Testing conditional independence in diagnostic palaeoepidemiology.

Leprosy was a well-known and dreaded disease in the Middle Ages. A substantial fraction of the adult population carried leprosy-related lesions. Previous research analyzed the occurrence and implications of seven such lesions in samples of medieval skeletons. These analyses were carried out under the assumption of conditional independence among lesion scores. The present paper examines this assumption by developing a test based on the odds ratios and applying the test to three rural medieval samples from Europe: Tirup from the 12th-14th century AD in Jutland, Denmark; Refshale from the 12th-14th century AD on the island of Lolland, Denmark; and Lauchheim from AD 460-680 in southern Germany. Signs of nonzero prevalence of leprosy at death were found in all three samples: Tirup, 25.5% (95% CI, 17.2-34.6%); Refshale, 39.1% (95% CI, 25.5-54.7%); and Lauchheim, 16.2% (95% CI, 10.0-22.9%). It is shown that when leprosy is the prime factor creating variation in the lesion scores in and between samples, the assumption of conditional independence cannot be rejected.