Early inflammatory and nutritional marker recovery patterns and short-term outcomes in critically ill older adults: a...

Admission severity scores describe acute organ dysfunction but may not capture early recovery in older intensive care unit (ICU) patients. We examined whether investigator-defined changes in routinely measured inflammatory and nutritional markers during the first 72 h identified distinct recovery patterns associated with short-term outcomes.

This single-center retrospective cohort included patients aged 65 years or older admitted from January 5, 2021, through May 31, 2026, who survived to a 72-h landmark and had first-window measurements at ICU admission (T0, 0-24 h) and 48-72 h (T1). Four rule-based recovery patterns were defined and assigned using internally derived marker thresholds and fixed directional/count rules finalized without reference to outcomes. The primary outcome was mortality from the landmark to day 28. Cox models sequentially adjusted for demographic factors, baseline clinical severity, and geriatric vulnerability. Within each bootstrap sample, pooled marker thresholds were re-estimated, patients were reassigned using the fixed operational rules, and the outcome models were refitted.

Among 934 screened admissions, 620 patients entered the landmark cohort and 122 died by day 28. Pattern 3 (initial extreme burden without improvement-rule fulfillment) included 180 patients and had the highest crude mortality (31.7%, 57/180), versus 9.8% (25/255) in Pattern 1. In the geriatric-adjusted model, the overall four-level association with mortality was not significant (likelihood-ratio p = 0.196), and none of the three non-reference patterns was independently associated with mortality; the hazard ratio for Pattern 3 versus Pattern 1 was 1.64 (95% confidence interval, 0.97-2.79; p = 0.066). In exploratory analyses, overall recovery-pattern associations were observed for mechanical ventilation longer than 7 days (FDR-adjusted p = 0.002) and ICU stay longer than 14 days (FDR-adjusted p = 0.001); the ventilation estimates were non-monotonic across patterns, whereas all three non-reference patterns had higher odds of prolonged ICU stay. Adding the recovery pattern did not improve optimism-corrected discrimination (0.747 to 0.745; DeLong p = 0.365).

Investigator-defined early marker recovery patterns described clinical heterogeneity but were not independently associated with 28-day mortality after geriatric adjustment and did not add predictive value. Exploratory overall associations with prolonged ventilation and ICU stay require external validation. Prospective evaluation is required before clinical use.