Inflammation, agitation and hallucinations as mortality markers in hospitalised delirium

Delirium carries a high risk of death, but it is unclear which bedside markers identify the patients at greatest risk once delirium is established. Electrolyte disturbance is usually modelled as a single binary abnormality, merging deficiency and excess states that arise through opposite mechanisms.

We studied 515 consecutive inpatients diagnosed with delirium by a consultation-liaison psychiatrist using DSM-5 criteria between January 2018 and April 2022 at a university hospital. Each electrolyte was modelled in three directional states against explicit reference ranges. The outcome was in-hospital mortality. Multivariable logistic regression was fitted on the complete cohort with age forced in.

In-hospital mortality was 51.1%. Non-survivors had higher C-reactive protein (CRP; median 96.1 vs 51.0 mg/L, p < 0.001), higher Richmond Agitation-Sedation Scale (RASS) scores (p < 0.001) and more often documented hallucinations (84.0% vs 71.0%, p = 0.001). Mortality was U-shaped across directional electrolyte states, peaking in hypernatremia (64.1%) and hyperkalemia (65.3%). In the primary model, CRP (odds ratio [OR] 1.069 per 10 mg/L, 95% confidence interval [CI] 1.042-1.096, p < 0.001), documented hallucinations (OR 1.840, 95% CI 1.158-2.924, p = 0.010) and RASS (OR 1.153, p = 0.048) were independently associated with death, whereas hyperkalemia was borderline (OR 1.876, p = 0.065). The optimism-corrected AUC was 0.667, with adequate calibration (Hosmer-Lemeshow p = 0.243).

Systemic inflammation was the strongest marker of in-hospital death in delirium, and the clinical phenotype itself, agitation and perceptual disturbance, carried independent prognostic information. Directional modelling located the electrolyte signal in excess rather than deficiency, an asymmetry that binary coding conceals. Prospective studies incorporating comorbidity indices are needed.