Skip to main content

Table 1 Performance of different detection algorithms with sensitivity of 100% to detect influenza (UrgIndex-hospitalisation and ICD10-consultations)

From: Can epidemic detection systems at the hospital level complement regional surveillance networks: Case study with the influenza epidemic?

Time series Detection algorithms Outbreaks periods Non-epidemic periods
   Mean timeliness (days) Number of days with false alarms N = 1094 Specificity (%)
UrgIndex - hospitalisations C2, k = 0.08, 1d −7.5 327 70.1
  C2, k = 0.08, 3d −3.3 226 79.3
  C2, k = 0.1, 1d −15.3 247 77.4
  C2, k = 0.1, 3d 5.0 155 85.8
  C3, k = 0.08, 1d −58.3 817 25.3
  C3, k = 0.08, 3d 20.3 739 32.4
  C3, k = 0.08, 5d −13.3 589 46.2
  C3, k = 0.1, 1d 18.3 745 31.9
  C3, k = 0.1, 3d −18.3 658 39.9
  C3, k = 0.1, 5d −13.3 589 46.2
  C3, k = 0.5, 1d 0.5 230 79.0
  C3, k = 1, 1d 5.5 78 92.9
ICD10 – consultations C1, k = 0.07, 1d −18.3 151 86.2
  C1, k = 0.07, 3d −11.5 127 88.6
  C1, k = 0.07, 5d −7.8 115 89.5
  C1, k = 0.1, 1d −18.3 148 86.5
  C1, k = 0.1, 3d −7.8 122 88.4
  C2, k = 0.07, 1d −19.8 147 86.6
  C2, k = 0.07, 3d 0.5 135 87.7
  C2, k = 0.07, 5d 2.5 125 88.6
  C2, k = 0.1, 1d −13.5 143 86.9
  C2, k = 0.1, 3d 0.8 131 88.0
  C2, k = 0.1, 5d 2.8 121 88.9
  C3, k = 0.07, 1d −32.3 172 84.3
  C3, k = 0.1, 1d −32.3 168 84.6
  C3, k = 0.1, 3d −15.8 151 86.2
  C3, k = 0.1, 5d −11.3 154 87.0
  C3, k = 0.5, 1d 9.8 74 93.2
  1. C1, C2, and C3 refer to the three different moving average calculations of CUSUM statistics (C1-mild, C2-medium, C3-ultra).
  2. k is the detectable difference to the mean used to the calculation of CUSUM statistics.
  3. Negative mean timeliness: first day signal before the outbreaks beginning, on average.
  4. Positive mean timeliness: first day signal after the outbreaks beginning, on average.