Interrater reliability of diagnostic methods in traditional Indian ayurvedic medicine

This study assesses the interrater reliability of Ayurvedic pulse (nadi), tongue (jivha), and body constitution (prakriti) assessments. Fifteen registered Ayurvedic doctors with 3-15 years of experience independently examined twenty healthy subjects. Subjects completed self-assessment questionnaires and software analyses for prakriti assessment. Weighted kappa statistics for all 105 pairs of doctors were computed for the pulse, tongue, and prakriti data sets. According to the Landis-Koch scale, the pairwise kappas ranged from poor to slight, slight to fair, and fair to moderate for pulse, tongue, and prakriti assessments, respectively. The average pairwise kappa for pulse, tongue, and prakriti was 0.07, 0.17, and 0.28, respectively. For each data set and pair of doctors, the null hypothesis of random rating was rejected for just twelve pairs of doctors for prakriti, one pair of doctors for pulse examination, and no pairs of doctors for tongue assessment. Thus, the results demonstrate a low level of reliability for all types of assessment made by doctors. There was significant evidence against random rating by software and questionnaire use and by the diagnosis preferred by the majority of doctors. Prakriti assessment appears reliable when questionnaire and software assessment are used, while other diagnostic methods have room for improvement.