Assessment of X-ray readers in TB prevalence surveys.

A prevalence study was carried out a rural community in Tiruvallur district in Tami Nadu to standardize the method of assessing an X-ray reader in tuberculosis (TB) prevalence surveys by means of different measures of agreement between the reader and a Standard Reader (SR). The exercise on assessing the X-ray readers was carried out on two occasions; one involving three trainee readers (R1, R2, and R3), and the other involving one trainee reader (R4). The extent of agreement was estimated using Kappa statistics (K), over-diagnosis, under-diagnosis, crude agreement and prevalence adjusted bias adjusted kappa (PABAK). The overall performance of readers R1, R2, and R3 was not satisfactory in terms of K (21, 34 and 14%) in the first assessment. The K, over-diagnosis and under-diagnosis were estimated to be 61, 28 and 4% for R1, 63, 18 and 4% for R2 and 58, 31 and 5% for R3 in the final assessment. This suggested that R2 performed well compared to the other two readers. The K was 68% for R4 in the first assessment. As the over-diagnosis was to the extent of 40%, the trainee reader underwent one more assessment. The K was 64% which was as good as before, but there was no improvement in the over-diagnosis (43.5%) in the second assessment. Based on the performance, only one reader (R2) was certified as qualified for X-ray reading in the first occasion while the reader (R4) assessed in the second occasion was not qualified. These findings were subject to the inherent variation in the SR's readings against which the readers were assessed.