Personalized Risk-based Follow-up of Cervical Cancer Screening in Practice, RCT

Cancer screening and management guidelines are moving from a "one size fits all" approach to one that is tailored to an individual. Cervical Cancer Screening Management (CCSM) is the first example of a guideline that incorporates longitudinal screening history to personalize risk assessment. While a personalized approach offers the promise of enhancing the balance of benefits vs. harms of screening and management, this ideal can only be realized through strong systems to support care delivery. Without IT tools that integrate personalized algorithms with care delivery, more errors of judgement may occur if practitioners and patients are confused about appropriate next steps. The investigators hypothesize that delivering risk based CCSM to all eligible individuals will require: 1) using a generalizable Clinical Decision Support (CDS) tool that is validated and open source to implement the American Society for Colposcopy and Cervical Pathology (ASCCP) recommendations; 2) leveraging a system level health IT platform to present risk-based, personalized recommendations; and 3) offering a stepped care approach that individually engages patients and primary care physicians (PCP); and 4) enhances team-functioning, with increasing intensity over time (Figure 1). Accordingly, the investigators propose to develop, implement, and rigorously test PREDICT (Personalized Risk-based Follow-up of Cervical Cancer Screening in Practice) within 4 primary care practice networks.

PREDICT components include supporting individual patient and practitioner engagement, and enhancing team coordination through an efficient "stepped care" approach. This will be compared to standard care by the patient's care team using a 3-arm, randomized design that will allow us to examine the marginal and cumulative effectiveness of the intervention components.

The study will be conducted in four primary care networks that are part of a large healthcare system (Mass General Brigham [MGB]): two affiliated with academic medical centers (Brigham and Women's Hospital [BWH] and Massachusetts General Hospital [MGH]), and two affiliated with community hospitals (Newton Wellesley Hospital [NWH] and North Shore Medical Center [NSMC])

PREDICT will include multilevel components. These will be tested using the 3-arm design that allows assessment of the sequential addition of these components. This design will also evaluate changing the responsibility for "opportunistic" follow-up, typically by the practitioner or patient at the time of a visit, to a systematic, multilevel approach, examining the cumulative and marginal effects of each subsequent level of intervention. In addition to the multilevel components, the individual- and team-level engagement algorithm will take a "stepped care" approach with increasing level of intensity of engagement

Step 1 will start 90 days before the date an individual is due for follow-up ("due date"), at the time of enrollment, based on the patient's personalized risk score. For example, Step 1 for a patient with a 5-year CIN 3+ risk of 0.6% in a practice randomized to an intervention arm would automatically have a 1-year surveillance follow-up reminder set in the EHR. In these intervention practices, practitioners could access this information when in the patient's EHR or when the patient views the reminder in their patient portal. For practices randomized to Arm 3, Step 2 would begin with a patient being sent a reminder letter 90 day prior the due date via the patient portal or mailed if no portal account (Outreach 1). If needed, four weeks later a phone call from the outreach coordinator reminds patients of the follow-up date if scheduled or helps establish follow-up (Outreach 2). The coordinator could also place an order for a referral and send a reminder to the practitioner to sign it if appropriate. The coordinator will review and document any "alternative care plans" discussed by the PCP and patient. A SQL study database will include functionality for coordinators to know when a patient should be contacted, how many contacts have been made, and the outcome of those efforts. Intervention components in Arm 3 will differ for individuals recommended for 3-year surveillance and for individuals 65 years and older who have prior abnormalities that would warrant additional surveillance or diagnostic management. In Arm 3, individuals recommended for 3-year surveillance will only receive Outreach 1 reminder letters (the study will have insufficient resources to make phone calls). Since individuals 65 years and older who have prior abnormalities that would warrant additional surveillance or diagnostic management are more likely to have other comorbid conditions that may affect recommended follow-up, study coordinators will message practitioners prior to performing patient outreach.