Cole, et al. found that logical discrepancies in a manuscript are associated with unreliable studies. So unreliable that they had to be retracted. In fact discrepancy counts are 2.7-fold higher in retracted reports than in unretracted reports, according to this blinded case-controlled study. However, the prevalence of retraction is extremely low, so the probability that a study with lots of discrepancies warrants retraction is also very low – remember posterior odds = prior odds x likelihood ratio. In short it is a lousy diagnostic (or even screening) test. This study reminded the CLAHRC WM Director of a study conducted in our precursor CLAHRC pilot  where it was found that doctors who make minor errors are more likely to make serious errors but where, again, statistical significance did not entail clinical significance.
— Richard Lilford, CLAHRC WM Director
- Cole GD, Nowbar AN, Mielewczik M, et al. Frequency of discrepancies in retracted clinical trial reports versus unretracted reports: blinded case-control study. BMJ. 2015; 351: h4708.
- Coleman JJ, Hemming K, Nightingale PG, et al. Can an electronic prescribing system detect doctors who are more likely to make a serious prescribing error? J R Soc Med. 2011; 104(5): 208-18.