Increasing uptake of evidence-based screening services though a community health worker-delivered multimodality program: study protocol for a randomized pragmatic trial

Our specific aim is to determine if a strategy in which CHWs deliver multimodality, home-based screenings results in an increase in the proportion of participants screened for all four conditions (HIV, HCV, CRC, and cervical cancer) as compared to a strategy in which patients are navigated to primary care by a CHW at one of our participating health centers. Our study will have over 95% power to examine our primary hypothesis that the CHW led multimodality screening strategy will result in a 15 percentage point increase in participants who are up to date in screening for these four conditions (three for men) as compared to a strategy of linkage to primary care. Secondary analysis will examine increases in screening for each of the four conditions individually. Subgroup analysis will include examining outcomes by race/ethnicity and by gender.

We are recruiting 900 participants from various underserved communities in Miami-Dade County. Of the 2.8 million county residents, 69% are Hispanic and 18% are Black [15]. Further, whereas 30 years ago, Cubans made up 80% of Miami Hispanics, over the last decades a rapid influx of immigrants from most of Latin America now make Miami as the most diverse Latino community in the USA. In 2018, 52%Latinos were Cuban, 20%¹⁷ were South American, 15% were Central American, and 9% were Puerto Rican or Dominican[15]. South Florida’s Black population is also diverse, and Miami has the largest population of Black immigrants. Nationally, the Black immigrant population has quadrupled and they will represent nearly 20% of Black Americans in 50 years. However, at present, more than a third of Black residents in Miami are foreign-born. This includes the 484,000 Haitians living in Florida, who represent 47% of all US Haitians [16]. For the current study, we are focusing recruitment primarily in three communities: Little Haiti (predominantly Haitian), Hialeah (predominantly Hispanic), and South-Dade (racially/ethnically mixed).

As with prior studies, our community outreach, recruitment, and intervention approaches are led by CHWs. Our study team includes three CHWs all from our target communities who are bilingual in either English and Spanish or English and Haitian Creole. All three are certified CHWs by the Florida Certification Bureau and each has over 5 years of experience as a CHW. All three have completed the University of Miami’s required research training in human subject research as well as formal CHW research training using modules developed by our study team. They also received additional study-specific training from our academic research team, organized into three modules: (1) health education on each of the four conditions, (2) specimen collection procedures, and (3) study data collection procedures. Each of these sessions was interactive and hands-on, and CHWs were able to practice and receive feedback on their data collection and study procedure skills.

Trial recruitment occurs via convenience sampling at community-based locations such as markets, churches, botanicas, and health clinics, as well as via the social networks of our CHWs and community partners. The CHWs approach potentially eligible individuals, screen for eligibility, and explain the study. Eligible and interested individuals are asked to provide their contact information, and CHWs will try up to ten times to contact eligible individuals to schedule them for informed consent and study intake. Participants who agree to participate complete a 30‑minute intake interview in their home or location of their preference. The CHWs work closely with our two bilingual research associates (RAs) to set up these visits during which informed consent and study intake occur. Our RAs are master’s level highly trained research professionals. Upon completion of the intake interview, participants are randomized into one of two interventions.

Following the study intake, participants are randomized to one of the two study interventions by the study biostatistician, using REDCap’s randomization module. The module randomizes study participants electronically, and randomizes within study sites to account for nesting. The CHWs are notified via email of the assignment of participants following the randomization, and will follow up with participants within a week to inform them of their assignment, and to schedule a home visit (if the participant is in the home-based testing arm). Notably, while participants and CHWs are aware of group assignment, the community health educator (CHE), who collects study outcome data, is blinded to participant assignment.

The baseline screening rates in our sample size calculation rates are obtained from our feasibility studies and local data. The overall sample size was chosen to ensure sufficient power to detect clinically meaningful effects associated with a 15% point increase in proportion of additional screening done in the home-based arm versus primary care arm using two-sided test for differences in proportion. Although we plan to conduct intention-to-treat analyses, in which all women lost to follow-up will be considered “unscreened,” we will also conduct analyses with lost to follow-up participants excluded. For these analyses, we need to account for attrition (21% in our previous study). In addition, in our subgroup analysis where we look at each individual test, we set alpha at 1.25% instead of 5% to account for multiple testing. Even with a 25% attrition rate, a proposed total sample size of 900 participants (450 per arm and 150 per site) will give us over 95% power for our primary outcome and 90% power for the subgroup analysis.

Intervention fidelity is assessed via weekly review of CHW encounter logs, which are documented every time the CHW has contact or attempts contact with a participant. Any issues or inconsistencies elucidated by this regular review are addressed and resolved at team meetings. In addition, we will assess participant knowledge of the four target conditions and screening following the intervention, which can help assess if disease education was administered effectively by the CHWs.

Participants may request to be withdrawn from the study at any time by contacting the CHE. If a participant elects to withdraw, their reason for withdrawal will be recorded and their study data will be deleted.

Study data will be collected and stored within the REDCap system, which is a web-based, secure, and HIPAA-compliant research management platform. REDCap fully supports several research processes for patient recruitment and scheduling, participant randomization, data entry and management, as well as data safety monitoring and adverse event reporting. A particular advantage of REDCap is that CHEs are immediately notified if fields are missing and/or if out or range values are entered. However, in case of loss of internet access, backup paper copies are provided, which would then be subsequently entered into REDCap. The data manager regularly screens the REDCap data for quality control, and any issues or inconsistencies are discussed with the CHEs. De-identified data is extracted from REDCap for data analysis.

Prior to any analysis, data screening for accuracy will be conducted and will focus on several aspects. We will determine the amount, pattern, and randomness of missing data. Outliers will be identified observations that appear to be very high or very low, and decide how to proceed based on the presumed cause. We will also identify inconsistencies within a single variable and between variables. As there are various types of measurements, we will use various descriptive statistics to visualize the study data with the appropriate graphical displays. Initially, we will perform an explanatory data analysis; visually via graphics/plots such as scatter and box-whisker plots and numerically by descriptive statistics such as range, median, means, standard deviations for measurements taken on a continuous scale, and percentages, various types of cross-tabulations for measurements taken on a categorical scale. Corresponding confidence intervals for means and proportions will also be calculated. To have a better understanding of the relationships among different study measurements, parametric or non-parametric correlation coefficients, bivariate, and multidimensional cross-tabulations, scatter and box-whisker plots will be constructed. Prior to analyses, baseline differences of key covariates from the PC and HBT arm will be examined with respect to our primary and secondary outcomes, particularly those related to race, ethnicity, SES and education. We will also examine differences on key variables between subjects who completed the study and those lost to follow-up. Our primary analysis will consist of chi-squared testing for the differences in proportions among the PC versus HBT study arms. Several univariate and multivariate logistic regression models will be used to calculate odds ratios and corresponding 95% confidence intervals for exploring the result of the screening and the relation to various types of key study variables in addition to study arms. Two-way interaction terms between for various study variables and the study arms may need to be included in the models. Transformations of the continuous data in order to meet statistical assumptions may be undertaken if indicated. Type-1 error will be set to 5% (α = 0.05) for calculating confidence intervals and testing various hypotheses for the primary outcome. Adjustment of type-I error for multiple comparisons will be performed by controlling the family-wise type-I error rate at the 0.05 level. Standard diagnostic tools will be used to assess any model fit. All statistical analyses will be carried out using SAS® v9.4 for Windows (SAS Institute, Cary, NC, USA) and/or R-project for Windows.