Comparing Strategies for Implementing Primary Human Papillomavirus Testing for Routine Cervical Cancer Screening

Background In 2018, the US Preventive Services Task Force updated its recommendation for routine cervical cancer screening to include primary human papillomavirus (HPV) screening alongside previously recommended options. Primary HPV screening is more sensitive than Papanicolaou test alone and requires fewer tests per screened woman than cytologic and HPV co-testing. Limited guidance exists, however, for health systems interested in substituting HPV testing for Papanicolaou testing. In response to the new recommendations, Kaiser Permanente Southern California (KPSC) decided to replace co-testing with primary HPV screening for routine cervical cancer screening for women 30 to 65 years of age. The system provides care for about 4.8 million members, of whom 52% are female. We partnered with health system leaders in designing, conducting, and evaluating the health system's practice substitution effort, using an embedded research approach to conduct a cluster randomized pragmatic trial comparing 2 implementation strategies to substitute primary HPV screening for co-testing (ClinicalTrials.gov identifier NCT04371887). Objectives We compared a centrally administered plus a locally tailored (hereafter referred to as locally tailored) approach with a centrally administered plus usual care (hereafter referred to as centrally administered) approach to implement primary HPV testing and deimplement co-testing for cervical cancer screening. Our primary study aim compared the 2 arms in terms of changes in rates of primary HPV screening; feasibility, acceptability, and appropriateness of the new screening approach; and clinician and patient knowledge, perception, behavior, and satisfaction. Our secondary aims included explaining variations in practice substitution strategy effectiveness and developing guidance for use of effective practice substitution strategies in additional settings and for additional implementation and deimplementation scenarios. Methods Twelve KPSC service areas were randomly assigned to receive the locally tailored or centrally administered practice substitution strategy. Physicians, nurses, medical assistants, and women undergoing cervical cancer screening from KPSC primary care and obstetrics-gynecology departments were included in this study. Centrally administered practice substitution included infrastructure updates, educational webinars, and socialization. Sites in the locally tailored arm received these services plus local need assessments to guide selection and deployment of site-selected interventions (eg, clinician and patient education materials and electronic health record [EHR] tools). The primary outcome was the proportion of primary HPV screenings used among all screenings performed. Secondary outcomes included patient- and clinician-centered outcomes; system outcomes such as volume of patient emails about cervical cancer screening and time to colposcopy among women with abnormal screening results; and implementation fidelity, sustainment, scalability for the locally tailored sites only. Surveys and interviews were conducted among women (baseline only) and health care professionals (HCPs) before the systemwide transition (August 2019), shortly after the transition (October-December 2020), and after locally tailored interventions (October-December 2021). Surveys measured site-level needs for the transition, stakeholder-centered outcomes, and fidelity; interviews measured perceived barriers, process evaluation, fidelity, and sustainment. Screening with primary HPV screening or co-testing was measured using KPSC's EHR system and supplementary manual chart review. Whereas the primary outcome was evaluated at the service area level using propensity weighting methods, secondary outcomes were modeled at the individual level using the generalized estimating equation framework and the difference-in-differences (DID) approach when applicable. A stakeholder advisory committee assisted with the study design and data collection tools and helped define stakeholder-centered outcomes. Results During the study period (July 2020-January 2022), an estimated 312 274 women underwent cervical cancer screening. The proportion of appropriate screenings (ie, use of primary HPV screening of the proportion being screened; primary outcome) during the postintervention period ranged from 97.6% to 99.7%. There was no clinically or statistically significant difference between study arms in this outcome: The overall proportion was estimated to be 98.4% (95% CI, 96.3%-100%) for the locally tailored arm and 99.1% (95% CI, 97.8%-100%) for the centrally administered arm. Clinician perception of the effectiveness of primary HPV screening; knowledge about the screening and follow-up algorithm; and ratings of acceptability, appropriateness, and feasibility of primary HPV screening increased in both study arms, whereas rates of HCP delivery of patient education about the new screening test decreased between the baseline and postintervention surveys. There were no significant differences (either clinically or statistically) in these outcomes, however, between study arms (P > .05). There were also no clear differences between the baseline and the postintervention surveys for any of the patient-centered outcomes and no significant differences (either clinically or statistically) between study arms. The volume of patient email inquiries decreased over time between baseline and after the intervention, again with no significant difference (clinically or statistically) between study arms (DID, 1.06-fold; P = .59). Health care professional survey findings suggested reasonable fidelity to the locally tailored interventions. Analyses of secondary outcomes showed that physician specialty and sex did not significantly modify the effect (or lack of effect) of the 2 practice substitution strategies on most HCP-centered outcomes, except that physician sex influenced rates of delivery of patient education. We also did not observe any significant effect modification by women's age and spoken language for the patient-centered outcomes. In our analyses of qualitative interview data, we found that overall, participants from both locally tailored and centrally administered sites reported smooth implementation and a high degree of fidelity to the new HPV primary screening process. The ease of implementation was often credited to the simple practice change of transitioning from 2 specimen collection vials to 1. Regarding fidelity to interventions, interview data revealed moderate fidelity to the centrally administered interventions; group discussion prompted by both interventions appears to have been a strong facilitator of practice change. In the locally tailored arm, few interviewees remembered the locally tailored resources provided (in contrast to the survey findings showing higher rates of recall). The COVID-19 pandemic probably affected recall, with several participants remarking that the pandemic was “front and center,” and therefore educational efforts focused on other issues were not attended or remembered. Several recommendations for future practice change strategies were obtained from the interviews. Conclusions The transition to primary HPV screening was highly successful in KPSC, with high fidelity to the new screening approach in both study arms. The locally tailored practice substitution strategy did not lead to higher uptake of primary HPV screening or better stakeholder-centered or system outcomes than the centrally administered approach. Limitations Generalizability of our findings and our health system's practice substitution approach may be limited because of the unique features of our health system and a key leadership decision to achieve the desired practice change through systemwide policy and procedural change, most notably restricting the co-testing option in the EHR system to follow-up and surveillance use only, essentially forcing the desired practice change for routine screening. Furthermore, the COVID-19 pandemic had negative impacts on the study conduct, significantly reducing HCP availability and attendance at local team meetings, thereby limiting the intensity of local tailoring. COVID-19 also prevented us from conducting site visits for process evaluation, necessitating reliance on remote interview and survey data collection.