV database.
Cox proportional hazards models with multiple imputation were used for the primary analysis.
Conclusions
The authors concluded that early ICU fever burden may provide time-integrated prognostic information but that its clinical utility requires further prospective validation.
The study was retrospective in design, using the MIMIC-IV database.
The authors stated that 'its clinical utility and therapeutic implications require further prospective validation.'
Subgroup analyses were also performed to further assess the robustness of findings.
What This Means
This research suggests that the total amount of fever a stroke patient experiences in the first three days of an ICU stay — not just whether they have a fever at any single point — is meaningfully linked to their risk of dying within 28 days. Researchers analyzed data from over 3,000 adult ICU patients with acute ischemic stroke (the most common type of stroke, caused by a blocked blood vessel) from a large hospital database. They measured 'fever burden' as the total area of elevated temperature above 37.9°C over time, divided patients into no, low, and high fever burden groups, and found that 28-day death rates were 15.3%, 28.7%, and 36.6% respectively. Even after accounting for many other factors, patients with the highest fever burden were nearly twice as likely to die within 28 days as those with no fever burden.
This research suggests that tracking cumulative fever exposure — rather than simply noting whether a patient has a fever at a given moment — may give doctors more useful information about a stroke patient's prognosis. The findings held up across multiple different statistical methods and sensitivity tests, including after removing patients who also had sepsis (a severe infection that independently causes both fever and death), which strengthens confidence in the association.
However, because this was a retrospective study using existing hospital records, it cannot prove that fever directly causes worse outcomes, nor does it tell us whether treating fever would improve survival. The authors specifically note that the clinical and therapeutic implications need to be tested in future prospective studies before any changes to care practices could be recommended.
Ma C, Zheng X, Tang S, Mo Y, Wang X, Zeng J, et al.. (2026). Association between early ICU fever burden and 28-day mortality in patients with acute ischemic stroke: a retrospective cohort study based on the MIMIC-IV database.. Frontiers in neurology. https://doi.org/10.3389/fneur.2026.1900723