Statistical method
In this study a case-crossover (CC) design was applied.
The technique is an adaptation of the case-control methodology
[14]. In the CC approach, a case defines its own
controls, i.e., the case itself considered on a predefined set
of control days proximate to the time of the occurrence
of the case. Usually, the control days are selected to be
the same days of the week in a month of the case occurrence.
In such an approach 3 or 4 controls are produced
for each case. For example the case on January 8, 2002
has the following 4 corresponding control days: January
1, 15, 22, and 29 in the year 2002. The method has been
widely adopted, as it has been shown to produce unbiased
conditional logistic regression estimates [15,16].
A p