Colorectal cancer (CRC) is the third most common cancer, and the second leading cause of cancer deaths, in the United States [1]. Unequivocal evidence demonstrates that guideline concordant screening decreases CRC incidence and mortality by 30–60% [2]. Numerous modalities are currently recommended for CRC screening in average risk adults aged 50–75 years, including colonoscopy every 10 years or a fecal occult blood test (FOBT) or fecal immunochemical test (FIT) within the past year [3–5]. However, current CRC screening rates are 63% across the United States [6], well below targets set by the National CRC Roundtable (80% by 2018) [7] and by Healthy People 2020 (70.5%) [8]. More striking are the consistent disparities in CRC screening in rural areas, among adults with low income, and in racial and ethnic minorities [6, 9–12]. Although improving, screening rates in these vulnerable populations may be 15%–30% lower than their non-rural, higher income, non-minority counterparts [13].

Multiple systematic reviews identify interventions that effectively increase CRC screening [14–16]. In 2016, the Community Preventive Services Task Force recommended the use of multicomponent interventions to increase screening for CRC. Multicomponent interventions combine two or more approaches to increase community demand (e.g., client reminders, small media), community access (e.g., reducing client costs), or provider delivery of screening services (e.g., provider reminders) or two or more approaches to reduce different structural barriers [17]. However, understanding how these interventions can be best implemented and in what populations (e.g., screening naïve versus experienced) and settings (e.g., rural versus urban settings, health system versus independent clinics) remains a neglected area of study [14, 18, 19].

Implementing interventions into routine care in clinic and community-based settings often involves the active engagement of multiple stakeholders and the adaptation of program elements to local contexts. In Oregon and elsewhere, primary care and health plan leaders are eager to identify, adapt, and implement interventions to improve CRC screening in order to achieve state performance benchmarks and to improve patient quality and experience of care. However, our work with primary care and health system partners found that stakeholders are interested not just in which interventions work to improve CRC screening, but which interventions would work best in their setting given specific patient populations, clinical settings, and community characteristics (see www.communityresearchalliance.org).

Therefore, we designed this systematic review to compare the effectiveness of interventions to improve fecal testing for CRC in clinic and community settings serving rural, low-income populations and their associated implementation strategies and contextual factors (see definitions and sources, Fig. 1) [17, 30, 32–37]. Determining how implementation strategies and contextual factors influence the uptake of interventions to increase fecal testing for CRC may help stakeholders identify the interventions best suited for use in their local settings.

We followed systematic review methods described in the Cochrane Handbook for Systematic Reviews of Interventions [25] and AHRQ [26]. The review was guided by a technical expert panel of primary care professionals, health system leaders, and academicians. Our protocol was registered with PROSPERO, the international prospective register of systematic reviews (registration number CRD42015019557). The review is reported in accordance with the PRISMA publication standards [27–29].

The combined literature searches initially yielded 4218 titles and abstracts, including 4203 from electronic database searches, and 15 from reference lists of systematic reviews and other relevant articles. As summarized in Fig. 2, we assessed 278 full-text articles for eligibility, of which 27 studies reported in 29 publications met inclusion criteria. We identified 20 RCTs (published in 21 articles [40–60], 2 non-randomized controlled trials [61, 62], 3 pre-post studies [63–65], 1 cohort study [66] and 1 feasibility study [67] that contained primary data relevant to the key questions. The descriptive characteristics and findings of the 27 included studies are found in Additional file 1: Appendix D, Table D1.

Studies occurred in primary care clinics (n = 20, 74.1%), communities (n = 5, 18.5%), or in both settings (n = 2, 7.4%). Over half of the studies used FOBT (n = 18, 66.6%); FIT was used in 8 studies (29.6%); one study used both FIT and FOBT (3.7%). As summarized in Additional file 1: Appendix D, Table D2, patient eligibility and the informed consent process varied across the studies. Of the 20 clinic-based studies, 5 (25%) did not report on the patient consent process, 8 (40%) received a waiver of informed consent, 4 (20%) utilized an informed consent process, 2 (10%) used opt-out, and 1 was conducted as a quality improvement study (5%). Six of the seven studies (85.7%) conducted in community or both clinic/community settings did not report on the patient consent process; the one remaining study required informed consent. Many of the included studies rated as high or unclear risk of bias, this was often related to insufficient detail in the methods as well as lack of blinding. Quality assessment details appear in Additional file 1: Appendix C, Tables C1 and C2.

We found that studies to promote fecal testing in rural and low-income populations used multicomponent interventions that were tested in varied combinations across multiple study arms. Over half of the studies (16/27, 59.3%) used an informed consent process or did not provide details on the consent process, leaving it difficult to interpret if the participating patients were “different” than the general population served by the clinic or in the community. The most frequently used interventions in clinics included client reminder or recall, small media, provision of pre-addressed stamped envelopes for FIT/FOBT return, and direct mail programs. Many studies also employed outreach using navigators. In clinic-based studies, the components that occurred most frequently among the highly effective and effective study arms were provision of kits by direct mail, use of a pre-addressed stamped envelope, client reminders, and provider ordered in-clinic distribution.

Few studies included an adequate description of the contextual factors and approaches used to implement the intervention needed for health system leaders, primary care providers, or researchers to replicate the study in their own settings or to determine if the intervention was a good “fit” for the local context. For example, only 25.9% (7/27) reported on baseline screening rates prior to study implementation and only three clinic-based studies (15.0%, 3/19) named the EHR system used. In relation to implementation factors, delivery of the interventions occurred by clinic/community staff, research team members, or both – yet over half of the studies did not report on the type or length of training used. Few studies indicated if new clinical staff were hired or additional resources were provided to support intervention implementation. Seven studies reported on intervention costs, yet methods were heterogeneous, estimates highly variable, and details inadequate for determining how much an intervention might cost in a target setting. Only six studies described monitoring intervention fidelity, which was often limited to an acknowledgement that some form of interaction between the study team and clinical staff occurred (e.g., meetings, supervision).

Our review describes the effectiveness of interventions to improve fecal testing in low-income and rural patients in clinic and community-based settings. As in the recent review on CRC interventions by the Community Preventive Services Task Force, we found that most interventions tested in these settings were multicomponent in nature (i.e., included two or more strategies) [17]. Strategies to increase community demand (i.e., client reminders) as well as to increase community access (i.e., direct mail, use of a pre-addressed stamped envelope, in-clinic distribution) were intervention components commonly found in highly effective/effective study arms tested in clinic settings. However, generating precise estimates of specific intervention components was difficult in part because of the limited information on baseline screening levels for participating clinics, different patient targets (e.g., prior FIT/FOBT versus screening naïve), and variation in the consent process (e.g., waived versus informed). Moreover, none of the articles included in our review explicitly compared implementation strategies or the impact of contextual factors on intervention effectiveness.

One key challenge identified by our review was that the articles frequently lacked information in relation to local contextual factors and implementation strategies that could be used to inform end users as to which interventions would work best in their setting given specific patient populations, clinical settings, and community characteristics – and how to implement them. In their recent review to develop a taxonomy for CRC screening promotion, Ritvo and colleagues emphasized the need to describe the engagement sponsor, population targeted, alternative screening tests, delivery methods, and support for test performance (EPADS) in future reporting [69]. Numerous recent articles suggest strategies and approaches that can improve reporting of intervention studies, including TIDieR (Template for Intervention Description and Replication [70]) or PARIHS (Promoting Action on Research Implementation in Health Services [71]). Moreover, recent work by Powell and colleagues identified 73 implementation strategies grouped in 9 categories; thus providing a common terminology that can be used to design and evaluate effectiveness as well as implementation research studies [32, 72]. However, as indicated in our review, this data is either not being gathered by research investigators or it is not being reported in publications.

We speculate that lack of reporting on the impact of implementation strategies and contextual factors may reflect funder preferences for novel discoveries, researcher bias toward intervention effectiveness measured by statistical significance rather than nuanced understanding of how, when, and why a specific intervention works, and journal restrictions on manuscript length. Our experience also suggests that pragmatic studies conducted in settings serving rural and vulnerable populations are often better equipped for implementing interventions owing to greater capacity for data reporting, workflow refinements, and leadership interest/engagement than would be needed to scale these interventions across average or “below average” settings. This presents challenges and opportunities for cancer control researchers and implementation scientists. Namely, it presents opportunities to compare the effectiveness of implementation strategies needed to translate effective multicomponent interventions into diverse clinic, community, and population settings; to use SMART designs to test stepped-implementation support in response to baseline setting capacity and intervention responsiveness over time; and to implement “layered” interventions which may first build clinic capacity and visit-based workflows prior to implementing population outreach programs [73].

We are seeing a shift to address these needs through funding opportunities through the Cancer Moonshot, in NCI and CDC’s continued support of collaborative research networks and implementation research, as well as the emergence of local initiatives in response to advanced payment models that reward improved CRC screening and disparities reduction [74, 75]. Taking the next step forward in implementation of evidence-based approaches to improve cancer screening also argues for investing in academic-community collaborations before research studies begin through the infrastructure of practice-based research networks [76], researcher in residence models [77], or participatory research methods [78, 79]. Such partnered approaches can help actualize the call for rapid and relevant science [80]. Shifting the research paradigm to support collaborative, partnered implementation enables leveraging “teachable moments” and “tipping points” when research evidence can be used to inform local, regional, or system-wide interventions that are under consideration or actively underway.

There are a few important limitations in the present study. First, we limited our review to studies targeting rural and low-income patients in the United States. These populations experience CRC screening disparities, and our stakeholder partners wanted to know what interventions worked best for settings serving these patients. It is also possible that interventions that are effective for these hard-to-reach populations need to be different or more intensive than those implemented in higher resourced populations. Second, we acknowledge that the requirement to have quantitative outcome data on changes in FIT/FOBT may have led to the exclusion of studies targeting any type of CRC screening or to studies focused on qualitative methods. Future systematic reviews on this topic would benefit from interviews with the lead or senior author or a realist review [81–83] which could generate a more contextualized understanding of how and why complex interventions achieve particular effects in particular contexts. Finally, we created cut-offs for evaluating the individual intervention components based on a proportional distribution of the outcomes. This approach identified four intervention components that commonly appeared in the highly effective/effective study arms in clinic based studies (i.e., direct mail, client reminders, addressed stamped envelopes, in-clinic distribution); the use of alternative cut-points may have identified additional strategies. Despite these limitations, our findings document the general deficiency in reporting on contextual factors, implementation strategies, and how these factors interact with the intervention. Our review provides a critical starting point for future reviews and informs research funding and publication in this area.

There is growing interest in the conduct and reporting of pragmatic research that can be used by stakeholders to inform practice in primary care and community settings. Such evidence can guide health systems administrators in selecting an appropriate intervention in their setting, as well as approaches for adapting a given intervention or implementation approach based on contextual factors. Interventions to improve FIT/FOBT for CRC screening should be considered generally as complex interventions (e.g., multicomponent, context-sensitive, and highly dependent on the behaviors of participants and providers) [84]. As the field of pragmatic research continues to advance, our findings suggest that the use of standard guidelines for the reporting of implementation strategies and contextual factors within trials may be useful. Building on the policy of the National Institute of Health (NIH) to require detailed on rigor and reproducibility in grant submissions [85], journals, editors, and funders could mandate scientists to more freely share protocols, implementation toolkits, or intervention websites. Tools used in our review may inform such guidelines, including more routine application of Proctor, Powell and colleague’s recommendations for tracking and reporting implementation strategies [32, 34, 72] or Stange and colleague’s conceptualization of contextual factors [35–37]. Rather than reducing the nuances of implementation in diverse, complex adaptive systems to soundbites, we need to apply the principles of precision medicine to research design to determine what works for whom and why, rather than what works after variation has been removed. Funders and journal editors should apply these policies so that investigation of how intervention effects are modified by context and implementation strategies moves from being a “new methodological frontier” to a standard of practice for research [86].

Multicomponent interventions can effectively increase fecal testing for CRC across diverse rural and low-income communities. Few community based studies were effective, but this may be because they targeted patients irrespective of CRC screening eligibility. For clinic-based studies, the intervention components frequently found in highly effective and effective treatment arms included mailing FIT/FOBT materials directly to patient’s homes (i.e., “direct mail programs”), provision of a pre-addressed stamped envelope to facilitate kit return, client reminder and recall, and provider ordered in-clinic distribution. Based on the community guide, these strategies are designed to improve community access by reducing structural barriers and to increase community demand. Few studies described contextual factors or implementation strategies in the detail needed for stakeholders to clearly determine how and which interventions would work in their local contexts. More robust application of guidelines to support reporting of implementation strategies and contextual factors in funding and publication are needed so that stakeholders can determine not just which interventions work to improve CRC screening, but which interventions would work best in their setting given specific patient populations, clinical settings, and community characteristics.