Prevalence of male circumcision in four culturally non-circumcising counties in western Kenya after 10 years of program implementation from 2008 to 2019

Abstract

Introduction Kenya started implementing voluntary medical male circumcision (VMMC) for HIV prevention in 2008 and adopted the use of decision makers program planning tool version 2 (DMPPT2) in 2016, to model the impact of circumcisions performed annually on the population prevalence of male circumcision (MC) in the subsequent years. Results of initial DMPPT2 modeling included implausible MC prevalence estimates, of up to 100%, for age bands whose sustained high uptake of VMMC pointed to unmet needs. Therefore, we conducted a cross-sectional survey among adolescents and men aged 10–29 years to determine the population level MC prevalence, guide target setting for achieving the goal of 80% MC prevalence and for validating DMPPT2 modelled estimates. Methods Beginning July to September 2019, a total of 3,569 adolescents and men aged 10–29 years from households in Siaya, Kisumu, Homa Bay and Migori Counties were interviewed and examined to establish the proportion already circumcised medically or non-medically. We measured agreement between self-reported and physically verified circumcision status and computed circumcision prevalence by age band and County. All statistical were test done at 5% level of significance. Results The observed MC prevalence for 15-29-year-old men was above 75% in all four counties; Homa Bay 75.6% (95% CI [69.0–81.2]), Kisumu 77.9% (95% CI [73.1–82.1]), Siaya 80.3% (95% CI [73.7–85.5]), and Migori 85.3% (95% CI [75.3–91.7]) but were 0.9–12.4% lower than DMPPT2-modelled estimates. For young adolescents 10–14 years, the observed prevalence ranged from 55.3% (95% CI [40.2–69.5]) in Migori to 74.9% (95% CI [68.8–80.2]) in Siaya and were 25.1–32.9% lower than DMMPT 2 estimates. Nearly all respondents (95.5%) consented to physical verification of their circumcision status with an agreement rate of 99.2% between self-reported and physically verified MC status (kappa agreement p-value<0.0001). Conclusion This survey revealed overestimation of MC prevalence from DMPPT2-model compared to the observed population MC prevalence and provided new reference data for setting realistic program targets and re-calibrating inputs into DMPPT2. Periodic population-based MC prevalence surveys, especially for established programs, can help reconcile inconsistencies between VMMC program uptake data and modeled MC prevalence estimates which are based on the number of procedures reported in the program annually.