How Good are Modern Vital Statistics?

From the book Drug War Heresies:

It has been said, for example, that French authorities will only record a death as drug-related if a needle is still sticking in the arm, whereas the Germans will include a driving fatality of a one-time client of a drug treatment clinic.

 Indeed, if you look at the figures in this link (click on "EMCDDA Selection B"), Germany inexplicably has more overdose deaths than France. In 2013, Germany had 1179 while France had 349. Germany had 80.65 million people and France had 66 million in 2013. Not enough to explain the difference. Dividing by populations, we get 1.46 deaths per 100k in Germany and 0.53 deaths per 100k in France. This is for two western European that don't differ in any obvious way (no obvious way that would explain a factor of three difference in drug poisoning deaths, anyway). If the reporting bias mentioned in Heresies is real, though, that would certainly explain a difference in the official statistics.

This is a theme I've written about numerous times. Too many commentators take the official statistics at face value. If mere reporting differences leads to enormous differences in national statistics, then it's likely that changes in reporting biases over time can induce a spurious trend in a time series. It's been a while since I looked at the by-state drug overdose data in any detail, but some patterns in those data suggested a similar bias creeping in. There is enormous state-to-state variation in overdose totals and in the trends over time. Are these differences real or spurious?