Fig 1.
Multiple dataset creation and exclusion criteria from each patient.
The dataset was created during entry and exit from the ICU/ECU while heparin was administered and dosing was continued. The dataset of interest included data in which the aPTT was measured prior to heparin administration and the target aPTT was measured between 6 and 24 h post-administration.
Fig 2.
Data processing and model construction procedure.
Table 1.
Patient characteristics.
Fig 3.
Histogram of aPTT values after the (A) first dose and (B) multiple doses of heparin.
Table 2.
Predictive performance of aPTT for first heparin infusion.
Fig 4.
Contributing variables for predictive performance in (A) first and (B) multiple heparin infusion patients.
The contributions to the prediction model were calculated as follows. For the first and multiple heparin infusion patient prediction, the random forest model with data imputed based on static data was used. aPTT: activated partial thromboplastin time; BHC after SA: blood heparin concentration after the starting point of administration; AT-III: antithrombin III; Urine volume: amount of urine output; CBV: circulation blood volume; PT-INR: prothrombin time-international normalized ratio; SBP: systolic blood pressure; AST: aspartate aminotransferase; ALT: alanine aminotransferase; Elapsed time: elapsed time from baseline aPTT measurement to target aPTT measurement; BHC before SA: blood heparin concentration before the starting point of administration; ECMO: extracorporeal membrane oxygenation; TFV: total fluid volume.
Table 3.
Predictive performance of aPTT for multiple heparin infusions.