First of all, you are apparently referring to a summary statement that does not contain methodology or detailed description of statistical methodologies. The actual publication did account for changes in volume and still reported an increase in accidents and injuries.
The traffictek document shows an increase in crashes, but a greater increase in ridership
First, a general comparison group is used to
account for crash trends. Second, changes in traffic volumes are taken into account. And
third, an analysis of long-term crash trends is made in order to check for abnormally high or
low crash counts, i.e. regression-to-the-mean, in the before period.
And the stunning conclusion:
If corrections for traffic volumes were not done at all, the expected number of crashes
and injuries in the after period on the roads, where bicycle tracks were constructed, would be
2,758 and 875, respectively. The comparable figures found when corrections for traffic
volumes were done, see Table 3, are 2-4 percent lower. This means that corrections for traffic
volumes result in a small worsening of the overall safety effect, i.e. the effect would be about
6 percent instead of about 10 percent as shown in Table 3.
This should not be surprising since this observation is a recurring theme in all large studies of physically separated infrastructure. As noted by Heine it also matches results from additional Danish studies as well as German Cycling Federation studies of German and Dutch data.
The weighted means or best estimates for safety effects of bicycle tracks in urban
areas are an increase of about 10 percent in crashes and injuries. This is due to a large
increase of 18 percent in intersections, which more than outweigh a small reduction on road
links between intersections. Pedestrians, bicyclists and moped riders safety at intersections
are significantly worsened.