Designing Guidelines for Those Who Do Not Follow Them: The Impact of Adherence Assumptions on Optimal Screening Guidelines

Abstract

Objectives

Model-based cost-effectiveness analyses can inform decisions about screening guidelines by quantifying consequences of alternative algorithms. Although actual screening adherence is imperfect, incorporating nonadherence into analyses that aim to determine optimal screening may affect the policy recommendations. We evaluated the impact of nonadherence assumptions on the optimal cervical cancer screening in Norway. Methods

We used a microsimulation model of cervical carcinogenesis to project the long-term health and economic outcomes under alternative screening algorithms and adherence patterns. We compared 18 algorithms involving primary human papillomavirus testing (5-yearly) that varied follow-up management of different human papillomavirus results. We considered 12 adherence scenarios: perfect adherence, 8 high- and low-coverage “random-complier” scenarios, and 3 “systematic-complier” scenarios that reflect conditional screening behavior over a lifetime. We calculated incremental cost-effectiveness ratios and considered a strategy with the highest incremental cost-effectiveness ratio < 55 000 US dollars/quality-adjusted life-year as “optimal.” Results

Under perfect adherence, the least intensive screening strategy was optimal; in contrast, assuming any nonadherence resulted in a more intensive optimal strategy. Accounting for lower adherence resulted in both lower costs and health benefits, which allowed for a more intensive strategy to be considered optimal, but more harms for women who screen according to guidelines (ie, up to 41% more colposcopies when comparing the optimal strategy in the lowest-adherence scenario with the optimal strategy under perfect adherence). Conclusions

Assuming nonadherence in analyses designed to inform national guidelines may lead to a relatively more intensive recommendation. Designing guidelines for those who do not adhere to them may lead to over-screening of those who do.