Background
Related work
Methods
Dataset
Variable | Mean | SD |
---|---|---|
Gender- N, % | ||
Female | 9755 | 40.73% |
Male | 14,195 | 59.27% |
Age (yr) | 60.89 | 16.09 |
Ethnicity | ||
African-American | 1668 | 6.96% |
White | 17,261 | 72.07% |
Hispanic | 845 | 3.53% |
Other | 4176 | 17.44% |
Creatinine Max during Day 2 and Day 3 (mg/dL) | 0.84 | 0.31 |
Heart Rate Maximum (bpm) | 105.40 | 19.86 |
Heart Rate Mean (bpm) | 86.97 | 15.15 |
Systolic BP Minimum (mmHg) | 92.51 | 16.79 |
Systolic BP Mean (mmHg) | 118.87 | 15.71 |
Diastolic BP Minimum (mmHg) | 45.31 | 11.06 |
Diastolic BP Mean (mmHg) | 61.98 | 10.24 |
Temperature Maximum (Celsius) | 37.60 | 0.75 |
SpO2 Minimum (%) | 92.23 | 6.71 |
SpO2 Mean (%) | 97.41 | 1.85 |
Glucose Level Maximum (mg/dL) | 173.36 | 74.94 |
Bicarbonate Level Minimum (mg/dL) | 23.87 | 4.34 |
Creatinine Level Minimum (mg/dL) | 0.74 | 0.21 |
Creatinine Level Maximum (mg/dL) | 0.80 | 0.22 |
Hemoglobin Level Minimum (g/dL) | 10.33 | 2.08 |
Platelet Count Minimum (K/μL) | 210.24 | 114.82 |
Potassium Level Maximum (mg/dL) | 4.39 | 0.79 |
Partial Thromboplastin Time Minimum (s) | 32.55 | 11.72 |
Partial Thromboplastin Time Maximum (s) | 40.22 | 23.55 |
International Normalized Ratio Minimum | 1.34 | 0.46 |
International Normalized Ratio Maximum | 1.48 | 0.70 |
Prothrombin Time Minimum (s) | 14.66 | 3.61 |
Prothrombin Time Maximum (s) | 15.63 | 4.95 |
Blood Urea Nitrogen Level Maximum (mg/dL) | 16.87 | 8.58 |
White Blood Cell Count Maximum (K/μL) | 12.78 | 9.02 |
Calcium Level Minimum (mg/dL) | 8.08 | 0.76 |
Mechanical Ventilation- N, % | ||
No (0) | 11,677 | 48.76% |
Yes (1) | 12,273 | 51.24% |
Average Urine Output (mL) | 2202.68 | 1395.37 |
Estimated Glomerular Filtration Rate (eGFR) | 110.20 | 52.49 |
Statistical analysis
class_weight
parameter to “balanced” for logistic regression and random forest classifiers to down weight the more popular class. We also explored the imlearn package for multiple over sampling and under sampling methods including SMOTE [34], ADASYN [35], random under sampling, Tomek Link, and Neighborhood Cleaning Rule. For our task, we found that random under sampling to bring the positive and negative class numbers to comparable level is a simple yet effective method.Results
Variable | Linear Regression | Logistic Regression | ||||
---|---|---|---|---|---|---|
Coeff | P-Value | 95% CI | OR | P-Value | 95% CI | |
Gender | ||||||
(M: 1,F: 0) | 0.2135 | < 0.001 | [0.206,0.221] | 1.23 | < 0.001 | [1.144,1.318] |
Age (yr) | 0.0019 | < 0.001 | [0.002,0.002] | 1.02 | < 0.001 | [1.016,1.020] |
Ethnicity | ||||||
African-American | ref | ref | ref | ref | ref | ref |
Caucasian | −0.0706 | < 0.001 | [−0.086,-0.055] | 1.12 | 0.106 | [0.976,1.294] |
Hispanic | −0.1026 | < 0.001 | [− 0.128,-0.077] | 0.73 | 0.014 | [0.566,0.938] |
Other | −0.0585 | < 0.001 | [− 0.076,-0.041] | 1.33 | < 0.001 | [1.136,1.552] |
Systolic BP Min (mmHg) | −0.0011 | < 0.001 | [−0.001,-0.001] | 0.98 | < 0.001 | [0.976,0.980] |
Systolic BP Mean (mmHg) | −0.0005 | < 0.001 | [−0.001,0.000] | 0.98 | < 0.001 | [0.982,0.987] |
Diastolic BP Min (mmHg) | −0.0003 | 0.067 | [− 0.001,0.000] | 0.98 | < 0.001 | [0.980,0.986] |
Diastolic BP Mean (mmHg) | −0.0009 | < 0.001 | [−0.001,0.001] | 0.98 | < 0.001 | [0.971,0.978] |
Temperature Max (Celsius) | −0.0035 | 0.190 | [−0.009,0.002] | 1.04 | 0.089 | [0.994,1.089] |
SpO2 Min (%) | −0.0014 | < 0.001 | [−0.002,-0.001] | 0.99 | < 0.001 | [0.983,0.991] |
SpO2 Mean (%) | −0.0028 | 0.010 | [−0.005,-0.001] | 1.08 | < 0.001 | [1.055,1.098] |
Bicarbonate Level Min (mg/dL) | −0.0049 | < 0.001 | [−0.006,-0.004] | 0.94 | < 0.001 | [0.934,0.949] |
Creatinine Level Minimum (mg/dL) | 0.9479 | < 0.001 | [0.933,0.963] | – | – | – |
Creatinine Level Max (mg/dL) | 0.9340 | < 0.001 | [0.921,0.947] | 3.20 | < 0.001 | [2.743, 3.737] |
Hemoglobin Level Min (g/dL) | −0.0048 | < 0.001 | [− 0.007,-0.003] | 0.77 | < 0.001 | [0.759,0.787] |
Platelet Count Min (K/μL) | − 0.0004 | < 0.001 | [− 0.000,0.000] | 1.00 | < 0.001 | [0.996,0.997] |
Potassium Level Max (mg/dL) | 0.0837 | < 0.001 | [0.079,0.089] | 1.90 | < 0.001 | [1.824,1.984] |
Partial Thromboplastin Time Min (s) | 0.0020 | < 0.001 | [0.002,0.002] | 1.01 | < 0.001 | [1.010,1.015] |
Partial Thromboplastin Time Max (s) | 0.0012 | < 0.001 | [0.001,0.001] | 1.01 | < 0.001 | [1.008,1.010] |
International Normalized Ratio Min | 0.0213 | < 0.001 | [0.013,0.030] | 1.04 | 0.266 | [0.970,1.118] |
International Normalized Ratio Max | 0.0274 | < 0.001 | [0.022,0.033] | 1.21 | < 0.001 | [1.158,1.260] |
Prothrombin Time Min (s) | 0.0032 | < 0.001 | [0.002,0.004] | 1.01 | 0.194 | [0.997,1.015] |
Prothrombin Time Max (s) | 0.0042 | < 0.001 | [0.003,0.005] | 1.03 | < 0.001 | [1.021,1.033] |
Blood Urea Nitrogen Level Max (mg/dL) | 0.0088 | < 0.001 | [0.008,0.009] | 1.01 | < 0.001 | [1.009,1.017] |
White Blood Cell Count Max (K/μL) | 0.0008 | < 0.001 | [0.000,0.001] | 1.01 | < 0.001 | [1.010,1.018] |
Calcium Level Min (mg/dL) | 0.0245 | < 0.001 | [0.019,0.030] | 0.72 | < 0.001 | [0.685,0.749] |
Mechanical Ventilation (Yes 1/No 0) - N, % | 0.0618 | < 0.001 | [0.054,0.070] | 3.04 | < 0.001 | [2.816,3.275] |
Estimated Glomerular Filtration Rate (eGFR) | −0.0025 | < 0.001 | [−0.003,-0.002] | 1.00 | < 0.001 | [1.004,1.005] |
Variable | Adjusted Regression Coefficient | P-Value | 95% Confidence Interval |
---|---|---|---|
Gender | |||
(M: 1,F: 0) | 0.026 | < 0.001 | [0.019,0.034] |
Age | 0.020 | < 0.001 | [0.016,0.023] |
SpO2 Mean | −0.013 | < 0.001 | [− 0.016,-0.009] |
Bicarbonate Level Minimum | −0.005 | 0.005 | [−0.008,-0.001] |
Creatinine Level Minimum | 0.364 | < 0.001 | [0.326,0.403] |
Creatinine Level Maximum | 0.137 | < 0.001 | [0.130,0.144] |
Hemoglobin Level Minimum | −0.024 | < 0.001 | [− 0.027,-0.021] |
Platelet Count Minimum | −0.018 | < 0.001 | [− 0.021,-0.015] |
Potassium Level Maximum | 0.021 | < 0.001 | [0.018,0.025] |
Partial Thromboplastin Time Minimum | 0.016 | < 0.001 | [0.013,0.019] |
International Normalized Ratio Minimum | −0.025 | < 0.001 | [− 0.035,-0.015] |
International Normalized Ratio Maximum | 0.016 | < 0.001 | [0.011,0.020] |
Blood Urea Nitrogen Level Maximum | −0.014 | < 0.001 | [−0.017,-0.011] |
Calcium Level Minimum | 0.014 | < 0.001 | [0.011,0.017] |
Mechanical Ventilation | 0.018 | < 0.001 | [0.015,0.021] |
Estimated Glomerular Filtration Rate | 0.009 | < 0.001 | [0.004,0.013] |
Variable | Adjusted Odds Ratio | P-Value | 95% Confidence Interval |
---|---|---|---|
Gender | |||
(M: 1,F: 0) | 0.59 | < 0.001 | [0.553,0.647] |
Age | 1.56 | < 0.001 | [1.494,1.635] |
Ethnicity | |||
African-American | ref | ref | ref |
Caucasian | 1.47 | < 0.001 | [1.248,1.721] |
Hispanic | 1.12 | 0.414 | [0.850,1.484] |
Other | 1.65 | < 0.001 | [1.382,1.970] |
Systolic BP Minimum | 0.94 | 0.003 | [0.897,0.977] |
SpO2 Mean | 0.89 | < 0.001 | [0.852,0.925] |
Bicarbonate Level Minimum | 0.90 | < 0.001 | [0.858,0.938] |
Creatinine Level Maximum | 2.68 | < 0.001 | [2.501,2.862] |
Hemoglobin Level Minimum | 0.74 | < 0.001 | [0.711, 0.776] |
Platelet Count Minimum | 0.80 | < 0.001 | [0.767,0.839] |
Potassium Level Maximum | 1.28 | < 0.001 | [1.236,1.334] |
Partial Thromboplastin Time Minimum | 1.12 | < 0.001 | [1.078,1.158] |
International Normalized Ratio Maximum | 1.06 | < 0.001 | [1.025,1.099] |
Blood Urea Nitrogen Level Maximum | 0.89 | < 0.001 | [0.857,0.933] |
Calcium Level Minimum | 1.08 | < 0.001 | [1.034,1.130] |
Mechanical Ventilation | 1.45 | < 0.001 | [1.386,1.519] |
Estimated Glomerular Filtration Rate | 2.93 | < 0.001 | [2.734,3.140] |
Root Mean Square Error Values | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Mean | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Run 6 | Run 7 | Run 8 | Run 9 | Run 10 | |
Linear Backward Selection Model | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 | 0.224 |
Linear All Variables Model | 0.224 | 0.224 | 0.224 | 0.223 | 0.224 | 0.224 | 0.224 | 0.223 | 0.224 | 0.224 | 0.224 |
Area Under the Curve Values | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Mean | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Run 6 | Run 7 | Run 8 | Run 9 | Run 10 | |
LR Backward Selection Model | 0.780 | 0.780 | 0.780 | 0.779 | 0.780 | 0.780 | 0.780 | 0.780 | 0.780 | 0.780 | 0.780 |
LR All Variables Model | 0.783 | 0.782 | 0.782 | 0.782 | 0.782 | 0.782 | 0.782 | 0.782 | 0.782 | 0.782 | 0.783 |
RF Backward Selection Model | 0.772 | 0.772 | 0.774 | 0.775 | 0.772 | 0.773 | 0.773 | 0.773 | 0.775 | 0.773 | 0.772 |
RF All Variables Model | 0.779 | 0.779 | 0.779 | 0.780 | 0.779 | 0.780 | 0.778 | 0.777 | 0.780 | 0.779 | 0.779 |
MLP Backward Selection Model | 0.792 | 0.794 | 0.794 | 0.791 | 0.792 | 0.793 | 0.794 | 0.792 | 0.793 | 0.794 | 0.792 |
MLP All Variables Model | 0.796 | 0.796 | 0.798 | 0.796 | 0.795 | 0.796 | 0.796 | 0.794 | 0.795 | 0.796 | 0.796 |
Accuracy | Sensitivity | Specificity | Positive Predictive Value | Negative Predictive Value | |
---|---|---|---|---|---|
LR Backward Selection Model | 0.724 | 0.697 | 0.730 | 0.337 | 0.924 |
LR All Variables Model | 0.729 | 0.698 | 0.736 | 0.342 | 0.925 |
RF Backward Selection Model | 0.739 | 0.660 | 0.754 | 0.346 | 0.918 |
RF All Variables Model | 0.742 | 0.673 | 0.756 | 0.352 | 0.921 |
MLP Backward Selection Model | 0.744 | 0.684 | 0.756 | 0.356 | 0.924 |
MLP All Variables Model | 0.743 | 0.694 | 0.753 | 0.357 | 0.926 |