Skip to main content
. Author manuscript; available in PMC: 2015 Jun 1.
Published in final edited form as: J Biomed Inform. 2008 Mar 28;41(3):452–460. doi: 10.1016/j.jbi.2008.03.011

Table 2.

Performance Comparison for Classification Tree Learning. CMS is the bedside monitor’s alarm algorithm. Threshold stands for a standard threshold-based alarm algorithm. The rightmost five columns are results for the classification models built after the five given times. PPV is positive predictive value. The measures shown are averages from 10 sessions of monitoring, with the ranges of values shown in brackets.

Metric CMS Threshold Tree built after n hours of data
12
 1 2 4 8
Sensitivity 1.00 1.00 0.00 [.00,.18] 0.01 [.00,.30] 0.11 [.04,.36] 0.43 [.29,.63] 0.84 [.70,.93]
Specificity 0.99 0.88 1.00 [.97,1.0] 0.98 [.69,.99] 0.98 [.90,.98] 0.98 [.94,.98] 0.98 [.96,.99]
PPV 0.82 0.70 0.63 [0.0,.72] 0.17 [0.0,.20] 0.14 [.02,.23] 0.37 [.15,.57] 0.72 [.60,.80]
Accuracy 0.99 0.96 0.96 [.92,.96] 0.95 [.89,.96] 0.95 [.93,.98] 0.97 [.95,.99] 0.97 [.96,.98]