Convicted by Statistics?

January 4, 2008 7:57 AM

Dutch nurse Lucia De Berk has had her case reopened 5 years after her conviction for multiple counts of murdering her patients.

Ben Goldacre's much linked Bad Science has an excellent summary and collection of links reviewing the history of the case, suggesting that the conviction was based upon poor statistics and weak circumstantial evidence.
"In the three years before Lucia worked on the ward in question, there were 7 deaths. In the three years that Lucia did work on that ward, there were 6 deaths. It seems odd that the death rate should go down on a ward at the precise moment that a serial killer “on a killing spree“ arrives on the scene."
Nature notes a parallel with the unfortunate case of Sally Clark, convicted largely by the erroneous statistical analysis of the now-discredited
Sir Roy Meadow previously.

posted by Jakey (6 comments total)

Great post.
posted by goodnewsfortheinsane at 8:29 AM on January 4, 2008

It must be noted that one of the alleged victims was Mr. Haopei Li, a (91-year old) Chinese judge at the International Court for War Crimes in Former Yugoslavia. That can't have helped Ms. de Berk.

More to the point, the whole case always had a whiff of the witchhunt about it. The privatised Dutch healthcare system is in a declining state, and it must have been tempting to blame an individual for an increase in mortality that could rather be caused by continuous cost-cutting...
posted by Skeptic at 8:56 AM on January 4, 2008

This is the identical statistical procedure used on athletes who fail drug tests (or two drug tests at once, etc.) Makes you wonder...
posted by noble_rot at 9:39 AM on January 4, 2008

From Wikipedia:

"...many experts still did not exclude a natural cause of death. In the case where digoxin poisoning was alleged, and supposedly detected by independent measurements in two Dutch laboratories, the (same) method used in those laboratories did not exclude that the substance found was actually a related substance naturally produced by the body itself (i.e., this could be a false positive, and the two measurements are not better than one)."

If the test is faulty or the substance it detected is naturally produced, two independent tests are not better.
posted by jeffamaphone at 10:05 AM on January 4, 2008

Three statisticians go out hunting. They spot a nice juicy deer, and the first statistician aims, shoots... and misses, a yard wide on the left. The second statistician aims, shoots... and misses, a yard wide on the right. Upon which the third one drops his rifle and jumps up and down triumphantly, shouting "we got him! we got him!".
posted by goodnewsfortheinsane at 10:23 AM on January 4, 2008

Wait, what? 6 deaths is considered the work of a serial killer? In my first 6 months after graduating I hit the double digits! Course, that was my own fault for continually volunteering to look after the heavy (and therefore, dying) patients a lot, but still....
posted by Silentgoldfish at 4:47 AM on January 5, 2008