Evaluating smartphone strategies for reliability, reproducibility, and quality of VIA for cervical cancer screening in the Shiselweni region of Eswatini: A cohort study

Abstract

Background Cervical cancer is among the most common preventable cancers with the highest morbidity and mortality. The World Health Organization (WHO) recommends visual inspection of the cervix with acetic acid (VIA) as cervical cancer screening strategy in resource-poor settings. However, there are barriers to the sustainability of VIA programs including declining providers’ VIA competence without mentorship and quality assurances and challenges of integration into primary healthcare. This study seeks to evaluate the impact of smartphone-based strategies in improving reliability, reproducibility, and quality of VIA in humanitarian settings. Methods and findings We implemented smartphone-based VIA that included standard VIA training, adapted refresher, and 6-month mHealth mentorship, sequentially, in the rural Shiselweni region of Eswatini. A remote expert reviewer provided diagnostic and management feedback on patients’ cervical images, which were reviewed weekly by nurses. Program’s outcomes, VIA image agreement rates, and Kappa statistic were compared before, during, and after training. From September 1, 2016 to December 31, 2018, 4,247 patients underwent screening; 247 were reviewed weekly by a VIA diagnostic expert. Of the 247, 128 (49%) were HIV–positive; mean age was 30.80 years (standard deviation [SD]: 7.74 years). Initial VIA positivity of 16% (436/2,637) after standard training gradually increased to 25.1% (293/1,168), dropped to an average of 9.7% (143/1,469) with a lowest of 7% (20/284) after refresher in 2017 (p = 0.001), increased again to an average of 9.6% (240/2,488) with a highest of 17% (17/100) before the start of mentorship, and dropped to an average of 8.3% (134/1,610) in 2018 with an average of 6.3% (37/591) after the start of mentorship (p = 0.019). Overall, 88% were eligible for and 68% received cryotherapy the same day: 10 cases were clinically suspicious for cancer; however, only 5 of those cases were confirmed using punch biopsy. Agreement rates with the expert reviewer for positive and negative cases were 100% (95% confidence interval [CI]: 79.4% to 100%) and 95.7% (95% CI: 92.2% to 97.9%), respectively, with negative predictive value (NPV) (100%), positive predictive value (PPV) (63.5%), and area under the curve of receiver operating characteristics (AUC ROC) (0.978). Kappa statistic was 0.74 (95% CI; 0.58 to 0.89); 0.64 and 0.79 at 3 and 6 months, respectively. In logistic regression, HIV and age were associated with VIA positivity (adjusted Odds Ratio [aOR]: 3.53, 95% CI: 1.10 to 11.29; p = 0.033 and aOR: 1.06, 95% CI: 1.0004 to 1.13; p = 0.048, respectively). We were unable to incorporate a control arm due to logistical constraints in routine humanitarian settings. Conclusions Our findings suggest that smartphone mentorship provided experiential learning to improve nurses’ competencies and VIA reliability and reproducibility, reduced false positive, and introduced peer-to-peer education and quality control services. Local collaboration; extending services to remote populations; decreasing unnecessary burden to screened women, providers, and tertiary centers; and capacity building through low-tech high-yield screening are promising strategies for scale-up of VIA programs.