Rural-Urban Disparities in Colorectal Cancer Screening, Diagnosis, Treatment, and Survivorship Care: A Systematic Review and Meta-Analysis

Abstract Background Rural residents have a higher prevalence of colorectal cancer (CRC) mortality compared to urban individuals. Policies have been aimed at improving access to CRC screening to reduce these outcomes. However, little attention has been paid to other determinants of CRC-related outcomes, such as stage at diagnosis, treatment, or survivorship care. The main objective of this analysis was to evaluate literature describing differences in CRC screening, stage at diagnosis, treatment, and survivorship care between rural and urban individuals. Materials and Methods We conducted a systematic review of electronic databases using a combination of MeSH and free-text search terms related to CRC screening, stage at diagnosis, treatment, survivorship care, and rurality. We identified 921 studies, of which 39 were included. We assessed methodological quality using the ROBINS-E tool and summarized findings descriptively. A meta-analysis was performed of studies evaluating CRC screening using a random-effects model. Results Seventeen studies reported disparities between urban and rural populations in CRC screening, 12 on treatment disparities, and 8 on staging disparities. We found that rural individuals were significantly less likely to report any type of screening at any time period (pooled odds ratio = 0.81, 95% CI, 0.76-0.86). Results were inconclusive for disparities in staging at diagnosis and treatment. One study reported a lower likelihood of use of CRC survivorship care for rural individuals compared to urban individuals. Conclusion There remains an urgent need to evaluate and address CRC disparities in rural areas. Investigators should focus future work on assessing the quality of staging at diagnosis, treatment, and survivorship care in rural areas.


Introduction
Approximately 14% of the US population lives in a rural area. 1 Individuals living in rural areas are often subject to poorer health outcomes compared to their urban counterparts.For example, the age-adjusted death rate in rural areas is 7% higher than that of urban areas. 2 These excess deaths have been linked to a higher prevalence of chronic conditions such as heart disease, chronic lower respiratory disease, and cancer. 3][4][5][6][7][8] Moreover, less than 8% of all physicians and surgeons in the US practice in rural settings and 64% of rural medical staff report difficulties in finding specialists for patient referrals. 9,10 higher incidence of cancer-related cases and mortality are among the most impactful causes of excess deaths among e432 The Oncologist, 2024, Vol. 29, No. 4   rural populations as a result of these health disparities. 11,12hile national cancer incidence has declined over time, the incidence and mortality rates of certain cancers have persisted in rural areas, such as those for colorectal cancer (CRC). 13n 2016, the annual age-adjusted death rate due to CRC in nonmetropolitan areas was estimated at 17.1 deaths per 100 000 persons, higher than the estimated rate in metropolitan areas (14.0 deaths per 100 000 persons). 13This trend was also noted for incident CRC cases, where the ageadjusted incident CRC rate among nonmetropolitan residents was higher than that of metropolitan residents (43.9 cases per 100 000 persons vs 39.6 cases per 100 000 persons, respectively). 13Of the hypothesized driving factors behind these outcome disparities in incidence and mortality, those involving the patient's provider and their care along the CRC care continuum may be particularly impactful. 14atients in rural areas experience limited access to medical and oncology providers and higher-quality care, which may lead to later stages of any cancer diagnosis and a lower likelihood of receiving standard-of-care treatment with supportive care. 12,15Therefore, inequities in each of these stages (CRC screening, stage at diagnosis, treatment, and survivorship care) of patients with CRC care may contribute to worsened downstream outcomes.Limited access to CRC screening among rural individuals has been previously characterized. 16A systematic review from Wang et al found that the most frequently reported barriers for rural individuals to CRC screening were high cost, lack of insurance coverage, and lack of perceived need for CRC screening. 16owever, it is unknown to what extent these barriers in screening, in addition to barriers in treatment, stage at diagnosis, or survivorship care differ from urban individuals.To better guide policymakers, decision makers, and healthcare providers in addressing the impact of these hypothesized CRC disparities on cancer-related incidence and mortality for rural patients, it is essential to understand how each of these categories independently contributes to patient with CRC care and subsequent outcomes.Moreover, it is critical to understand how various existing definitions of ruralurban status can affect analyses evaluating CRC care disparities, as different rural-urban categorizations can lead to a variation in results for studies evaluating disparities or barriers in access to care. 17To better understand these urban-rural CRC disparities, we performed a systematic literature review to evaluate original research investigating differences in CRC screening, stage at diagnosis, CRC treatment, and survivorship care between rural and urban adults in the US.Secondary to this, we performed a metaanalysis to synthesize findings across these gaps where sufficient data were available.

Materials and Methods
We conducted a systematic review to assess differences in CRC screening, stage at CRC diagnosis, treatment, and survivorship care between rural and urban adults in the US in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols (PRISMA-P). 18This study was exempt from IRB approval and informed consent as it collected and synthesized nonidentifiable data from previously published studies.The protocol for this study was approved and made available on PROSPERO (ID #CRD42022350943).

Data Sources, Search Strategy, and Inclusion/ Exclusion Criteria
The review was conducted using PubMed and EMBASE for primary sources of evidence published between January 2012 and July 2022.From our original search, we identified and excluded duplicate indices and screened abstracts for inclusion criteria (Supplementary Materials).Those qualifying for inclusion were further screened using the full-text source and once again screened for final inclusion/exclusion criteria.We used a "snowball" approach in reviewing studies to ensure that references of included studies that also qualified for inclusion were not missed in the original search strategies.We used search terms related to rural populations, colorectal cancer, screening/prevention, treatment, diagnosis, and survivorship care and combined several databasespecific search terms, such as Medical Subject Heading (MeSH) terms and free-text search terms.The full search strategy may be found in the Supplementary Materials.We specified eligibility criteria using the PICOT framework (population, intervention, comparator, outcome, time frame), summarized in the Supplementary Materials.We included randomized controlled trials, observational cohort studies, and case-control studies in English evaluating rural vs urban individuals as either a primary or secondary outcome in any of the 4 "disparity categories" (screening, stage at diagnosis, treatment, and survivorship care).We excluded publications that were other systematic reviews or meta-analyses, guidelines, letters to the editor, case studies, ethnographic or qualitative studies, and surveys.We did not exclude studies based on a specific type of definition or categorization method used to distinguish rural-urban individuals nor on any specific age group.We included studies that reported incidence rates for rural and urban populations and studies that only reported regression output(s) as a method of assessment.CRC screenings of interest included fecal-occult blood testing (FOBT), fecal immunochemical testing (FIT), flexible sigmoidoscopies, colonoscopies, CT colonographies, and barium enema exams.Studies evaluating disparities in stage at diagnosis must have reported data on the type of staging definition used (SEER staging classification system, American Joint Committee on Cancer 7th Edition Staging, etc.).Those assessing treatment disparities must have reported data on the type of treatments evaluated.Treatments of interest included surgical resection (laparoscopic colectomy, open resection), chemotherapy, and radiotherapy.For studies that also included data on other types of cancers, we extracted results for CRC only.

Screening, Study Design, and Data Abstraction
The original search, screening, and abstraction were performed using Covidence (Melbourne, AUS).Three investigators (A.S., M.N., L.C.) screened initial database search results (title and abstract) for studies adherent to the predetermined PICOT criteria.Indices screened and included were screened once more using the publication full text to ensure final study inclusion.Data were abstracted from each full-text publication and included the following: first author, year of publication, study design, geographic location, years of data evaluated, age group, inclusion criteria, rural-urban classification and categorizations, data source, screening types evaluated, staging definition and categories, treatment types evaluated, and results of primary/secondary outcomes.To assess for study bias, 2 investigators (A.S., M.N.) independently reviewed e433 each full-text article using the ROBINS-E tool. 19At each stage, any disagreements between reviewers were resolved by discussion, and, if necessary, adjudicated by a fourth reviewer.

Meta-Analysis of CRC Screening
Data for all outcomes except CRC screening were not sufficient to permit meta-analysis.Therefore, we limited the meta-analysis to studies reporting differences in CRC screening.The primary outcome was the odds ratio (OR) comparing rural and urban populations for any screening method reported at any time.All studies were weighted based on the generic inverse-variance method. 20A random-effects model was developed and built on the assumption that betweenstudy variance results from factors other than measured treatment differences. 21The random-effects model assumes a normal distribution of between-study variance, which is facilitated by the generally large sample sizes (>100) of the included studies. 21All analyses were performed using the "meta" package in R with a significance level of 0.05. 22The code for this analysis is publicly available (https://tinyurl.com/3kknzasx).

Assessment of Heterogeneity and Publication Bias
Quantifiable heterogeneity between studies reporting CRC screening differences was evaluated using the I 2 statistic.The I 2 statistic developed by Higgins et al describes the percent of total variation across studies attributable to heterogeneity beyond random chance. 20Generally, values of 0% indicate no heterogeneity, and 25%, 50%, and 75% indicate low, moderate, or high heterogeneity, respectively. 20We assumed an acceptable I 2 value of 50% or less. 23In the event of an I 2 value > 50%, we specified a priori methods on outlier assessment and removal.We used methods from Viechtbauer and Cheung to identify and remove outlier studies with effect sizes outside of the 95% CI of the original pooled result. 24ublication bias was evaluated using Peter's test and funnel plots, which evaluate the relationship between the effect size of each study and its precision. 25If there is a detected systematic relationship between effect size and precision, publication bias may be present.Pooled results were reported for without outliers and with outliers as a pooled odds ratio with 95% CI in a forest plot.
]31,38,50 Two studies used the Office of Management and Budget (OMB) criteria, 1 study used CDC rural-urban definitions, and 2 studies used Metropolitan Statistical Area (MSA) classifications to define rural-urban individuals. 31,36,41,42,60One study used the Medical Service Study Area (MSSA) classification scheme to categorize individuals and later used RUCA codes as a sensitivity analysis. 53One study also tested different ruralurban classification criteria (OMB, RUCC, UIC, National Center for Health Statistics definition) as a primary analysis. 31[37][38][39][40]42,64 Risk of bias assessments revealed a mostly overall high risk of bias for included studies (Supplementary Figs.S2, S3).This was primarily driven by a high risk of bias due to missing data, exposure measurement, and confounding.

Studies Reporting CRC Treatment Disparities
8][59][60][61] Four studies    examined differences in chemotherapy or radiation administration between rural-urban patients with CRC. 55,56,62An analysis of rural individuals with stage III CRC reported higher rates of no adjuvant chemotherapy, adjuvant 5-FU only, and lower rates of adjuvant oxaliplatin compared to urban individuals. 55Neoadjuvant chemoradiotherapy rates were similar across urban and rural areas, with CRC stage II-III rural individuals having the highest odds of treatment. 56ariation was observed in pre-and post-operative neoadjuvant radiation rates between urban and rural hospitals treating patients with stage I-IV CRC, with rural hospitals having a non-significant lower rate of post-operative radiation use (23.0%urban vs 3.5% rural, P = .08). 54Compliance with high-quality measures, such as lymph node removal and adjuvant chemotherapy, was lower in rural patients with stage II colon cancer and stages II-III rectal cancer. 53Overall, rural residents had lower odds of receiving adjuvant chemotherapy compared to urban residents. 44,56,61,62udies Reporting CRC Survivorship Care Disparities One study assessed rural-urban disparities in CRC survivorship care (168 urban individuals, 109 rural individuals). 63cDougall et al evaluated adherence to surveillance colonoscopy in 30-74 year old CRC survivors at 1, 3, 5, and 10 years post-diagnosis using New Mexico Cancer Registry data. 63ompared to urban survivors, rural survivors were 2.28 times (95% CI, 1.07-4.85)more likely to report nonadherence to surveillance colonoscopy guidelines. 63Moreover, financial hardship was independently associated with nonadherence to surveillance colonoscopies (OR = 2.17, 95% CI, 1.01-4.5). 63

Discussion
In this analysis, we found substantial data supporting a disparity in CRC screening between urban and rural CRC individuals, minimal evidence supporting poorer standard of care treatment for rural patients with CRC, mixed evidence suggesting that rural patients with CRC had a higher stage at diagnosis, and only one study evaluating survivorship care disparities.Our results for CRC screening were consistent with previously published data.One included study directly assessed the effects of limited access to care due to cost by describing improved CRC screening rates after introduction of the Affordable Care Act (ACA), which improved access to CRC screening by eliminating financial barriers. 40However, the persistence of this gap in screening over time suggests that non-financial factors may still exist to limit rural utilization of appropriate CRC screening.Our mixed results with respect to disparities in stage at diagnosis largely differ from other studies evaluating this rural-urban disparity for other types of cancers. 47,65,66This may have been due to a variety of reasons, but given the variation in time periods assessed among the included studies evaluating the CRC stage at diagnosis, we note that it is challenging to account for the long latency period between environmental exposures (rural-urban status) and cancer diagnosis.Furthermore, we did not include a specific definition of "stage" at diagnosis for this review, leading to the inclusion of studies with various staging definitions used which may have contributed to our mixed results.We found, paradoxically, that research to date suggests a disparity among rural and urban patients with respect to CRC screening, but the distribution of stages at diagnosis was very similar between the 2. While there may be several explanations for this observation, one hypothesis stems from a large amount of missing staging/registry data for patients who tend to do very poorly, such as rural patients with cancer. 67This may have biased results in staging studies more favorably for rural patients with CRC.9][70][71] Finally, the single study evaluating CRC survivorship care differences noted a disparity in access to and adherence to survivorship CRC surveillance care. 63As patients with CRC live longer lives, it is critical to understand and measure how disparities noted upstream in the cancer care continuum may carry over to survivorship care and affect clinical outcomes.Given the lack of data in this area, researchers may want to focus more e442 The Oncologist, 2024, Vol. 29, No. 4 on disparities of the quality survivorship care received between rural and urban CRC survivors.
Underlying mechanisms driving the rural-urban disparities we observed in this study may be better understood through a framework that considers factors such as low socioeconomic status, lack of insurance coverage, limited healthcare access, and transportation issues. 14These factors in turn contribute toward inadequate surveillance, late-stage diagnosis, and lower quality of cancer treatment. 14Qualitative research has expanded on some of these underlying mechanisms.For example, work from Lee et al describes distance to travel in rural areas as a consistently reported barrier toward achieving appropriate CRC screening. 72Others have reported clear barriers for rural patients in receiving timely specialist care after diagnosis, and a lack of clear communication between providers and patients, which may lead to the poorer care received that we observed in this analysis. 73,74A series of structured interviews among rural Nebraskan patients with CRC also suggests that providers may not even discuss alternative screening methods unless rural patients with CRC resist colonoscopies. 75Ultimately, the multitude of rural barriers to appropriate receipt of CRC screening, early stage at diagnosis, standard of care treatment, and survivorship care are complex and interdependent and involve patients, providers, and community-level effects.Therefore, a multifactorial and cross-functional effort between patients, providers, local decision makers, and national policymakers is essential for addressing these underlying mechanisms.For example, national and state-level policymakers may advocate for the expansion of CRC screening and treatment services covered under Medicaid, which in turn may increase the timely use of screening and treatment for patients with lower socioeconomic status.These policymakers may also opt to increase financial support for local rural providers to coordinate with other urban cancer centers, such as through the National Cancer Institute Community Oncology Research Program.Finally, national funding bodies may consider pushing to incentivize rural recruitment into CRC-related clinical trials, potentially improving access to timely care.
A notable trend observed in this review was the wide variety of used rural-urban definitions.This definition variability severely limited our synthesis of results.While federal bodies such as the National Institutes of Health (NIH) have often called for harmonization of person-level and contextual variables for research purposes, there is currently no call for standardization of the rural-urban definition.While this may be for several reasons, we recognize the difficulty in utilizing only one standard definition for rural-urban status.For example, certain definitions, such as those used by the US Census Bureau or the OMB, focus exclusively on population density as a means of defining rural-urban categories, while others use more specific criteria that incorporate other contextual factors, such as proximity to urban areas or commuting distance. 76,77Moreover, certain definitions are confined to the US census-tract level, which may change over time and are large enough such that they may contain both "rural" and "urban" areas.Others utilize county-level estimates (eg, OMB definition, RUCC codes). 77,78Finally, certain taxonomies can further subdivide areas past rural or urban status, which may improve the precision of certain estimates, but comes with the risk of inadequate sample size.We note that the use of varying definitions of rural-urban status is also frequently a product of data availability.For example, SEER data, curated by the National Cancer Institute (NCI), only includes spatial data using RUCA codes, RUCC codes, or MSAs.RUCA and RUCC codes, in particular, were fairly prominent in our review. 79This may be because RUCA and RUCC codes combine population density, proximity to urban areas, and other factors. 78,80While RUCA codes are measured at the census tract level, RUCC codes are smaller subdivisions of county areas. 78,80Because county areas are smaller than census tracts, differences in results may be observed when comparing studies that use these different definitions.Due to these nuances in how one may define "rural-urban" status, there naturally exists disagreement as to which definition is better suited for analysis.The consequences of a lack of consensus surrounding best practices for rural-urban criteria are evident when attempting to synthesize published data, as various definitions can introduce bias.For example, Hao et al and Hines et al reported similar rates of therapy between urban and rural patients with CRC, while Panchal et al reported much lower rates among rural individuals.While there may be several reasons for this observation, it is worth noting that Hao et al and Hines et al used the RUCA criteria for rural-urban status, which allowed for the inclusion of smaller towns with local commuting to urban areas as "rural."In contrast, Panchal et al used a stricter definition of rural using RUCC codes, limiting "rural" to areas with <2500 residents.Analyses utilizing rural-urban criteria that allow for urban-adjacent areas to qualify as "rural" may bias results toward a null hypothesis.Therefore, careful attention must also be paid to studies that report no significant difference between rural and urban receipt of CRC-related care to ensure that bias from the use of various rural-urban definitions is limited.
Outside of the challenges involving rural-urban definition, analytic considerations may also pose a challenge in the analysis of rural-urban CRC disparities.Most studies in our review used regression methods where rural status was considered as a person-level variable, instead of a contextual one.A few studies used more advanced 2-level hierarchical models where the rural-urban definition was used as a nesting level, allowing investigators to account for the effect of either the same residential or hospital location among different patients CRC.While this is a reasonable approach, one may also consider the use of cross-classified multilevel models to simultaneously account for nesting in both residential areas and hospital care areas. 81For example, patients with CRC may be nested within the same geographic location, but seek care in different locations.Cross-classified multilevel models provide a unique method to model the diverse settings that may affect CRC care and outcomes, yet are underused in spatial research. 81Alternatively, investigators may also consider the use of 2-step floating catchment area (2SFCA) methods to aid in the determination of access to healthcare services. 82The 2SFCA method is particularly insightful, as it not only considers the distance to care but also how accessible a provider is to a patient given the surrounding population density. 824][85] Overall, while a variety of methods exist to quantify spatial disparities in CRC screening and care, future investigators may want to consider the use of different, more robust methods of analysis.
The Oncologist, 2024, Vol. 29, No. 4   e443   This had several limitations to acknowledge, such as the heterogeneity of the studies included and assessed with regard to rural-urban definition, definition of "screening," definition of "treatment," and definition of "staging."While we were able to perform a meta-analysis of studies with adequate statistical heterogeneity, the studies included vary widely in study design, cohort definition, and outcome definitions.Therefore, readers should cautiously interpret our meta-analysis results.Similarly, while statistically we demonstrated low publication bias for CRC screening studies, the same could not be said for studies evaluating the CRC stage at diagnosis or treatment.We also note that differences in rural-urban receipt of CRC treatment noted for this study may have been biased by each patient's health status at baseline.For example, rural individuals may be "sicker" at baseline, leading to a difference in the type of CRC treatment received as opposed to urban individuals.We only found one study evaluating CRC survivorship care, limiting the applicability of our findings in this area.The studies included in this review were all retrospective, as no prospective studies were found.While retrospective studies confer specific strengths, an amount of selection bias is inherent to these studies, limiting results.Finally, as this was primarily a descriptive analysis in scope, we were unable to account for the effect of time on our results, which may be relevant as the studies included in this review varied widely.While it does not appear that the time period evaluated had a relationship to the disparities we noted here, important interventions such as policy (eg, Affordable Care Act) and changes in guidelines may have affected the results noted in screening, stage at diagnosis, and treatment.

Conclusions
Our findings suggest the presence of rural disparities in equitable access to CRC screening.Investigators should focus on understanding the driving factors behind our noted disparities in CRC screening, such as inequity in socioeconomic status, distance and access to appropriate care, and appropriate healthcare coverage.Given the mixed findings for rural disparities in CRC staging at diagnosis and treatment, it may be that the largest barrier rural individuals face may simply be at the initial screening level.Decision-makers should focus their policy efforts on improving this initial access to care for rural individuals.Future investigators should also weigh the benefits and limitations of using certain definitions for rurality and may want to consider the use of more robust methods suited for spatial analysis to better understand these disparities.

Figure 1 .
Figure 1.PRISMA flow diagram for screening article selection and evaluation.
not report definitions.Abbreviations: FAR, frontier and remote; MSA, Metropolitan Statistical Area Codes; MSSA, medical service study area; NR: none reported; RUCA, rural-urban commuting area code; RUCC, rural-urban continuum code; UIC, urban influence code.

Table 1 .
Study population age, rural-urban classification, urban, and rural categorization.

Table 2 .
Colorectal cancer screening study characteristics.

Table 3 .
Colorectal cancer staging study characteristics.

Table 4 .
Colorectal cancer treatment study characteristics.