Biology Reference
In-Depth Information
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-150
-100
-50
0
Time Lag in Week
50
100
150
Figure 1.4
Cross correlation between “respiratory” (RS) and the “pneumonia and influenza” (P and I)
deaths: maximum correlation occurs when RS and “influenza chief complaints” (IS) are
2 weeks earlier than the P and I curve. (From Tsui et al. 2002. Value of ICD-9-coded chief com-
plaints for detection of epidemics. J Am Med Inform Assoc 9: S41-S47.)
Overall, ambulatory-care episodes were highly correlated (correlation = 0.92)
with hospital admissions, and they preceded them by 2 weeks.
Tsui et al. (2002) measured the value of ICD-9 coded emergency depart-
ment chief complaints for influenza surveillance. They applied a detection
system based on a standard algorithm (the Serfling method) and the cross-
correlation function. The timeliness of “respiratory” (RS) and “influenza”
(IS) chief complaint ICD-9 codes were compared with national pneumonia
and influenza (P and I) deaths. The time to detection for RS and IS data based
on the algorithmic method was one week earlier than P and I deaths. In this
study, the cross-correlation function demonstrated that the signal provided
by ICD-9-coded chief complaints occurred 2 weeks earlier than the signal
provided by P and I data (see Figure 1.4) (Tsui et al. 2002).
Benefits of the correlation method:
This method is useful for showing that variation in one data source
can predict changes in another data source (Suyama et al. 2003).
Limitations of the correlation method:
This method requires data collected over a continuous period of
time (i.e., cannot have missing data, e.g., weekend data).
Search WWH ::




Custom Search