Skip directly to search Skip directly to A to Z list Skip directly to page options Skip directly to site content

Volume 14, Number 7—July 2008

Dispatch

Optimizing Use of Multistream Influenza Sentinel Surveillance Data

Eric H. Y. Lau*, Benjamin J. Cowling*Comments to Author , Lai-Ming Ho*, and Gabriel M. Leung*
Author affiliations: *University of Hong Kong, Hong Kong Special Administrative Region, People’s Republic of China;

Main Article

Table

Performance of alerts generated by individual monitoring of aggregate data and separate data streams, and simultaneous monitoring of multiple data streams by using univariate and multivariate time series models, Hong Kong, 1998–2007*

Data Univariate models
Multivariate models†
AUWROC Sensitivity‡ Timeliness, wk‡ AUWROC Sensitivity‡ Timeliness, wk‡
Aggregated data
GP 0.78 1.00 2.41
GOPC
0.86
1.00
1.50




Single stream
GP
HK 0.75 1.00 2.36 0.73 0.87 2.64
KL 0.66 1.00 2.71 0.62 0.88 3.06
NTE 0.89 1.00 2.00 0.76 0.90 2.04
NTW 0.80 1.00 2.07 0.80 0.91 2.24
GOPC
HK 0.79 1.00 2.21 0.71 0.89 2.42
KL 0.78 1.00 2.46 0.62 0.96 3.15
NTE 0.79 0.95 2.22 0.79 0.96 2.26
NTW
0.73
1.00
2.55

0.72
1.00
2.52
Multiple streams
M1: First aberration 0.84 1.00 1.57 0.86 1.00 1.66
M2: 2 simultaneous aberrations 0.89 1.00 1.22 0.82 1.00 1.77
M3: 3 simultaneous aberrations 0.90 1.00 1.47 0.80 1.00 1.70
M4: Any 2 aberrations in 2 wk 0.81 1.00 2.63 0.72 1.00 2.43
M5: Any 2 aberrations in 2 wk 0.83 1.00 2.44 0.77 1.00 2.11

*AUWROC, area under the weighted receiver operating characteristic curve; GP, general practitioner; GOPC, general outpatient clinic; HK, Hong Kong Island; KL, Kowloon; NTE, New Territories East; NTW, New Territories West.
†See Technical Appendix for more detailed description of the multivariate model.
‡At a fixed specificity of 0.95.

Main Article

TOP