Secular trend in interobserver agreement of VIA diagnosis for cervical cancer screening in Nigeria

Objective In low resource settings, visual inspection with acetic acid (VIA) by allied health workers, has been suggested as an alternative for cervical cancer screening. However, there are concerns about the objectivity and time to diagnostic concordance with specialists. We evaluated the secular trend in interobserver agreement between nurse providers and a gynecologist/colposcopist over a five-year period. Methods Nurses provided VIA screening with digital cervivography to 4,961 participants in five screening clinics from October 2010 to May 2014 in Nigeria in this observational study. Cer-vigraphs were reviewed at meetings where a gynaecologist/colposcopist made an assessment from the cervigraphs. We used weighted kappa statistics to calculate agreement in diagnosis between nurse providers and the gynecologist/colposcopist; linear regression models to examine overall trend and investigate potential clinic characteristics that may influence agreement; and time series models to characterize month to month variations. Results Mean age of participants was 37±8 years. Overall agreement was 0.89 at Site D, 0.78 and 0.73 at Sites A and C respectively, 0.50 for Site E and 0.34 for Site C. The number of trainings attended by nurse providers(β = 0.47,95%CI:0.02–0.93, p = 0.04), high level of engagement by site gynecologists(β = 0.11,95%CI:0.01–0.21,p = 0.04) were associated with increased agreement; while increasing distance from the coordinating site(β = -0.47,95%CI:-0.92–0.02, p = 0.04) was associated with decreased agreement. There were no associations between number of years screening clinics were operational(β = 0.01,95%CI: -0.01–0.03,p = 0.29), cumulative experience of nurse providers(β = 0.04,95%CI:-0.03–0.12,p = 0.19) and agreement. There were no significant increases in weighted kappa statistics over time for all sites considered. Monthly variations were significant for only one of two sites considered in time series models (AR1 term = -0.40, 95%CI:-0.71–0.09,p = 0.01). Conclusion Our results showed a lack of objectivity, persistent variation and lack of convergence of diagnostic capabilities of nurse led VIA cervical cancer screening with the diagnostic capabilities of a specialist in a cervical cancer screening program in Nigeria.