Answers are grounded in this paper and cite their sources.
Background: Toxic encephalopathy (TE) covers a group of neurological disorders triggered by external toxins. Mortality is high, yet clinicians still lack a simple marker for early risk sorting. The albumin-corrected anion gap (ACAG) combines the anion gap with serum albumin and may summarize both metabolic and inflammatory strain. We tested whether ACAG tracks mortality in TE. Methods: This retrospective study drew on MIMIC-IV version 3.1. Adults with TE were eligible when anion gap and albumin were available at the first ICU admission. ACAG was modeled as a continuous variable and as tertiles. Cox models estimated 28-, 90-, and 365-day all-cause mortality under stepwise adjustment. Restricted cubic splines described the dose-response pattern; Kaplan-Meier curves, subgroup analyses, and incremental prediction measures (AUC, NRI, IDI, calibration, and decision curve analysis) were also applied. E-values were calculated. Results: Altogether 1,120 patients entered the analysis. Mean age was 63.6 ± 16.7 years and 53.3% were male. Mortality at 28, 90, and 365 days was 22.3%, 30.3%, and 38.3%. In the fully adjusted model, each 1-mmol/L rise in ACAG carried an HR of 1.05 (95% CI 1.02-1.08) at 28 days, 1.04 (1.02-1.07) at 90 days, and 1.04 (1.02-1.06) at 365 days. The top tertile also had higher mortality than the bottom tertile at all three horizons (HR 1.58, 1.61, and 1.55). E-values were 2.54, 2.60, and 2.47. The non-linearity tests were not significant ( P = 0.908, 0.313, and 0.211). AUC gains were small; NRI and IDI reached significance. Conclusions: Admission ACAG was independently linked to short- and long-term death in TE, and the pattern was close to linear. Its added predictive value was modest. These data are observational and need external confirmation.
No takes yet. Share an insight, caveat, or question.
Sun et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: