We should stick to the original “to err” methodology - with improvements for specificity, in which case we’ve reduced error 2-5x since 1980-90s when that data was first collected. Avoidability is also a critical factor. This methodology puts figure at 3%.