(I addressed per-captia deaths vs. distance driven in another comment earlier, I do quite agree)

I think averaging the rates across the EU is a counterproductive, or outright obfuscating, metric. While it is a basis for comparison, the data varies so wildly from country to country it serves only to obscure the realities of the situation. The report I’m sourcing from does make it clear that countries may be underreporting their accident rates:

The comparability of data is also subject to differences in road accident reporting standards. Notably, the number of accidents and the number of injuries may be underreported in some countries due to administrative or practical limitations.

And the cited article from that section is quite interesting. In particular,

For estimations of the real volumes it is not necessary to have all accidents reported by the police. By using other sources, such as cause of death statistics and medical and insurance databases, it is possible to get a more accurate assessment. These additional databases can also provide supplementary information on accidents or the outcome of an accident, including the types and severity of injuries.

which does appear to be how CARE is sourcing their data (although their dashboard is some godawful Qlik thing so I was not able to determine sources for all of their metrics)