County-level social, preventive care, and behavioral correlates of colorectal cancer mortality in the United States: a nationwide ecological analysis

Abstract

Background:

Geographic disparities in colorectal cancer (CRC) outcomes remain substantial across U.S. counties. Associations of social conditions, preventive care, behavioral risk environments, and rurality with CRC mortality remain incompletely characterized in analyses addressing state-level clustering and spatial dependence.

Methods:

We linked CDC PLACES, American Community Survey, United States Cancer Statistics, and 2023 National Center for Health Statistics rurality data in a nationwide county-level ecological study. Outcomes were age-adjusted CRC mortality and all-cancer mortality during 2019–2023 and CRC incidence during 2018–2022. Population-weighted core-adjusted linear models evaluated each exposure separately, adjusting for the county proportion aged ≥65 years, nonmetropolitan status where applicable, and state fixed effects. Standard errors were clustered by state using CR2 estimation with Satterthwaite degrees of freedom.

Results:

The ACS-complete dataset included 3,142 counties; 3,134 remained after linkage to the 2023 NCHS rurality classification. Final fixed samples included 2,126 counties for CRC mortality, 3,005 for all-cancer mortality, and 2,628 for CRC incidence. In core-adjusted CRC mortality models, greater broadband deprivation (β = 1.82, 95% CI 1.44 to 2.20), higher current smoking prevalence (β =2.08, 95% CI 1.74 to 2.42), and nonmetropolitan status (β=2.73, 95% CI 2.35 to 3.11) were associated with higher mortality. Higher CRC screening uptake (β = −1.64, 95% CI −1.97 to −1.31) and recent dental visit prevalence (β = −1.84, 95% CI −2.27 to −1.40) were associated with lower mortality. Key associations were directionally consistent in log-rate and matched spatial error model sensitivity analyses, with similar patterns for all-cancer mortality and CRC incidence.

Conclusions:

CRC mortality was associated with adverse social and behavioral profiles, lower preventive care engagement, and nonmetropolitan residence. Overlap across CRC mortality, all-cancer mortality, and CRC incidence suggests that these indicators may reflect broader geographic cancer disadvantage rather than CRC-specific mechanisms. Integrated structural, preventive care, behavioral, and rural health strategies may help reduce county-level disparities in CRC outcomes.

1 Introduction

Colorectal cancer (CRC) remains a major contributor to cancer burden in the United States, yet substantial geographic disparities in CRC outcomes persist across counties (1, 2). Although population-level declines in CRC incidence and mortality have been observed over time, these improvements have not been evenly distributed, and some counties continue to experience disproportionately high CRC burden (3, 4). Understanding the contextual factors associated with these county-level disparities is important for identifying high-burden communities and informing geographically targeted cancer control strategies (5, 6).

County-level disparities in CRC outcomes likely reflect the combined influence of structural disadvantage, access to preventive care, local health behavior environments, and rurality (7). Traditional social determinants of health, including poverty, unemployment, and lower educational attainment, have been associated with lower cancer screening uptake and reduced access to timely care (8, 9). At the same time, preventive and follow-up care increasingly depends on digital infrastructure, including patient portals, telehealth communication, appointment coordination, access to health information, and navigation of multistep screening pathways (10). Broadband access may therefore serve as a contextual marker of healthcare connectivity and care navigation capacity in contemporary CRC prevention and management (11). However, its county-level association with CRC outcomes, together with correlated socioeconomic, preventive care, behavioral, and rurality indicators, remains incompletely characterized (12).

It is also important to determine whether county-level correlates of CRC mortality overlap with broader cancer mortality and CRC incidence patterns. Comparing CRC mortality with all-cancer mortality can help characterize whether observed associations reflect broader geographic cancer disadvantage, whereas examining CRC incidence can indicate whether similar contextual patterns extend across the CRC disease continuum (13–17). These comparisons should be interpreted as descriptive ecological contrasts rather than tests of disease-specific causal mechanisms.

In this nationwide county-level ecological study, we linked county-level data from CDC PLACES, the American Community Survey, United States Cancer Statistics, and the 2023 National Center for Health Statistics urban-rural classification scheme. Our objectives were to: (1) characterize county-level variation in CRC mortality and describe differences between high- and low-burden counties; (2) estimate core-adjusted associations of social, preventive care, behavioral, and rurality indicators with CRC mortality; and (3) compare key associations across CRC mortality, all-cancer mortality, and CRC incidence. We hypothesized that adverse social and behavioral profiles, lower preventive care engagement, and nonmetropolitan county status would be associated with higher county-level CRC burden.

2 Methods

2.1 Study design and data sources

We conducted a nationwide county-level ecological study of U.S. counties using linked public datasets. The county-level analytic base was constructed from the 2023 CDC PLACES county GIS-friendly release and the American Community Survey (ACS) 2017–2021 county-level non-medical factor measures dataset (18, 19). During preprocessing, variable names were standardized, county identifiers were harmonized, and ACS measures were reshaped from long to wide format before linkage with PLACES data by county Federal Information Processing Standards code. Counties with complete ACS-derived measures were retained for the initial merged dataset.

Rurality was classified using the 2023 National Center for Health Statistics Urban-Rural Classification Scheme for Counties. Counties classified in categories 1–4 were considered metropolitan, whereas counties classified in categories 5–6 were considered nonmetropolitan. Eight counties with legacy Connecticut county FIPS codes did not have direct matches in the 2023 NCHS classification file and were excluded before construction of outcome-specific analytic samples (20).

Cancer outcomes were derived from county-level United States Cancer Statistics ASCII files. Records were restricted to county-level observations for all racial/ethnic groups and both sexes. For all-cancer mortality, records were restricted to overall cancer mortality across all sites. For CRC outcomes, records were restricted to malignancies of the colon and rectum and subsequently separated into incidence and mortality datasets. Suppressed or unavailable age-adjusted rates were treated as missing. County FIPS codes were extracted from the USCS area field before linkage with the PLACES/ACS/NCHS analytic dataset (21).

This study used publicly available, de-identified county-level data and did not involve direct participant contact. Therefore, institutional review board approval and informed consent were not required.

2.2 Outcomes

The primary outcome was county-level age-adjusted CRC mortality for 2019–2023 combined. The secondary outcome was county-level age-adjusted all-cancer mortality for 2019–2023 combined. The exploratory outcome was county-level age-adjusted CRC incidence for 2018–2022 combined. Outcome-specific analytic datasets were created after merging these USCS outcomes with the county-level PLACES/ACS/NCHS analytic base. The outcome windows differed because publicly available county-level USCS files were released for different periods. CRC incidence was assessed during 2018–2022, whereas mortality outcomes were assessed during 2019–2023. Each outcome was analyzed separately in an outcome-specific analytic sample, and cross-outcome comparisons were interpreted as descriptive ecological contrasts rather than formal tests of disease-specific differences.

2.3 County-level correlates

County-level explanatory variables were grouped a priori into three domains (13, 22). The social determinants domain included poverty below 150% of the federal poverty level, no high school diploma, unemployment, high housing cost burden, crowded housing, single-parent households, racial/ethnic minority composition, and lack of broadband internet access. The healthcare access and preventive care domain included lack of health insurance, routine checkup, recent dental visit, and up-to-date CRC screening. The behavioral risk indicator domain included obesity, current smoking, binge drinking, short sleep duration, and physical inactivity. Continuous exposure variables were standardized to z scores. This standardization facilitated comparison of association magnitudes across county-level indicators with different distributions and variability, such that regression coefficients represented the absolute difference in the county-level outcome associated with a 1-standard-deviation higher prevalence of the corresponding exposure. The proportion of county residents aged ≥65 years was included as a demographic covariate. Rurality was represented by nonmetropolitan versus metropolitan county status.

2.4 Study sample and descriptive analyses

The ACS-complete merged dataset initially included 3,142 counties. After exclusion of eight counties without direct matches in the 2023 NCHS rurality classification, the county analytic base included 3,134 counties. Outcome-specific samples were then defined according to availability of each cancer outcome. CRC mortality was available for 2,188 counties, all-cancer mortality for 3,072 counties, and CRC incidence for 2,694 counties.

To ensure comparability of estimates across individual-exposure models within each outcome, we constructed outcome-specific fixed complete-case analytic samples using all variables required by the planned analytic framework. Within each outcome, the same fixed complete-case sample, defined by complete data on all candidate exposure variables and core covariates, was used for every individual-exposure model. The final fixed analytic samples included 2,126 counties for CRC mortality, 3,005 counties for all-cancer mortality, and 2,628 counties for CRC incidence. The study flow and outcome-specific exclusions are shown in Figure 1.

Study flow diagram. The ACS-complete merged county dataset included 3,142 counties. Eight counties with legacy Connecticut county FIPS codes lacked direct matches in the 2023 National Center for Health Statistics rurality classification and were excluded, leaving 3,134 counties in the county analytic base. Outcome-specific exclusions reflected unavailable or suppressed cancer outcome data and missing variables required for the fixed analytic samples. Final samples included 2,126 counties for CRC mortality, 3,005 counties for all-cancer mortality, and 2,628 counties for CRC incidence.

County-level characteristics were summarized across quartiles of CRC mortality burden in the final primary analytic sample. Geographic distributions of CRC mortality, all-cancer mortality, and CRC incidence were visualized using county-level maps.

2.5 Statistical analysis

The primary analyses used population-weighted linear regression models with state fixed effects. Because the outcomes were county-level age-adjusted rates rather than individual-level event counts, coefficients were interpreted as absolute differences in county-level rates per 100,000 population. County population was used as the analytic weight so that associations were estimated with greater influence from counties contributing larger population denominators. State fixed effects were included to account for time-invariant state-level differences in cancer control policies, healthcare systems, data-reporting context, and other unmeasured state characteristics that could confound cross-county associations. Population-weighted linear regression was selected because the publicly available outcomes were county-level age-adjusted rates rather than individual-level data or raw event counts analyzed with population offsets; this approach provides directly interpretable absolute differences in rates per 100,000 population. Poisson and negative binomial models were therefore not used as primary models. State fixed effects were preferred to random-effects multilevel models because the aim was to estimate within-state county-level associations while controlling for unmeasured, time-invariant state context without assuming that state effects were independent of county-level exposures. Geographically weighted regression was not used as the primary approach because the objective was to estimate national average associations rather than location-specific coefficients. Residual spatial dependence was evaluated using Moran’s I and matched spatial error model sensitivity analyses.

For each outcome, each focal exposure was evaluated in a separate core-adjusted individual-exposure model. Continuous exposure variables were standardized to 1-standard-deviation increments. Primary models adjusted for the county proportion aged ≥65 years, nonmetropolitan county status, and state fixed effects. For the rurality model, nonmetropolitan county status was the focal exposure; therefore, the model adjusted for age and state fixed effects but did not self-adjust for rurality.

Standard errors were clustered at the state level using the CR2 variance estimator, and inference was based on Satterthwaite degrees of freedom (23, 24). Multicollinearity was assessed using variance inflation factors for focal exposure terms in the core-adjusted models. Supplementary sequential conditional-association models were estimated for CRC mortality to transparently describe changes in estimates across nested domain-block specifications. Model 1 included social determinants, the county proportion aged ≥65 years, nonmetropolitan county status, and state fixed effects. Model 2 additionally included healthcare access and preventive care indicators. Model 3 additionally included behavioral risk indicators. Because social, preventive care, and behavioral indicators were correlated at the county level, these sequential models were treated as supplementary conditional-association analyses and were not used as the primary basis for variable-level inference. No formal mediation analysis was undertaken, and changes in estimates across the sequential models were not interpreted as evidence of mediation, confounding control, or independent explanatory power.

Sensitivity analyses used the same core-adjusted individual-exposure specifications with log(rate + 0.1) as the outcome, with estimates expressed as percentage changes in county-level rates. Residual spatial autocorrelation was assessed using Moran’s I for the five prespecified principal exposure models. For five prespecified principal exposures—lack of broadband internet access, up-to-date CRC screening, recent dental visits, current smoking, and nonmetropolitan county status—we fitted matched spatial error models using queen-contiguity county neighbor structures. These models used the same outcome, focal exposure, covariates, state fixed effects, and population weights as the corresponding primary models (25).

All analyses were performed in R. Two-sided P values <0.05 were considered statistically significant.

3 Results

3.1 Study sample, county characteristics, and geographic distribution

The ACS-complete merged dataset included 3,142 counties. Eight counties with legacy Connecticut county FIPS codes did not have direct matches in the 2023 National Center for Health Statistics rurality classification and were excluded, leaving a county analytic base of 3,134 counties. CRC mortality data were available for 2,188 counties, all-cancer mortality data for 3,072 counties, and CRC incidence data for 2,694 counties. Counties with unavailable or suppressed outcome data accounted for 946 exclusions from the CRC mortality analysis, 62 exclusions from the all-cancer mortality analysis, and 440 exclusions from the CRC incidence analysis. After additional exclusions for missing variables required for the outcome-specific fixed analytic samples, the final samples included 2,126 counties for CRC mortality, 3,005 counties for all-cancer mortality, and 2,628 counties for CRC incidence (Figure 1).

County-level characteristics across quartiles of CRC mortality burden are summarized in Table 1. Compared with counties in the lowest CRC mortality quartile, counties in the highest quartile had a less favorable social, preventive care, behavioral, and rurality profile. Specifically, counties in the highest quartile had a higher prevalence of poverty below 150% of the federal poverty level (29.31% vs 18.69%), lack of broadband internet access (23.35% vs 13.28%), and current smoking (21.98% vs 15.44%), whereas CRC screening uptake was lower (66.56% vs 70.18%; all P <0.001). Recent dental visit prevalence was also lower in the highest than in the lowest mortality quartile (53.82% vs 65.31%; P <0.001).

CharacteristicOverallQ1 (Lowest)Q2Q3Q4 (Highest)P valueCRC Mortality Rate (per 100,000)15.56 (4.45)10.93 (1.14)13.64 (0.63)16.12 (0.86)21.58 (3.90)<0.001Age ≥65 Years (%)18.31 (4.10)17.57 (4.49)18.24 (4.48)18.64 (3.80)18.78 (3.45)<0.001Crowded Housing (%)2.28 (1.78)2.36 (1.93)2.24 (1.55)2.19 (1.62)2.35 (1.98)0.430Single-Parent Households (%)6.11 (2.02)5.51 (1.56)6.14 (1.86)6.24 (2.04)6.55 (2.41)<0.001Lack of Broadband Internet Access (%)18.17 (7.20)13.28 (5.52)16.51 (5.51)19.56 (6.68)23.35 (6.85)<0.001High Housing Cost Burden (%)23.11 (4.59)24.16 (4.72)23.78 (4.65)22.59 (4.32)21.92 (4.33)<0.001Racial/Ethnic Minority Population (%)25.20 (19.49)25.60 (18.27)25.39 (18.83)23.59 (18.88)26.21 (21.76)0.019Income Below 150% of Poverty (%)23.91 (7.96)18.69 (6.74)22.60 (6.28)25.07 (6.85)29.31 (7.92)<0.001No High School Diploma (%)11.93 (5.38)9.01 (4.77)10.93 (4.56)12.61 (4.78)15.17 (5.40)<0.001Unemployment (%)5.40 (2.11)4.84 (1.62)5.25 (1.70)5.50 (2.05)6.02 (2.71)<0.001Lack of Health Insurance (%)11.83 (5.18)9.57 (4.00)11.46 (4.41)12.32 (5.21)13.99 (5.88)<0.001Routine Checkup (%)73.03 (4.35)71.63 (5.07)72.73 (4.52)73.79 (3.72)73.99 (3.53)<0.001Up-to-Date Colorectal Cancer Screening (%)68.44 (4.77)70.18 (5.38)68.86 (4.74)68.15 (4.13)66.56 (4.00)<0.001Recent Dental Visit (%)59.50 (7.32)65.31 (5.83)60.96 (5.72)57.92 (5.81)53.82 (6.60)<0.001Obesity (%)37.26 (4.56)33.55 (4.79)36.87 (3.77)38.53 (3.21)40.11 (3.47)<0.001Current Smoking (%)18.84 (4.02)15.44 (3.39)18.08 (2.98)19.88 (2.93)21.98 (3.54)<0.001Binge Drinking (%)17.25 (2.49)18.22 (2.64)17.54 (2.29)17.02 (2.27)16.23 (2.30)<0.001Short Sleep Duration (%)34.83 (3.51)32.53 (3.14)34.54 (3.17)35.65 (3.13)36.59 (3.22)<0.001Physical Inactivity (%)26.48 (5.18)22.21 (4.30)25.45 (3.99)27.81 (3.97)30.46 (4.52)<0.001Nonmetropolitan Counties, n (%)1124 (52.9%)153 (28.8%)237 (44.5%)315 (59.3%)419 (78.9%)<0.001

County-level characteristics by quartiles of CRC mortality in the final primary analytic sample.

Values are mean (SD) unless otherwise indicated. P values were calculated using Kruskal-Wallis tests for continuous variables and chi-square tests for nonmetropolitan county status. Quartiles were defined using the CRC mortality rate in the final primary fixed complete-case sample (n =2,126).

Geographic patterns of county-level cancer outcomes are shown in Figure 2. CRC mortality, all-cancer mortality, and CRC incidence each displayed substantial geographic heterogeneity. The patterns of CRC mortality and all-cancer mortality partially overlapped, whereas CRC incidence did not completely mirror the distribution of CRC mortality. Because the maps use outcome-specific rate scales, the panels describe within-outcome geographic variation rather than directly comparable absolute color intensities across outcomes. Counties with unavailable or suppressed outcome data are shown in grey.

Geographic distribution of county-level cancer outcomes in the United States. (A) shows age-adjusted colorectal cancer mortality rates for 2019–2023; (B) shows age-adjusted all-cancer mortality rates for 2019–2023; and (C) shows age-adjusted colorectal cancer incidence rates for 2018–2022. Rates are shown per 100,000 population using outcome-specific scales. Grey counties had unavailable or suppressed outcome data.

3.2 Core-adjusted associations with CRC mortality

Core-adjusted associations between county-level factors and CRC mortality are presented in Table 2; Figure 3. Each exposure was estimated in a separate population-weighted model adjusted for the county proportion aged ≥65 years, nonmetropolitan county status where applicable, and state fixed effects.

DomainCounty-level correlate†β (95% CI), per 100,000P valueSocial determinantsCrowded housing0.17 (−0.64 to 0.98)0.447Single-parent households1.13 (0.52 to 1.74)0.005Lack of broadband internet access1.82 (1.44 to 2.20)<0.001High housing cost burden0.47 (0.04 to 0.90)0.038Racial/ethnic minority population0.29 (−0.19 to 0.78)0.198Income below 150% of the federal poverty level1.26 (0.89 to 1.63)<0.001No high school diploma1.06 (0.63 to 1.48)0.005Unemployment1.34 (0.85 to 1.84)<0.001Healthcare access and preventive careLack of health insurance1.28 (−0.37 to 2.94)0.082Routine checkup‡0.91 (0.38 to 1.45)0.009Up-to-date colorectal cancer screening−1.64 (−1.97 to −1.31)<0.001Recent dental visit−1.84 (−2.27 to −1.40)<0.001Behavioral risk indicatorsObesity1.17 (0.97 to 1.37)<0.001Current smoking2.08 (1.74 to 2.42)<0.001Binge drinking−0.51 (−1.04 to 0.03)0.058Short sleep duration1.47 (1.20 to 1.74)<0.001Physical inactivity1.65 (1.23 to 2.07)<0.001RuralityNonmetropolitan county versus metropolitan county2.73 (2.35 to 3.11)<0.001

Core-adjusted associations of county-level factors with CRC mortality (n = 2,126 counties across 49 state clusters).

CI, confidence interval; CRC, colorectal cancer.

Each row was estimated in a separate population-weighted core-adjusted linear regression model. Models adjusted for the county proportion aged ≥65 years, nonmetropolitan county status, and state fixed effects. The rurality model adjusted for age and state fixed effects but did not self-adjust for rurality. Standard errors were clustered at the state level using the CR2 variance estimator, and inference was based on Satterthwaite degrees of freedom.

†

Continuous county-level correlates were standardized to 1-SD increments. β coefficients represent absolute differences in age-adjusted county-level CRC mortality rates per 100,000 population associated with a 1-SD higher prevalence of the corresponding correlate. For rurality, the estimate compares nonmetropolitan with metropolitan counties. ‡ Routine checkup showed high focal-term variance inflation factors across outcomes and should be interpreted cautiously; see Supplementary Table 3.

Core-adjusted associations with county-level colorectal cancer mortality. Each exposure was estimated in a separate population-weighted core-adjusted linear regression model with state fixed effects and state-clustered CR2 standard errors. Continuous exposures are expressed per 1-SD higher county-level prevalence. Nonmetropolitan county status compares nonmetropolitan with metropolitan counties. Models adjusted for the county proportion aged ≥65 years and nonmetropolitan county status where applicable; the rurality model adjusted for age and state fixed effects but did not self-adjust for rurality. Points indicate β coefficients and horizontal bars indicate 95% confidence intervals. β coefficients represent absolute differences in county-level colorectal cancer mortality rates per 100,000 population.

Several social determinants were associated with higher CRC mortality. A 1-SD higher prevalence of lack of broadband internet access was associated with a 1.82-per-100,000 higher CRC mortality rate (95% CI 1.44 to 2.20; P <0.001). Higher prevalence of single-parent households, poverty below 150% of the federal poverty level, no high school diploma, unemployment, and high housing cost burden was also associated with higher CRC mortality.

Within the healthcare access and preventive care domain, higher up-to-date CRC screening prevalence was associated with lower CRC mortality (β = −1.64, 95% CI −1.97 to −1.31; P <0.001), as was higher recent dental visit prevalence (β = −1.84, 95% CI −2.27 to −1.40; P <0.001). Higher current smoking prevalence was associated with higher CRC mortality (β = 2.08, 95% CI 1.74 to 2.42; P <0.001). Obesity, short sleep duration, and physical inactivity were also positively associated with CRC mortality. Nonmetropolitan counties had higher CRC mortality rates than metropolitan counties (β = 2.73, 95% CI 2.35 to 3.11; P <0.001).

Routine checkup prevalence was positively associated with CRC mortality; however, the focal exposure term showed high VIF values across outcomes, and this estimate should therefore be interpreted cautiously. Figure 3 displays the full pattern of exposure-wise core-adjusted associations.

3.3 Core-adjusted associations with all-cancer mortality

Core-adjusted associations with all-cancer mortality are presented in Table 3. Several key associations were directionally consistent with the CRC mortality analysis. Higher lack of broadband internet access was associated with higher all-cancer mortality (β = 12.28, 95% CI 7.98 to 16.57; P <0.001). Higher current smoking prevalence was also strongly associated with higher all-cancer mortality (β = 19.95, 95% CI 18.08 to 21.82; P <0.001).

DomainCounty-level correlate†β (95% CI), per 100,000P valueSocial determinantsCrowded housing−3.66 (−9.58 to 2.26)0.115Single-parent households8.95 (4.25 to 13.65)0.004Lack of broadband internet access12.28 (7.98 to 16.57)<0.001High housing cost burden2.94 (−1.09 to 6.98)0.111Racial/ethnic minority population−3.26 (−7.03 to 0.51)0.080Income below 150% of the federal poverty level9.30 (5.48 to 13.13)<0.001No high school diploma5.80 (0.90 to 10.69)0.033Unemployment8.95 (4.56 to 13.35)0.002Healthcare access and preventive careLack of health insurance6.36 (−10.65 to 23.38)0.265Routine checkup‡10.35 (6.08 to 14.63)0.003Up-to-date colorectal cancer screening−9.61 (−14.20 to −5.02)0.004Recent dental visit−13.43 (−17.27 to −9.60)<0.001Behavioral risk indicatorsObesity10.94 (8.29 to 13.60)<0.001Current smoking19.95 (18.08 to 21.82)<0.001Binge drinking−0.42 (−5.92 to 5.08)0.848Short sleep duration12.47 (10.35 to 14.60)<0.001Physical inactivity12.34 (7.88 to 16.79)<0.001RuralityNonmetropolitan county versus metropolitan county13.93 (11.32 to 16.53)<0.001

Core-adjusted associations of county-level factors with all-cancer mortality (n = 3,005 counties across 49 state clusters).

Each row was estimated in a separate population-weighted core-adjusted linear regression model. Models adjusted for the county proportion aged ≥65 years, nonmetropolitan county status, and state fixed effects. The rurality model adjusted for age and state fixed effects but did not self-adjust for rurality. Standard errors were clustered at the state level using the CR2 variance estimator, and inference was based on Satterthwaite degrees of freedom.

† Continuous county-level correlates were standardized to 1-SD increments. β coefficients represent absolute differences in age-adjusted county-level all-cancer mortality rates per 100,000 population associated with a 1-SD higher prevalence of the corresponding correlate. For rurality, the estimate compares nonmetropolitan with metropolitan counties. ‡ Routine checkup showed high focal-term variance inflation factors across outcomes and should be interpreted cautiously; see Supplementary Table 3.

Preventive care indicators showed inverse associations with all-cancer mortality. Higher CRC screening uptake was associated with lower all-cancer mortality (β = −9.61, 95% CI −14.20 to −5.02; P = 0.004), and higher recent dental visit prevalence was associated with lower all-cancer mortality (β = −13.43, 95% CI −17.27 to −9.60; P <0.001). Higher poverty, unemployment, single-parent household prevalence, obesity, short sleep duration, and physical inactivity were also associated with higher all-cancer mortality. Nonmetropolitan counties had higher all-cancer mortality rates than metropolitan counties (β = 13.93, 95% CI 11.32 to 16.53; P <0.001).

The overlap between CRC mortality and all-cancer mortality findings indicates that several contextual indicators, including broadband deprivation, lower preventive care engagement, smoking prevalence, and rurality, may reflect broader geographic cancer disadvantage. The observed associations should not be interpreted as evidence that these factors have disease-specific or causal effects.

3.4 Exploratory, sensitivity, collinearity, and spatial analyses

Core-adjusted associations with CRC incidence are presented in Supplementary Table 1. Several findings were directionally consistent with the CRC mortality analysis. Higher lack of broadband internet access was associated with higher CRC incidence (β = 2.76, 95% CI 2.33 to 3.20; P <0.001), whereas higher CRC screening uptake and recent dental visit prevalence were associated with lower CRC incidence. Higher current smoking prevalence, obesity, short sleep duration, physical inactivity, and nonmetropolitan county status were also associated with higher CRC incidence.

Sensitivity analyses using log(rate + 0.1) as the outcome are presented in Supplementary Table 2. The direction of associations for the principal exposures was generally consistent with the primary rate-scale analyses across CRC mortality, all-cancer mortality, and CRC incidence. For example, higher lack of broadband internet access and current smoking prevalence remained associated with higher rates, whereas higher CRC screening uptake and recent dental visit prevalence remained associated with lower rates. Supplementary Figure 1 visually compares five principal core-adjusted associations across the three outcomes.

Variance inflation factor diagnostics are presented in Supplementary Table 3. Most focal exposure terms showed no material VIF concern. Up-to-date CRC screening showed moderate VIF values across outcomes, whereas routine checkup showed high VIF values. These findings support cautious interpretation of correlated healthcare-access indicators and reinforce the use of core-adjusted individual-exposure models as the primary framework for variable-level inference.

Residual spatial autocorrelation was detected for the primary core-adjusted models of the prespecified principal exposures. However, matched spatial error models produced directionally consistent estimates for lack of broadband internet access, CRC screening, recent dental visits, current smoking, and nonmetropolitan county status across CRC mortality, all-cancer mortality, and CRC incidence (Supplementary Table 4). Supplementary Figure 2 compares the state-clustered CR2 and matched spatial error model estimates for CRC mortality.

Supplementary sequential conditional-association models for CRC mortality are presented in Supplementary Table 5. Estimates for several social and preventive care indicators changed after additional correlated domain variables were included. Because these models represent increasingly conditional specifications, they are presented for transparency and are not the primary basis for variable-level inference.

4 Discussion

In this nationwide county-level ecological study, we found substantial geographic and contextual variation in CRC mortality across U.S. counties. Counties in the highest quartile of CRC mortality had a less favorable social, preventive care, behavioral, and rurality profile than counties in the lowest quartile. In the primary core-adjusted individual-exposure models, higher prevalence of broadband deprivation, poverty, lower educational attainment, unemployment, current smoking, physical inactivity, and nonmetropolitan residence was associated with higher CRC mortality (5, 26). In contrast, higher CRC screening uptake and recent dental visit prevalence were associated with lower CRC mortality. These findings should be interpreted as county-level associations rather than independent effects or causal estimates. Several principal associations were directionally similar across CRC mortality, all-cancer mortality, and CRC incidence. Higher current smoking prevalence, lower CRC screening uptake, lower recent dental visit prevalence, and nonmetropolitan county status were associated with less favorable outcomes across the three measures. Lack of broadband internet access was positively associated not only with CRC mortality and CRC incidence but also with all-cancer mortality. Accordingly, the revised results do not support an interpretation that broadband deprivation is uniquely related to CRC outcomes. Instead, broadband access may function as a broader contextual marker of healthcare connectivity, social disadvantage, care navigation capacity, and geographic cancer burden.

The association of broadband deprivation with adverse cancer outcomes is plausibly related to the growing role of digital infrastructure in healthcare access and care coordination (27–29). Broadband access may facilitate use of patient portals, telehealth services, appointment scheduling, access to health information, communication with healthcare teams, and completion of multistep screening and follow-up pathways (27–29). CRC prevention and early detection frequently require several linked processes, including screening invitation, test completion, diagnostic evaluation after an abnormal result, colonoscopy, treatment referral, and surveillance (30–34). However, the present ecological analysis cannot establish whether county-level broadband access directly influences any individual patient’s screening, diagnosis, treatment, or survival.

The inverse associations of CRC screening and recent dental visits with CRC mortality should also be interpreted at the county level. CRC screening uptake may reflect both use of evidence-based screening and the availability of prevention-oriented healthcare infrastructure. Screening can reduce CRC burden through earlier detection and removal of precancerous lesions, although county-level screening prevalence cannot identify the timing, modality, completion, or quality of screening received by individual residents (30, 35–37). Recent dental visits are not a CRC prevention intervention. Rather, they may serve as a proxy for routine healthcare engagement, preventive care orientation, outpatient service access, or community-level healthcare connectivity (38, 39). Therefore, the association between dental visits and lower cancer burden should not be interpreted as evidence of a direct protective effect of dental care on CRC.

Current smoking showed strong positive associations with CRC mortality, all-cancer mortality, and CRC incidence. This pattern is consistent with smoking serving both as a known CRC-related risk factor and as a marker of adverse behavioral and chronic disease environments at the county level (40, 41). Other behavioral indicators, including obesity, short sleep duration, and physical inactivity, were also associated with higher CRC mortality in the core-adjusted analyses (42, 43). These findings support the importance of considering behavioral environments alongside structural and preventive care conditions when describing geographic cancer inequities. Routine checkup prevalence was positively associated with CRC mortality but showed high VIF values across outcomes; therefore, this estimate should be interpreted cautiously.

The supplementary analyses strengthened the robustness of the principal findings. Log-rate sensitivity analyses yielded generally similar directions of association for the key social, preventive care, behavioral, and rurality indicators. Although residual spatial autocorrelation was detected for the prespecified principal exposure models, matched spatial error models produced directionally consistent estimates for broadband deprivation, CRC screening, recent dental visits, current smoking, and nonmetropolitan county status. These results indicate that the main patterns were not materially altered when using the alternative spatial error specification. Nevertheless, neither the CR2 models nor the spatial error models establish causality or eliminate the possibility of residual spatially patterned confounding.

The comparison of CRC mortality with all-cancer mortality and CRC incidence provides useful descriptive context. The partial geographic overlap between outcomes suggests that some county-level indicators reflect broader cancer disadvantage, whereas the incomplete overlap suggests that CRC mortality, all-cancer mortality, and CRC incidence should not be treated as interchangeable outcomes. The outcome windows also differed slightly, with CRC incidence measured during 2018–2022 and mortality outcomes measured during 2019–2023. Therefore, cross-outcome comparisons should be interpreted cautiously and should not be considered formal evidence of disease-specific pathways.

From a public health perspective, the results support geographically targeted strategies that extend beyond individual-level risk modification. High-burden counties may benefit from multicomponent approaches that increase CRC screening uptake, strengthen follow-up after abnormal screening results, improve preventive care engagement, reduce smoking prevalence, address barriers to healthcare navigation, and improve access to services in nonmetropolitan communities (44–47). Digital inclusion may be relevant within such strategies, particularly when it supports care communication, screening navigation, and continuity of care; however, digital interventions should be implemented alongside, rather than instead of, broader efforts to reduce structural barriers and improve local healthcare capacity (27, 46). For rural counties with high poverty and limited broadband access, multicomponent strategies could combine affordable broadband expansion and digital navigation support with mailed or mobile stool-based CRC screening outreach, patient navigation for timely diagnostic colonoscopy after abnormal screening results, and stronger referral linkages between dental, primary care, and preventive services.

The key strengths of this study include the use of a comprehensive, multi-source national county-level dataset and a rigorous statistical framework incorporating population weighting, state fixed effects, CR2 clustered standard errors, and extensive spatial and sensitivity analyses. Several limitations should be acknowledged. First, the ecological and cross-sectional design precludes causal inference and raises the possibility of ecological fallacy; county-level associations should not be interpreted as person-level effects. Second, CRC incidence and mortality were measured during slightly different multi-year periods. Third, CRC mortality had more unavailable or suppressed county-level rates than the other outcomes, reducing the primary analytic sample and potentially introducing selection bias. Fourth, several variables, including broadband access, dental visits, and routine checkups, are proxies for broader contextual constructs and cannot identify the specific mechanisms responsible for observed associations. Fifth, the primary individual-exposure models were intentionally designed to avoid treating correlated social, preventive care, and behavioral variables as mutually independent predictors. Consequently, the estimated associations should not be interpreted as mutually adjusted effects across all domains. Finally, public county-level data did not permit assessment of tumor stage, treatment, person-level screening history, individual insurance status, or individual socioeconomic circumstances.

5 Conclusion

In this national county-level ecological analysis, higher CRC mortality was associated with adverse social and behavioral profiles, lower preventive care engagement, and nonmetropolitan residence. Higher prevalence of current smoking and greater broadband deprivation were associated with higher CRC mortality, whereas higher CRC screening uptake and recent dental visit prevalence were associated with lower CRC mortality in core-adjusted models. Similar directional patterns across CRC mortality, all-cancer mortality, and CRC incidence suggest that these contextual indicators may reflect broader geographic cancer disadvantage rather than CRC-specific mechanisms alone. Integrated structural, preventive care, behavioral, and rural health strategies may help reduce county-level disparities in CRC outcomes.

Statements

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions

KX: Conceptualization, Formal analysis, Methodology, Writing – original draft. LL: Data curation, Methodology, Validation, Writing – original draft. CF: Formal analysis, Software, Visualization, Writing – original draft. HZ: Investigation, Validation, Writing – review & editing. XY: Data curation, Resources, Writing – review & editing. CL: Conceptualization, Project administration, Supervision, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This study was supported by the High-Level Innovative Talents Training Project of Guizhou Province (GCC (2024) 016), Science and Technology Fund Project of Guizhou Provincial Health Commission (gzwkj2026-061) and Guizhou Provincial Basic Research Program (Natural Science) (QKHJC MS (2025) 061).

Acknowledgments

The authors thank the Centers for Disease Control and Prevention, the U.S. Census Bureau, the National Center for Health Statistics, and the United States Cancer Statistics program for making the data resources used in this study publicly accessible.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Abbreviations

ACS, American Community Survey; CI, Confidence interval; CR2, Bias-reduced cluster-robust variance estimator; CRC, Colorectal cancer; GIS, Geographic information system; NCHS, National Center for Health Statistics; PLACES, Population Level Analysis and Community Estimates; USCS, United States Cancer Statistics; VIF, Variance inflation factor.

References

• SiegelRLWagleNSCercekASmithRAJemalA.
  Colorectal cancer statistics, 2023. CA Cancer J Clin. (2023) 73:233–54. doi: 10.3322/caac.21772

• LuoCKhanSJinLJamesASColditzGADrakeBF.
  Where should the cancer control interventions target: A geospatial hotspot analysis for major cancer mortality 2018 to 2022 in the United States. Cancer Epidemiol Biomarkers Prev. (2025) 34:1074–9. doi: 10.1158/1055-9965.EPI-24-0957

• SiegelRLGiaquintoANJemalA.
  Cancer statistics, 2024. CA Cancer J Clin. (2024) 74:12–49. doi: 10.3322/caac.21820

• GBD US Health Disparities Collaborators.
  Life expectancy by county, race, and ethnicity in the USA, 2000-19: a systematic analysis of health disparities. Lancet. (2022) 400:25–38. doi: 10.1016/S0140-6736(22)00876-5

• SepassiALiMZellJAChanASaundersIMMukamelDB.
  Rural-urban disparities in colorectal cancer screening, diagnosis, treatment, and survivorship care: A systematic review and meta-analysis. Oncologist. (2024) 29:e431–46. doi: 10.1093/oncolo/oyad347

• Snider-HoyNGBuchalterRBHastertTADysonGGronlundCRuterbuschJJet al.
  Social-environmental burden is associated with increased colorectal cancer mortality in metropolitan Detroit. Cancer Res Commun. (2025) 5:694–705. doi: 10.1158/2767-9764.CRC-24-0503

• CarethersJMDoubeniCA.
  Causes of socioeconomic disparities in colorectal cancer and intervention framework and strategies. Gastroenterology. (2020) 158:354–67. doi: 10.1053/j.gastro.2019.10.029

• BauerCZhangKXiaoQLuJHongY-RSukR.
  County-level social vulnerability and breast, cervical, and colorectal cancer screening rates in the US, 2018. JAMA Netw Open. (2022) 5:e2233429. doi: 10.1001/jamanetworkopen.2022.33429

• KornARWalsh-BaileyCCorrea-MendezMDelNeroPPilarMSandlerBet al.
  Social determinants of health and US cancer screening interventions: A systematic review. CA Cancer J Clin. (2023) 73:461–79. doi: 10.3322/caac.21801

• SieckCJSheonAAnckerJSCastekJCallahanBSieferA.
  Digital inclusion as a social determinant of health. NPJ Digit Med. (2021) 4:52. doi: 10.1038/s41746-021-00413-8

• DeGuzmanPBBernacchiVCuppCADunnBGhamandiBJFHintonIDet al.
  Beyond broadband: digital inclusion as a driver of inequities in access to rural cancer care. J Cancer Surviv. (2020) 14:643–52. doi: 10.1007/s11764-020-00874-y

• AcharyaMShoultsCCHayesCJBrownCC.
  Association between broadband capacity and social vulnerability factors in the United States: A county-level spatial analysis. Popul Health Manag. (2022) 25:798–806. doi: 10.1089/pop.2022.0189

• MehtaAJeonWJNagarajG.
  Association of US county-level social vulnerability index with breast, colorectal, and lung cancer screening, incidence, and mortality rates across US counties. Front Oncol. (2024) 14:1422475. doi: 10.3389/fonc.2024.1422475

• JaliliFHajizadehMMehrabaniSGhoreishySMMacIsaacF.
  The association between neighborhood socioeconomic status and the risk of incidence and mortality of colorectal cancer: A systematic review and meta-analysis of 1,678,582 participants. Cancer Epidemiol. (2024) 91:102598. doi: 10.1016/j.canep.2024.102598

• ChenKYBlackfordALSedhomRGuptaAHussainiSMQ.
  Local social vulnerability as a predictor for cancer-related mortality among US counties. Oncologist. (2023) 28:e835–8. doi: 10.1093/oncolo/oyad176

• BhattacharyaMCroninKAFarriganTLKennedyAEYuMSrinivasanS.
  Description of census-tract-level social determinants of health in cancer surveillance data. J Natl Cancer Inst Monogr. (2024) 2024:152–61. doi: 10.1093/jncimonographs/lgae027

• IslamiFBaeker BispoJLeeHWieseDYabroffKRBandiPet al.
  American Cancer Society’s report on the status of cancer disparities in the United States, 2023. CA Cancer J Clin. (2024) 74:136–66. doi: 10.3322/caac.21812

• GreenlundKJLuHWangYMatthewsKALeClercqJMLeeBet al.
  PLACES: Local data for better health. Prev Chronic Dis. (2022) 19:E31. doi: 10.5888/pcd19.210459

• Bureau UC.
  American community survey 5-year data. In: United States Census Bureau (2024). Bureau UC. American Community Survey 5-Year Data. Suitland, MD:
  U.S. Census Bureau.

• National Center for Health Statistics.
  NCHS urban-rural classification scheme for countiesNational Center for Health Statistics (2023). Available online at: https://www.cdc.gov/nchs/data-analysis-tools/urban-rural.html (Accessed June 30, 2026).

• National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control.
  Data visualizations tool technical notes. In: U.s. Cancer Statistics (2025). National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control. Data Visualizations Tool Technical Notes. Atlanta, GA:
  Centers for Disease Control and Prevention

• SiddiqueSBaumLVMDezielNCKellyJRWarrenJLMaX.
  Using a Bayesian analytic approach to identify county-level ecological factors associated with survival among individuals with early-onset colorectal cancer. PloS One. (2024) 19:e0311540. doi: 10.1371/journal.pone.0311540

• BellRMMccaffreyD.
  Bias reduction in standard errors for linear regression with multi-stage samples. Survey Method. (2002) 28:169–81.

• PustejovskyJTiptonE.
  Small-sample methods for cluster-robust variance estimation and hypothesis testing in fixed effects models. J Business Economic Stat. (2016) 36:672–83. doi: 10.1080/07350015.2016.1247004

• AnselinL. Spatial Econometrics: Methods and Models. Dordrecht:
  Kluwer Academic Publishers (1988).

• CoughlinSS.
  Social determinants of colorectal cancer risk, stage, and survival: A systematic review. Int J Colorectal Dis. (2020) 35:985–95. doi: 10.1007/s00384-020-03585-z

• ShafferKMTurnerKLSiwikCGonzalezBDUpasaniRGlazerJVet al.
  Digital health and telehealth in cancer care: A scoping review of reviews. Lancet Digit Health. (2023) 5:e316–27. doi: 10.1016/S2589-7500(23)00049-3

• LiuLOnegaTMoenELTostesonANASmithREWangQet al.
  Digital divides in telehealth accessibility for cancer care in the United States. NPJ Digit Med. (2025) 8:534. doi: 10.1038/s41746-025-01931-5

• BayardSFasanoGGillotTBrattonBIbalaRTaylor FortsonKet al.
  Breast cancer disparities and the digital divide. Curr Breast Cancer Rep. (2022) 14:205–12. doi: 10.1007/s12609-022-00468-w

• LinJSPerdueLAHenriksonNBBeanSIBlasiPR.
  Screening for colorectal cancer: Updated evidence report and systematic review for the US preventive services task force. JAMA. (2021) 325:1978–98. doi: 10.1001/jama.2021.4417

• CorleyDAJensenCDQuinnVPDoubeniCAZauberAGLeeJKet al.
  Association between time to colonoscopy after a positive fecal test result and risk of colorectal cancer and cancer stage at diagnosis. JAMA. (2017) 317:1631–41. doi: 10.1001/jama.2017.3634

• San MiguelYDembJMartinezMEGuptaSMayFP.
  Time to colonoscopy after abnormal stool-based screening and risk for colorectal cancer incidence and mortality. Gastroenterology. (2021) 160:1997–2005.e3. doi: 10.1053/j.gastro.2021.01.219

• ZorziMBattagelloJSelbyKCapodaglioGBaraccoSRizzatoSet al.
  Non-compliance with colonoscopy after a positive faecal immunochemical test doubles the risk of dying from colorectal cancer. Gut. (2022) 71:561–7. doi: 10.1136/gutjnl-2020-322192

• MohlJTCieminsELMiller-WilsonL-AGillenALuoRColangeloF.
  Rates of follow-up colonoscopy after a positive stool-based screening test result for colorectal cancer among health care organizations in the US, 2017-2020. JAMA Netw Open. (2023) 6:e2251384. doi: 10.1001/jamanetworkopen.2022.51384

• KronborgOFengerCOlsenJJørgensenODSøndergaardO.
  Randomised study of screening for colorectal cancer with faecal-occult-blood test. Lancet. (1996) 348:1467–71. doi: 10.1016/S0140-6736(96)03430-7

• BretthauerMLøbergMWieszczyPKalagerMEmilssonLGarborgKet al.
  Effect of colonoscopy screening on risks of colorectal cancer and related death. N Engl J Med. (2022) 387:1547–56. doi: 10.1056/NEJMoa2208375

• DoubeniCACorleyDAJensenCDLevinTRGhaiNRCannavaleKet al.
  Fecal immunochemical test screening and risk of colorectal cancer death. JAMA Netw Open. (2024) 7:e2423671. doi: 10.1001/jamanetworkopen.2024.23671

• MosenDPihlstromDSnyderJSmithNShusterERustK.
  Association of dental care with adherence to HEDIS measures. Perm J. (2016) 20:33–40. doi: 10.7812/TPP/15-075

• McDonaldCJohnsonJRCastroJCReitzCTorrisiFMMehtaSJ.
  Colorectal cancer screening among patients who receive medical and dental care at an urban community health center. J Community Health. (2020) 45:626–34. doi: 10.1007/s10900-019-00782-y

• BotteriEIodiceSBagnardiVRaimondiSLowenfelsABMaisonneuveP.
  Smoking and colorectal cancer: A meta-analysis. JAMA. (2008) 300:2765–78. doi: 10.1001/jama.2008.839

• BotteriEBorroniESloanEKBagnardiVBosettiCPeveriGet al.
  Smoking and colorectal cancer risk, overall and by molecular subtypes: A meta-analysis. Am J Gastroenterol. (2020) 115:1940–9. doi: 10.14309/ajg.0000000000000803

• WolinKYYanYColditzGALeeI-M.
  Physical activity and colon cancer prevention: A meta-analysis. Br J Cancer. (2009) 100:611–6. doi: 10.1038/sj.bjc.6604917

• MaYYangYWangFZhangPShiCZouYet al.
  Obesity and risk of colorectal cancer: A systematic review of prospective studies. PloS One. (2013) 8:e53916. doi: 10.1371/journal.pone.0053916

• RubinLOkitondoCHainesLEbellM.
  Interventions to increase colorectal cancer screening adherence in low-income settings within the United States: A systematic review and meta-analysis. Prev Med. (2023) 172:107522. doi: 10.1016/j.ypmed.2023.107522

• SelbyKBaumgartnerCLevinTRDoubeniCAZauberAGSchottingerJet al.
  Interventions to improve follow-up of positive results on fecal blood tests: A systematic review. Ann Intern Med. (2017) 167:565–75. doi: 10.7326/M17-1361

• NelsonHDCantorAWagnerJJungbauerRFuRKondoKet al.
  Effectiveness of patient navigation to increase cancer screening in populations adversely affected by health disparities: A meta-analysis. J Gen Intern Med. (2020) 35:3026–35. doi: 10.1007/s11606-020-06020-9

• DoughertyMKBrennerATCrockettSDGuptaSWheelerSBCoker-SchwimmerMet al.
  Evaluation of interventions intended to increase colorectal cancer screening rates in the United States: A systematic review and meta-analysis. JAMA Intern Med. (2018) 178:1645–58. doi: 10.1001/jamainternmed.2018.4637

Summary

Keywords

broadband access, cancer mortality, colorectal cancer, geographic disparities, social determinants of health

Citation

Xia K, Luo L, Fan C, Zhang H, Yang X and Lin C (2026) County-level social, preventive care, and behavioral correlates of colorectal cancer mortality in the United States: a nationwide ecological analysis. Front. Oncol. 16:1844633. doi: 10.3389/fonc.2026.1844633

Edited by

Aditi Bhargava, University of California, San Francisco, United States

Reviewed by

Chengcheng Wei, First Affiliated Hospital of Chongqing Medical University, China

Dajana Terzic, Norwegian Centre for Violence and Traumatic Stress Studies, Norway

Updates

Check for updates

Copyright

© 2026 Xia, Luo, Fan, Zhang, Yang and Lin.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Chaohuang Lin, m15285933663_1@163.com

†These authors have contributed equally to this work

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.