Early procalcitonin kinetics and 28-day mortality in critical illness: a pragmatic Bayesian modelling approach

Serial laboratory measurements are routinely obtained in critically ill patients, yet prognostic assessment often relies on isolated baseline values. Using procalcitonin (PCT) as a worked example, we evaluated whether simple kinetic features derived from early serial measurements were associated with 28-day mortality in ICU patients with suspected sepsis.

We analysed a retrospective ICU cohort in which PCT measurements were aligned to the first available value (day 0) and evaluated over an early observation window (days 0-5). PCT concentrations were log-transformed [log(1 + PCT)]. Two kinetic features were derived: the slope of log-transformed PCT over time and the area under the curve (AUC). Bayesian logistic regression models were fitted to predict 28-day mortality using (i) baseline PCT, (ii) early PCT kinetic features, and (iii) kinetic features with selected clinical covariates.

The analytic cohort comprised 128 ICU patients, with a 28-day mortality of 32.8 %. Early PCT trajectories demonstrated substantial inter-individual heterogeneity. In the kinetics-only model, higher AUC was associated with increased mortality (odds ratio [OR] 1.83, 95 % credible interval [CrI] 1.25-2.74), while slope showed greater uncertainty (OR 1.42, 95 % CrI 0.98-2.12). The kinetics-only model achieved the highest expected log predictive density, although differences relative to the baseline PCT model were modest and uncertain. Addition of selected clinical covariates did not materially improve predictive performance.

Simple kinetic summaries derived from early serial biomarker measurements may retain prognostic information beyond baseline values alone. Using PCT as an example, this study illustrates a pragmatic approach to incorporating early laboratory dynamics into prognostic assessment in critically ill patients.