Table 2.
All days |
Unstable days |
|||||
---|---|---|---|---|---|---|
Test set | ||||||
Nephrocast | Reference SCr | P-value | Nephrocast | Reference SCr | P-value | |
MAE, median (IQR), mg/dL | 0.09 (0.085, 0.09) | 0.13 (0.131, 0.135) | <.01 | 0.20 (0.197, 0.203) | 0.54 (0.532, 0.548) | <.01 |
RMSE, median (IQR), mg/dL | 0.15 (0.146,0.152) | 0.25 (0.245, 0.253) | <.01 | 0.31 (0.307, 0.327) | 0.67 (0.663, 0.694) | <.01 |
Temporal validation | ||||||
MAE, mg/dL | 0.08 | 0.13 | N/A | 0.19 | 0.54 | N/A |
RMSE, mg/dL | 0.14 | 0.25 | N/A | 0.31 | 0.66 | N/A |
Prospective validation | ||||||
MAE, mg/dL | 0.09 | 0.13 | N/A | 0.18 | 0.50 | N/A |
RMSE, mg/dL | 0.14 | 0.23 | N/A | 0.28 | 0.60 | N/A |
Reference SCr was defined as the measured SCr on the day of making the prediction. Predicted SCr was defined as the predicted SCr by our model for the next patient-day. Measured SCr was defined as the laboratory measured SCr on the next patient-day. Model error was defined as the difference between predicted SCr and measured SCr, which was compared against the difference between reference SCr and measured SCr. Error was summarized using MAE and RMSE. Unstable patient-days were defined as those in which there was a change of 30% or 0.3 mg/dL or more in SCr concentration between the reference SCr and the measured SCr on the following patient-day.
Abbreviations: IQR = interquartile range; MAE = mean absolute error; RMSE = root mean squared error; SCr = serum creatinine.