Multicenter validation of PIM3 and PIM2 in Brazilian pediatric intensive care units

Figures

• 

  Figure 1. Study flowchart, showing eligibility, exclusion criteria, and final population included in the study.

• 

  Figure 2. Receiver operating characteristic (ROC) curves for PIM3 and PIM2, and areas under the curve (AUROC). PIM2: 0.89 (95% CI 0.88-0.90); PIM3: 0.88 (95% CI 0.87-0.89) (P = 0.0018).

• 

  Figure 3. Calibration belts for PIM2 (A) and PIM3 (B), assessing the concordance of observed vs. expected outcome in 10 deciles of patient risk. The dashed line represents the mean line compared to the bisector, which indicates a perfect match between the PIM results and the outcomes it tries to predict. The p-value expresses a Wald-like statistic that tests the null hypothesis that there is no difference between this line and the bisector, which was rejected for PIM2 and PIM3. The inner light gray area marks the 80% CI boundary, and the dark gray outer belt marks the 95% CI boundary. For PIM2, the calibration belt is never over the bisector and is under the bisector (overestimating mortality) in a very low risk range below 2% mortality risk and over 70% predicted mortality, indicating that PIM2 overestimated the mortality for these high-risk patients. For PIM3, the calibration belt is over the bisector between 0.09 and 0.46 predicted mortality, underestimating the mortality in this risk range, and it is under the bisector (overestimating mortality) in a small range in less than 1% and above 99% mortality risk.

• 

  Figure 4. Receiver operating characteristic (ROC) curves for PIM3 and PIM2, and areas under the curve (AUROC) in the three periods from October 2013 to September 2019. (A) Period 1(Oct-2013 to Sep-2015): PIM2: 0.89 (95% CI 0.86-0.91); PIM3: 0.86 (95% CI 0.84-0.89) (P=0.0018); (B) Period 2 (Oct-2015 to Sep-2017): PIM2: 0.89 (95% CI 0.86-0.91); PIM3: 0.87 (95% CI 0.85-0.90) (P=0.0136); (C) Period 3 (Oct-2017 to Sep-2019): PIM2: 0.90 (95% CI 0.88-0.92); PIM3: 0.90 (95% CI 0.88-0.92) (P=0.652).

• 

  Figure 5. Calibration belts for PIM2 and PIM3 in the three periods from October 2013 to September 2019. PIM2 maintained a good calibration for most risk ranges in all three periods, never underestimating mortality, overestimating it in the second period in the range of more than 91% risk of death and in the third period in the range of 0 to 5% risk of mortality. PIM3, on the other hand, had excellent calibration in the first two years, underestimated mortality in the risk-of-death range of 8%–44% in the second period, and overestimated it in a small range of 0%–3% risk mortality in the last two years.