Multilevel Follow-up of Cancer Screening (mFOCUS): Protocol for a multilevel intervention to improve the follow-up of abnormal cancer screening test results

Abstract

Introduction:

While substantial attention is focused on the delivery of routine preventive cancer screening, less attention has been paid to systematically ensuring that there is timely follow-up of abnormal screening test results. Barriers to completion of timely follow-up occur at the patient, provider, care team and system levels.

Methods:

In this pragmatic cluster randomized controlled trial, primary care sites in three networks are randomized to one of four arms: (1) standard care, (2) “visit-based” reminders that appear in a patient’s electronic health record (EHR) when it is accessed by either patient or providers (3) visit based reminders with population health outreach, and (4) visit based reminders, population health outreach, and patient navigation with systematic screening and referral to address social barriers to care. Eligible patients in participating practices are those overdue for follow-up of an abnormal results on breast, cervical, colorectal and lung cancer screening tests.

Results:

The primary outcome is whether an individual receives follow-up, specific to the organ type and screening abnormality, within 120 days of becoming eligible for the trial. Secondary outcomes assess the effect of intervention components on the patient and provider experience of obtaining follow-up care and the delivery of the intervention components.

Conclusions:

This trial will provide evidence for the role of a multilevel intervention on improving the follow-up of abnormal cancer screening test results. We will also specifically assess the relative impact of the components of the intervention, compared to standard care.

Trial Registration:

ClinicalTrials.gov NCT03979495

Introduction

Routine screening for breast, cervical, colorectal, and lung cancer reduces cancer-specific mortality and is recommended by the US Preventive Services Task Force (USPSTF) and other national guidelines.^1–5 Cancer screening is a process and not a discrete event. A considerable focus of research and quality improvement has been the timely completion of screening in eligible populations. Yet to realize the maximal benefits of screening, timely follow-up of any abnormal screening result is critical but often not achieved.⁶ Abnormal screening results are not rare, with abnormalities found for: 11% of mammograms;⁶ 8% of cervical screens,^6,7 5–10% of fecal occult blood test (FOBT)/ fecal immunochemical test (FIT) screens,⁶ 20% of colonoscopy screenings,⁸ and 10–25%of lung cancer screenings.^9,10

Apart from the legislated requirement for radiologists to follow-up on abnormal mammograms, responsibility for comprehensive screening follow-up falls to the ordering provider, typically a primary care provider (PCP). Unfortunately, few PCPs or their practices have integrated management systems to track abnormalities and manage follow-up. PCPs face the challenge of managing the diagnostic evaluation of multiple cancers with increasingly complex recommendations and timeframes for test completion.¹¹ The number of abnormal results that PCPs are responsible for is staggering;¹² the sheer volume of results supports the need to have a unified, systematic management approach.

The transition from screening to diagnostic evaluation often requires a coordination of responsibility between primary and specialist care.^13,14 Depending on the test and finding, a patient and their care team must track when follow-up testing is needed, typically from a few weeks to many years. Failure to receive appropriate diagnostic testing after an abnormal cancer screening test result (hereafter called an “abnormal screen”) undermines the benefits of screening, violates the trust that patients place in their providers and health systems, and may create medicolegal risk.^15,16

Barriers to follow-up of abnormal screens exist at multiple levels, including the patient, PCP, care team, and health system. Patient barriers such as literacy, financial resources, anxiety, logistical challenges such as transportation and scheduling, and knowledge and beliefs about tests and treatments have been shown to result in delays in timely follow-up.^17,18 PCPs and care teams may face uncertainty about responsibility, lack systems to identify and track when patients are overdue,¹⁹ and lack the time needed to perform outreach. While electronic health records (EHRs) often provide reminders for screening, prompts for follow-up of abnormal screens are uncommon and any such reminders are often displayed only when an individual’s record is open, typically during a visit.²⁰ Even if an EHR delivers non-visit based reminders to the ordering provider, there are rarely tools to facilitate a standard, easily actionable next step.

Though considerable progress has been made in advancing the use of cancer screening, evidence supports an opportunity to improve the systematic follow-up of abnormal screens.⁶ We therefore developed a multilevel intervention to improve the follow-up of abnormal screens which we evaluated in a pragmatic practice-randomized trial. Here we report on the design and implementation of a pragmatic practice-randomized controlled trial that will evaluate the effectiveness of the multilevel intervention. Our main hypothesis is that ensuring follow-up of an abnormal screen requires: 1) care coordination for diagnostic evaluation leveraging a system level health informatics platform; 2) a stepped care approach that individually engages patients and PCPs, and 3) a population health intervention to enhance clinical team functioning.

Methods

Overall Approach

While the majority of work on cancer screening and diagnostic evaluation has focused on a single type of cancer, PCPs take a “whole person” approach with responsibility for a wide range of preventive medicine efforts. Systems to improve comprehensive cancer detection require coordination of the PCP, who is responsible for cancer screening, with specialists who are often involved in the diagnostic evaluation of abnormal screens. The mFOCUS (multilevel Follow-up of Cancer Screening) trial is designed from the perspective of PCPs who need to ensure diagnostic follow up for all four of the USPSTF-endorsed cancer screening tests.^1,2,4,5 Because we recognize there may be multilevel factors that influence receipt of a diagnostic evaluation, our study design allows us to examine the effect of intervention components at the individual (patient, PCP), clinical team (coordination between patient, PCP, and specialty care team members), and system (health information technology infrastructure) levels (Figure 1).

Figure 1.

Multilevel factors that influence the follow-up of an abnormal cancer screening test result.

Experimental Design

mFOCUS is a 4-arm cluster (primary care site) randomized controlled trial (RCT) of individuals who are overdue for follow-up of an abnormal screen. The four arms are: (1) standard care, (2) “visit-based” reminders that appear in an patient’s EHR record when it is accessed and can be viewed by both patient and providers (3) the visit based reminders with additional population health outreach, and (4) visit based reminders, population health outreach, and patient navigation with systematic screening and referral to address social barriers to care (Figures 1 and 2). In the intervention arms, components are added sequentially using the timeframe depicted in Figure 2. Our stepped care approach allows us to compare the cumulative addition of each level of intervention to standard care as well as the marginal effect of each additional level to the prior group.

Figure 2.

Study Flow

Ethical

This study protocol has been reviewed and approved by the Mass General Brigham (MGB) Institutional Review Board (IRB). The MGB IRB is the reviewing IRB for all participating institutions, as determined by SMART IRB. Since mFOCUS is designed as a “fail safe” system to supplement rather than replace standard care, the care of all patients remains under the direction of their personal care team, with randomization at the practice level. The trial obtained a waiver of informed consent.

Study Setting

The study is being conducted in three primary care networks: two affiliated with MGB in Massachusetts (MA), Brigham and Women’s Hospital (BWH) and Massachusetts General Hospital (MGH), and a third affiliated with Dartmouth-Hitchcock Health (D-HH). The practices in these networks provide diversity in location (urban, suburban, rural) and patient characteristics including race/ ethnicity and socioeconomic status. The MGB primary care networks share a single Epic EHR; the D-HH affiliated practices share a separate Epic EHR.

Practice Randomization

Primary care site is the unit of randomization; 44 primary care sites were randomized into one of the four study arms prior to patient enrollment. To achieve balance of both the number of patients in each arm as well as the characteristics of patients within each arm, practices were divided within strata determined by practice size, and the number of people who were female or had Medicaid coverage (as a proxy for socioeconomic status). We then ran simulations of the randomization 100 times using the groups created by matching within strata to examine the likelihood of significant imbalance across the study arms for MGB and D-HH practices separately. Once we were assured that our strata would maximize the probability of having similar patient populations across the arms, practices were randomly assigned to one of the four study arms.

Patient Inclusion Criteria

Individuals with an abnormal screen who are overdue for follow-up are eligible for mFOCUS, including:

• Breast: females 40–80 years with an abnormal mammogram or digital breast tomosynthesis (DBT) exam.

• Cervical: females 21–65 years with an abnormal Pap test with or without an HPV test.

• Colorectal: adults 40–80 years with a positive fecal immunochemical test (FIT) or a short interval colonoscopy as specified by the care team (1, 2, 3, 5 years).

• Lung: adults 55–80 years, current and former smokers, with an abnormal low-dose computed tomography (LDCT) result.

For all abnormal screens, except short-interval colonoscopies, we use relevant guideline recommendations and clinical input from experts at participating institutions to determine a follow-up timeframe for each specific abnormality. The follow-up timeframe for short-interval colonoscopy is determined by the gastrointestinal specialist performing the procedure at all three participating sites. Since mFOCUS is designed to operate as a “fail safe,” to ensure follow-up for individuals who are overdue for diagnostic evaluation, we allow an additional 2 months (for high risk abnormalities) to 6 months (for low risk abnormalities) after the recommended follow-up interval, before a patient becomes eligible for mFOCUS to allow the care team to schedule follow-up (Table 1). Patients are eligible for the study on the basis of their first abnormal screening result during the study period. To minimize confusion for the care team, if a patient has more than one abnormal screen during the study period, they receive the intervention defined by the randomization status of their primary care site for all abnormalities. However, only the first abnormal screen is included in the trial.

Table 1.

Summary of recommended follow-up interval and mFOCUS eligibility window based on screening abnormality

Organ	Follow-up Group	Abnormal result	Appropriate diagnostic test (next test only)	Recommended follow-up (months since diagnostic test due)	Eligible for mFocus (months since diagnostic test due)
Cervix	High-Risk	Any pap cytology, HPV high risk genotype positive (HPV 16/18 or 45 (MGB only)) *	Colposcopy	3	6
	High-Risk	LSIL or ASCUS, HPV+, age 25 years or above	Colposcopy	3	6
	High-Risk	Higher cytology abnormality (ASCH, High grade SIL, HSIL)	Colposcopy	3	6
	High-Risk	LSIL pap, HPV not done, age 25 years or above	Colposcopy	3	6
	High-Risk	AGC pap	Colposcopy	3	6
	High-Risk Endometrial	Endometrial cells, age 50 years or above	Endometrial biopsy	3	6
	Med-Risk	LSIL or ASCUS Pap, HPV +, under age 25 years	Pap with HPV (reflex or cotest) OR Colposcopy	12	15
	Med-Risk	LSIL pap, HPV −	Pap with HPV (reflex or cotest) OR Colposcopy	12	15
	Med-Risk	ASCUS pap, HPV not done OR LSIL pap, HPV not done, under age 25 years	Pap with/ without HPV (reflex or cotest)	12	15
	Med-Risk Cotest	Normal pap, HPV +	Pap with HPV OR Colposcopy	12	15
	Low-Risk	ASCUS pap, HPV−	Pap with or without HPV	36	42
Breast	High-Risk	Birads 5	Tissue sample (FNA, biopsy, surgery)	1	3
	High-Risk	Birads 4	Tissue sample (FNA, biopsy, surgery)	1	3
	Low-Risk	Birads 3	Follow-up imaging	6 or 12**	9 or 15**
CRC	High-Risk	Positive FIT	Colonoscopy	3	6
	Med-Risk	Colonoscopy requiring 1-year follow-up	Colonoscopy	12	18
	Low-Risk	Colonoscopy requiring 2 or 3-year follow-up	Colonoscopy	36	42
	Low-Risk	Colonoscopy requiring 5-year follow-up	Colonoscopy	60	66
Lung	High-Risk	Lrads 4b or 4x	Chest CT, Biopsy/ Tissue sample, Thoracoscopy, Bronchoscopy, Appointment in a multi-disciplinary nodule clinic	1	3
	High-Risk	Lrads 4a	Chest CT, LDCT, Biopsy/ Tissue sample, Thoracoscopy, Bronchoscopy, Appointment in a multi-disciplinary nodule clinic	1	3
	Low-Risk	Lrads 3	Chest CT, LDCT, Biopsy/ Tissue sample, Thoracoscopy, Bronchoscopy, Appointment in a multi-disciplinary nodule clinic	6	9

NOTES:

^*DH variation: Women < 25 years with NILM or LSIL pathology were recommended to receive follow-up at 12 months and were eligible for mFOCUS at 15 months and could receive Pap with HPV (reflex or cotest) or colposcopy.

^**As specified by radiologist on care team

Exclusion criteria

We exclude patients who: (1) are not English or Spanish-speaking as we are unable to deliver the intervention in other languages; (2) have a history of cancer of the organ for each screening test as these individuals may be at higher risk and therefore require personalized evaluation (e.g., women with prior breast cancer are ineligible for an abnormal mammogram). For colorectal cancer, we also exclude individuals with a history of Crohn’s disease or inflammatory colitis because we cannot differentiate individuals who are undergoing colonoscopy for disease/treatment surveillance from those receiving cancer screening. Of note, we do not exclude individuals who are undergoing short interval colonoscopy screening for a high risk family history alone because we cannot distinguish them from those with an abnormal screen.

Enrollment

Practice and Provider Engagement.

We obtained support of network leadership for the study and attended meetings with practice leadership at intervention sites to educate staff about the study and its procedures. PCPs are not required to recruit individuals for the study as identification of eligible patients, as this is done centrally by study staff using the new EHR functionality described below. Practices and providers received an email prior to the study launch to describe the study with additional information relevant to the study arm of their practice.

Patient Identification and Enrollment.

Implementation of the informatics functionality varies between MGB and D-HH in order to accommodate the different configurations of the two EHRs (Table 2). In contrast to the D-HH EHR, the MGB EHR has limited availability of coded result data accessible in a data warehouse. In both systems, algorithms were built and validated to identify eligible patients who fit the inclusion criteria and were assigned them to a participating primary care site with an associated study arm. At MGB, the algorithms are performed external to the EHR in a secure study database, and coded data is passed back into the EHR using web services. Within the D-HH network, patient registries were built using EHR functionality and the same inclusion criteria as MGB.

Table 2.

Summary of Informatics Implementations at Each Site

	Site
Informatics Task	Mass General Brigham	Dartmouth Hitchcock Health
Patient idenfication, establish eligibility
EHR		Identify eligible patients using existing codified result fields (i.e. BIRADS, LRADS, cervical cytology) Indicate eligible patients in EHR by adding mFOCUS-specific labels to patient record
Study Database	Query the EHR data warehouse daily to extract relevant reports (mammography, Pap smear, LDCT) Run reports through organ-specific natural language processing (NLP) to identify abnormal results Identify eligible patients using extracted data Indicate eligible patients in the EHR using a study-specific health maintenance reminder and automated problem list entry added to patient charts via web services
Intervention Delivery
EHR	Workbenches generated using mFOCUS-specific health maintenance reminder and automated problem list entries Send patient portal messages and generate letters via workbenches	Workbenches generated by searching for mFOCUS-specific labels in the patient record Send patient portal messages and generate letters via workbenches
Tracking Intervention Delivery
EHR		Record dates, outcomes of calls in workbenches Automated tracking of study outcomes (e.g. cancer diagnoses, completed follow-up tests)
Study Database	Record dates, outcomes of calls in sidecar Automated tracking of study outcomes (e.g. cancer diagnoses, completed follow-up tests)	Copy of the tracking information from EHR is pushed to sidecar for study monitoring

Intervention Components

Intervention components are intended to augment the care delivered by the primary care team. The 4-arm design permits comparing “standard care” as delivered by their PCPs and practices (Arm 1) to three intervention arms that represent the sequential/stepped addition of: (Arm 2) a system-level re-design that displays provider and patient facing reminders when an individual record is open. Also included are health maintenance reminders that are specific to the abnormal screen in need of follow-up and, in two of the three networks, automated problem list entries;²¹ (Arm 3) the addition of population-based outreach that approximates many population management programs. There is an automated single patient portal message (if the patient participates) or a single mailed letter followed by single reminder phone call from study staff; and (Arm 4) the addition of a patient navigation which includes phone calls to build trust, assess and address social barriers to care. At MGB documentation of outreach is recorded in the study database and is recorded in the EHR at D-HH but is automatically pushed to the study database for tracking (Table 2). Protocol fidelity is monitored weekly by assessing outreach attempts for active patients using pre-specified reports in the study database. Study staff meet weekly to review protocol deliver and outcome assessment and adjudicate cases of uncertain eligibility.

A central role of the patient navigator is to conduct a systematic screening for social barriers across nine domains including: housing and food insecurity, paying for basic utilities, family caregiving, legal, transportation, financial compensation for treatment, education, and employment. The study used Aunt Bertha (https://company.auntbertha.com/) to develop a mFOCUS-specific screening and referral system to support the patient navigators’ ability to connect patients with social services available in their community. Aunt Bertha is an online network of verified social service programs including nonprofits and social care providers.²² Navigators work with patients to identify the most pressing social domains and then identify available community services to address any issue preventing a patient from receiving follow-up.

The four-arm design allows us to evaluate changing the responsibility for “opportunistic” follow-up, typically by the PCP at the time of a visit (Arm 1), to a systematic, multilevel approach, examining the cumulative and marginal effects of a “stepped care” approach with increasing level of intensity of engagement across the study arms. The stepped timing of the intervention components is diagramed in Figure 2.

Outcomes

For each abnormal result, we use guideline recommendations and clinical input from experts at the participating institutions to identify follow-up tests or procedures that would represent appropriate follow-up. The primary outcome is whether an individual receives follow-up, defined based on the type of screening abnormality and organ type, within 120 days of becoming eligible for mFOCUS (Table 1). Completion of the outcome is documented using automated algorithms that run daily in the EHR at D-HH or the study database at MGB. Secondary outcomes assess multi- and cross-level (individual, team, system) outcomes (Table 3). Finally, we assess the delivery of the intervention components, using documentation from the study database, patient and PCP surveys described below.

Table 3.

Summary of Secondary Outcomes

Level of Measurement	Level Specific Measures (data source)	Cross Level Measures
Individual Patient	# of days to follow-up (EHR) Satisfaction with follow-up care (patient survey)	Patient-Reported Assessment of: Individual-PCP: Satisfaction with care provided by PCP Care Team: Satisfaction with care provided by care team System: Receipt of reminder portal message, mailed letter, or call
PCP	Satisfaction with ability of patients to get timely follow-up (PCP survey) Practice barriers to management (PCP survey)	PCP-Reported Assessment of: Individual-Patient: Patient understanding of need for follow-up Care Team: Barriers to coordination of care System: Satisfaction with IT functionality to support follow-up

Data Sources

EHR

The EHR is the data source for all socio-demographic and clinical covariates, as well as the primary outcome. EHR data are captured in two ways: (a) automated extraction of all demographic variables and discrete clinical data elements from structured EHR fields. (b) At MGB some clinical findings (e.g. cervical cytology, mammography and lung cancer screening results) are available in free-text reports and are not available in a structured format. Therefore, we use natural language processing (NLP) to apply a set of regular expressions and logic rules to identify abnormal findings from narrative reports and convert findings to structured data elements.²³ In addition, we obtain data from the EHR on: age, race/ ethnicity, primary language, education, insurance status, prior cancer history, body mass index, comorbidity including tobacco use,²⁴ and vital status.

Study Database

A SQL database (PostgreSQL at Dartmouth, Microsoft SQL Server at Partners) with a web-based interface (React, a JavaScript library for building user interfaces, as well as scripts written in PHP at Dartmouth and C# at MGB) is used to assess eligibility (at MGB), and track participant enrollment and outreach status (at MGB and D-HH).

PCP Surveys

A brief, self-administered survey was administered to PCPs before the trial was launched in February-March 2020 at MGB and September-October 2020 at D-HH and will be repeated at the end of the study to assess change associated with the intervention. Topics include perceptions of who is responsible for the follow-up of an abnormal result, mechanisms to track whether follow-up has been obtained, difficulty scheduling follow-up, provider and system barriers to follow-up, practice resources to assist with follow-up, and satisfaction with the process of managing an abnormal result. PCP surveys are administered using REDCap (up to 5 email contacts with a REDCap link over 4 weeks).

Patient surveys

A patient survey collects information about the experience of receiving follow-up care (if any). Topics include communication with the care team, knowledge of follow-up recommendation, barriers to receiving a workup and how the COVID-19 pandemic influenced patient experience of, and plans for follow-up evaluation. We randomly sample 10% of eligible participants for the survey. Surveys are administered 4–12 weeks after the follow-up period has finished. Patients are sent an introductory letter from the clinical director of each primary care network paired with a letter from the study investigators that explains the study purpose and includes a unique link for accessing the web-based survey (REDCap). Patients can opt out of the survey at any time. If patients do not opt-out within 2 weeks of receiving the letters, they are contacted by phone and asked to either opt out or complete the survey using the REDCap link or by phone.

Statistical Analysis

The primary analysis is intention-to-treat (ITT), including all eligible patients. We expect that a small number of patients may change their primary care site within our systems during the study period. These individuals will be evaluated according to their initial randomization status. Through the stratified randomization we attempt to balance patient, provider and practice characteristics between study groups. However, since randomization occurs at the practice level, we will be cautious and compare the patient, provider and practice characteristics using Fisher exact tests, t-tests and Wilcoxon tests as appropriate to the covariate distribution. Any characteristics that show substantial clinical or statistical difference (p<.05) will be entered into the regression model and retained as covariates if they alter the effect estimate of the intervention by >20%. We anticipate minimal missing data for our primary analyses since outcome and covariate information will come from the electronic health record. However, multiple imputation will be used to account for any missing data.

The primary analysis model will be a random effects logistic regression, implemented through the SAS Glimmix procedure. Timely follow-up (yes/no) will be the patient-level outcome with a random effects for practice and PCP to allow for exchangeable correlation between patients seen within the same practice and by the same PCP. The primary fixed predictors will be indicator variables representing the three intervention arms and we will use a global likelihood ratio test to compare the 4 study arms. If the global test is significant (p<0.05), we will compare the intervention arms to the control group (our primary comparisons with a Bonferroni-adjusted significance level of 0.0167) and to each other (secondary comparisons). Comparisons to the control arm will capture the cumulative effects of the interventions, while comparisons between the intervention arms will capture the marginal effects of each level of intervention. We will include covariates for the type of cancer, level of initial screening abnormality, and any patient, provider or practice characteristics which are identified as confounders. Results will be presented as adjusted follow-up rates, with 95% confidence intervals, calculated using marginal standardization.

Secondary analyses will model time-to-follow-up, using a clustered proportional hazards regression to examine how quickly abnormal screens are followed-up with censoring at the end of study for patients who never received follow-up. Patients will be clustered within practices using the generalized estimating equation approach, implemented as a “frailty” analysis in the SAS Phreg procedure. An additional correlation component for patients within practices cannot be included, but we will evaluate the robustness of our findings by using an alternative model, clustering by practice rather than provider. Predictors and covariates in this model will be identified in the same way as described above. The proportional hazards assumption for the intervention effects will be verified by entering a time-varying version of those predictors. Other secondary outcomes based on patient, provider and team surveys will use clustered linear regression models to compare satisfaction scores between study arms. The model building will be analogous to the approach detailed above.

Sample Size and Power

We initially planned to recruit approximately 3,324 participants from 40 clinics estimated to be eligible for the study when the study was designed. This sample size would give us 80% power to detect an 11% difference in the follow-up rate, a clinically relevant change in the process of care. The sample size calculation allowed for a within-provider correlation of 0.02 (assuming 550 providers), a within-practice correlation of 0.01 (assuming 40 practices), and a Bonferroni-adjusted significance level of 0.0167 to allow for comparison of each of the 3 active intervention groups to the control arm. The pandemic, however, resulted in a much larger than anticipated number of eligible individuals who had delayed care and unanticipated waits for some diagnostic tests. Because this is a pragmatic trial embedded in clinical practice, we estimate that there will be approximately 12,000 eligible individuals by the time recruitment closes at the end of 2021. To date, we have enrolled over 6,000 individuals. Likewise, we initially anticipated that there would be 40 participating clinics but 44 were interested in participating when the project launched during the pandemic.

Discussion

The true benefit of cancer screening can only be attained if we assure timely follow-up care for all individuals with an abnormal screen. Prior research underscores the prevalence of incomplete follow-up, notes variation in follow-up across organ types and health care settings, and highlights the importance of developing systematic, multilevel interventions to address multilevel barriers to care.⁶ mFOCUS leverages several key advances to improve follow-up of abnormal screens including: (1) an EHR-integrated registry to facilitate identification and tracking of patients from a population-based perspective; (2) automated EHR-integrated reminders and problem list entries to remind patients and providers that follow-up is overdue; (3) population management using a team-based approach widely adopted in primary care whereby, outside of visits, staff members identify and reach out to patients with unmet care needs to improve adherence; and (4) patient navigation with systematic screening for social determinants of health to address barriers to receiving timely care.

Conceptual models suggest the importance of multilevel influences on cancer care delivery and identify “interfaces” where failures can occur to inform interventions.^13,14 mFOCUS is one of the first pragmatic trials where intervention components target these multiple levels of influence and also targets the interfaces between individuals and teams that require communication or transfer of responsibility (Figure 1).¹⁴ We hypothesize that these components will lead to productive interactions between an informed, active patient and a prepared, proactive PCP and care team, resulting in improved outcomes. Other key innovations of our approach include: 1) taking a patient- and PCP-centered “whole person” perspective to integrate follow-up management for four cancers; 2) stratifying patients based upon abnormality risk; 3) a health IT platform for population management based in primary care with pre-defined information exchange to support multilevel team engagement; and 4) applying a 4-arm randomized design to evaluate the effectiveness of mFOCUS components at the system, team, and individual levels. While this design is innovative and comprehensive it is not without limitations. Randomization status is not blinded and our system does not actively engage specialist physicians. All the eligible practices have agreed to participate, and none have withdrawn after learning their randomization status. Our study is not designed to assess the sustainability of our intervention, although we are beginning to have conversations about transitioning some of the infrastructure from research to clinical operations. Because our study launched prior to the most recent USPSTF for lung cancer screening, we include individuals who are at least 55 (and not 50).⁵ Given low volumes of lung cancer screening this is unlikely to limit the generalizability of our findings. Our study was also designed prior to the 2020 release of the American Society of Colposcopy and Cervical Pathology (ASCCP) risk-based consensus guidelines¹¹ and are based on the most recent Pap/HPV result. Although both MGB and DH use the same EHR vendor, the local implementation differs requiring somewhat different approaches to the informatics implementation. Our design has several strengths: the intervention is deployed in three primary care networks, uses a commonly available EHR allowing dissemination to other settings, and uses a pragmatic design. While not anticipated when this study was conceived, the COVID-19 pandemic has created both opportunities and challenges for mFOCUS. Because of the mandated closure of many elective medical services during 2020 there is a larger than anticipated backlog of patients in need of follow-up for an abnormal screening result who can therefore benefit from our intervention components. This resulted in the decision to increase our planned enrollment to accommodate these individuals. Unfortunately this backlog, as well as the ever-changing complexities of infection control in a pandemic created less capacity for certain diagnostic procedures particularly colonoscopy. Certain aspects of our study design, most notably the centralized, remote delivery of the intervention components and the systematic screening and referral for social determinants of care are even more relevant now than they were when the study was designed.

Completion of the trial will provide evidence for the role of a multilevel intervention on improving the follow-up of abnormal cancer screening test results. We will also be able to assess the relative impact of the components of the intervention to see if they improve outcomes compared to standard care.

Highlights.

• Barriers to the completion of timely follow-up of abnormal cancer screening are common.

• Barriers occur at the patient, provider, care team and system levels.

• This pragmatic cluster randomized trial, will test sequentially more intensive multi-level intervention.

• This trial will provide evidence for the role of a multilevel intervention on improving follow-up