Table 3

Descriptive statistics for summary performance characteristics for the EARS and NBC algorithms: receiver operating characteristic area under the curve (AUC) analyses

Algorithm

median (mean) AUC0–1

median (mean) AUC0–0.01

median (mean) AUC0–0.005

median (mean) WtAUC0–1

median (mean) WtAUC0–0.01


EARS C1

0.995 (0.989)

0.0011 (0.0019)

0.0004 (0.0015)

0.959 (0.968)

<0.0001 (0.0008)

EARS C2

0.996 (0.990)

0.0017 (0.0026)

0.0002 (0.0016)

0.994 (0.978)

0.0002 (0.0013)

EARS C3

0.996 (0.993)

0.0061 (0.0059)

0.0011 (0.0022)

0.991 (0.980)

0.0011 (0.0033)

NBC 7-day

1.0 (0.998)

0.0095 (0.0081)

0.0049 (0.0040)

0.990 (0.979)

<0.0001 (0.0032)

NBC 14-day

1.0 (0.999)

0.0099 (0.0088)

0.0049 (0.0040)

0.988 (0.981)

<0.0001 (0.0026)

NBC 28-day

1.0 (1.0)

0.0095 (0.0092)

0.0049 (0.0047)

0.990 (0.983)

0.0014 (0.0032)

NBC 56-day

1.0 (0.999)

0.0095 (0.0088)

0.0042 (0.0036)

0.977 (0.928)

<0.0001 (0.0018)

p-value¥

<0.0001

<0.0001

<0.0001

0.046

0.006


Evaluated at a false alarm rate ≤ 0.01

Evaluated at a false alarm rate ≤ 0.005

¥Friedman rank sum test

WtAUC: Weighted AUC using a reference value of 7 days for valid detection

EARS: Early Aberration Reporting System

NBC: Negative binomial cusum with an out of control state defined as 2 standard deviations greater than the mean

Watkins et al. BMC Medical Informatics and Decision Making 2008 8:37   doi:10.1186/1472-6947-8-37

Open Data