Interpretation of colposcopy in population-based cervical screening services in north-eastern Italy: an online...

An innovative web-based colposcopy quality assurance programme was implemented in population-based cervical screening services in three north-eastern Italian administrative regions with different colposcopists' training background. In this study, the levels of intra- and interregional intercolposcopist diagnostic agreement were evaluated.

Of the 158 registered colposcopists, 125 accessed the website of the programme, logged-in, viewed a posted set of 50 digital colpophotographs selected by an expert steering committee, and classified them for the colposcopic impression, the visibility of the squamocolumnar junction, and the need for biopsy. Anonymous data were downloaded and analysed using the crude, or observed, proportion of agreement and the kappa coefficient.

There were 113 eligible colposcopists. Overall, crude agreement on the colposcopic impression, the visibility of the squamocolumnar junction, and the need for biopsy was 0.72, 0.72, and 0.87, with kappa values of 0.60, 0.36, and 0.69, respectively. The homologous kappa values were 0.61, 0.41, and 0.69 in one region, 0.57, 0.36, and 0.69 in another, and 0.66, 0.38, and 0.74 in the third. Total intra- and interregional agreement were nearly identical, with kappa values of 0.59 and 0.60 for the colposcopic impression, 0.38 and 0.35 for the visibility of the squamocolumnar junction, and 0.69 and 0.69 for the need for biopsy. The width of 95% confidence intervals around the above kappa values was ≤0.01.

The levels of agreement varied between moderate and substantial both within and between regions. Regional differences in training background had minor effects. The interpretation of colposcopy is potentially well-reproducible.