Table 3 Mortality prediction performance of APRICOT-M compared to acuity baseline for episode and step levels

From: Real-time prediction of intensive care unit patient acuity and therapy requirements using state-space modelling

Cohort

Model

Prevalence (episode)

AUROC (episode)

AUPRC (episode)

Prevalence (step)

AUROC (step)

AUPRC (step)

Development

SOFA

0.052

0.88 (0.87–0.89)

0.36 (0.33–0.38)

0.003

0.83 (0.82–0.84)

0.02 (0.02–0.02)

APRICOT-M

0.93 (0.93–0.94)*

0.55 (0.52–0.58)*

0.94 (0.94–0.95)*

0.22 (0.20–0.24)*

External

SOFA

0.058

0.85 (0.85–0.85)

0.33 (0.32–0.35)

0.003

0.78 (0.78–0.79)

0.02 (0.01–0.02)

APRICOT-M

0.93 (0.93–0.94)*

0.59 (0.58–0.61)*

0.95 (0.94–0.95)*

0.23 (0.22–0.24)*

Prospective

SOFA

0.027

0.92 (0.83–0.99)

0.28 (0.12–0.57)

0.001

0.80 (0.65–0.92)

0.00 (0.00–0.01)

APRICOT-M

0.99 (0.97–1.00)

0.78 (0.45–1.00)*

0.99 (0.97–1.00)*

0.48 (0.05–0.86)*

  1. AUPRC area under the precision-recall curve, AUROC area under the receiving operating characteristic, SOFA sequential organ failure assessment. Performance is the median AUROC and AUPRC across a 100-iteration bootstrap with replacement, with 95% Confidence Intervals in parenthesis. P-values are based on pairwise two-sided Wilcoxon rank sum tests.
  2. Bold values represent performance metrics that are significantly higher based on statistical analysis.
  3. *p-value < 0.05 compared to SOFA.
Back to article page