Evaluation of inter-observer concordance of the Swede score for digital colposcopic images

Abstract

Objective: To assess inter-observer concordance by training level when applying the Swede score to digital colposcopy images. Materials and methods: Concordance study assembled in a cross-sectional study. The population consisted of patients referred for colposcopy to a cervical pathology reference center in Bogotá, Colombia, due to abnormal cytology. Convenience sampling was used. Measured variables were age, level of education, marital status, risk factors for cervical cancer such as parity, age of sexual activity initiation and number of sexual contacts, contraceptive method, cytology report and colposcopy result according to the criteria of the Swede score. Concordance was calculated using the weighted kappa (k) index. Results: The analysis was based on a set of 3 digital colposcopic images of 251patients. For the total score on the scale, low agreement (k = 0.38; 95% CI: 0.26-0.44) was found between two expert gynecologists; low agreement (k = 0.27; IC 95 %: 0.20-0.38) was also shown for a pair of gynecologists with different training levels, while moderate agreement (k = 0.45; 95% CI: 0.34-0.53) was found for the third pair. For individual criteria on the Swede scale, the best agreement was found for "acetowhite reaction" and "lesion size" (k = 0.43; 95% CI: 0.33-0.53). Conclusions: Inter-observer concordance using the Swede score for digital colposcopic images ranged from low to moderate according to the level of training of the gynecologists.