Table 2 APRICOT-M performance for 4-h predictions in terms of AUROC for three study cohorts

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

Outcomes

Development

External

Prospective

Primary

   

Discharge

0.85 (0.85–0.86)

0.85 (0.85–0.85)

0.86 (0.82–0.90)

Stable

0.91 (0.91–0.91)c

0.87 (0.87–0.87)a,c

0.92 (0.91–0.92)a,b

Unstable

0.92 (0.92–0.92)

0.88 (0.88–0.88)a,c

0.92 (0.92–0.93)b

Deceased

0.94 (0.94–0.95)b,c

0.95 (0.94–0.95)a,c

0.99 (0.96–1.00)a,b

Transition

   

Unstable–stable

0.77 (0.76–0.77)b

0.70 (0.70–0.70)a

0.73 (0.69–0.77)

Stable–unstable

0.77 (0.76–0.77)c

0.77 (0.77–0.77)c

0.71 (0.67–0.75)a,b

Life-sustaining therapy

   

MV

0.80 (0.79–0.81)c

0.81 (0.80–0.81)c

0.73 (0.69–0.77)a,b

VP

0.81 (0.80–0.82)b

0.76 (0.76–0.77)a,c

0.83 (0.78–0.87)b

CRRT

0.94 (0.93–0.96)b

0.86 (0.85–0.87)a

0.93 (0.79–0.98)

  1. CRRT continuous renal replacement therapy, MV mechanical ventilation, VP vasopressors. Performance is the median AUROC across a 100-iteration bootstrap with replacement, with 95% confidence intervals in parenthesis.
  2. P values are based on pairwise two-sided Wilcoxon rank sum tests.
  3. ap value < 0.05 compared to the development cohort.
  4. bp value < 0.05 compared to external cohort.
  5. cp value < 0.05 compared to prospective cohort.