Skip to main content

Table 4 Performance (AUC) of predicting unplanned readmissions following the unplanned discharges

From: A framework for feature extraction from hospital medical data with applications in risk prediction

  BASELINE (95% CI)   
  Period 1 M 3Y MR (95% CI) MR + Comorbidities (95% CI)
COPD      
1 M 0.57 (0.55,0.60) 0.60 (0.57,0.63) 0.730 (0.695,0.766) 0.730 (0.695,0.766)
2 M 0.59 (0.56,0.61) 0.60 (0.57,0.62) 0.719 (0.689,0.750) 0.719 (0.689,0.750)
3 M 0.58 (0.56,0.61) 0.60 (0.58,0.63) 0.719 (0.692,0.746) 0.720 (0.693,0.746)
6 M 0.59 (0.57,0.61) 0.61 (0.59,0.64) 0.724 (0.703,0.746) 0.724 (0.702,0.745)
12 M 0.60 (0.57,0.62) 0.62 (0.59,0.64) 0.720 (0.701,0.739) 0.720 (0.701,0.739)
Diabetes      
1 M 0.60 (0.57,0.62) 0.60 (0.58,0.63) 0.708 (0.674,0.741) 0.704 (0.670,0.738)
2 M 0.61 (0.59,0.63) 0.63 (0.61,0.65) 0.718 (0.692,0.744) 0.718 (0.692,0.743)
3 M 0.60 (0.58,0.622) 0.63 (0.61,0.65) 0.724 (0.703,0.745) 0.724 (0.703,0.745)
6 M 0.62 (0.60,0.633) 0.64 (0.62,0.66) 0.714 (0.697,0.731) 0.715 (0.698,0.732)
12 M 0.64 (0.62,0.653) 0.66 (0.64,0.68) 0.718 (0.705,0.732) 0.718 (0.704,0.732)
Mental disorders      
1 M 0.56 (0.53,0.59) 0.57 (0.54,0.60) 0.748 (0.709,0.787) 0.747 (0.708,0.786)
2 M 0.58 (0.55,0.61) 0.60 (0.57,0.62) 0.756 (0.727,0.784) 0.756 (0.728,0.785)
3 M 0.59 (0.57,0.62) 0.60 (0.58,0.63) 0.738 (0.713,0.764) 0.737 (0.711,0.762)
6 M 0.61 (0.59,0.64) 0.63 (0.61,0.65) 0.718 (0.697,0.740) 0.718 (0.696,0.739)
12 M 0.65 (0.63,0.67) 0.66 (0.64,0.68) 0.713 (0.694,0.732) 0.713 (0.694,0.732)
Pneumonia      
1 M 0.58 (0.55,0.60) 0.61 (0.59,0.63) 0.749 (0.717,0.782) 0.750 (0.718,0.782)
2 M 0.61 (0.59,0.63) 0.66 (0.64,0.68) 0.753 (0.729,0.777) 0.756 (0.733,0.780)
3 M 0.62 (0.60,0.64) 0.67 (0.65,0.68) 0.760 (0.739,0.780) 0.762 (0.742,0.782)
6 M 0.64 (0.62,0.66) 0.68 (0.67,0.70) 0.748 (0.731,0.764) 0.749 (0.733,0.765)
12 M 0.65 (0.63,0.67) 0.70 (0.68,0.71) 0.744 (0.731,0.758) 0.747 (0.733,0.761)
  1. AUC stands for Area Under ROC Curve; Feature sets are Elixhauser comorbidities as baselines, automatically extracted features from medical records (MR), and the combination of MR and comorbidities.