Changes in Colorectal Cancer 5-Year Survival Disparities in California, 1997–2014

This article is featured in Highlights of This Issue, p. 1079

Colorectal cancer is the third most common cancer in women and men in the United States, and the second leading cause of cancer-related mortality, with a total of 53,200 deaths estimated in the United States for 2020 (1). With increased screening and scientific advances in treatment and prevention, colorectal cancer incidence and mortality rates have declined in recent decades (2, 3). However, progress in colorectal cancer outcomes has not been equal for all groups. Patients without health insurance and those with Medicaid have more advanced stage at diagnosis, less access to treatment, more postoperative complications, and higher mortality than those with private insurance or Medicare (4–10). Racial/ethnic disparities also persist for colorectal cancer outcomes including higher incidence, higher mortality, and diagnoses at more advanced stages in various racial/ethnic minority groups as compared to non-Hispanic Whites (NHW; refs. 2, 4–6, 11–14). In addition, people of lower socioeconomic status (SES) are less likely to have access to treatment, have more postoperative complications, and have higher mortality than their more affluent counterparts (7, 10, 15–17).

To examine the extent to which improvements in colorectal cancer survival were observed across sociodemographic and health insurance status groups over time, we analyzed California Cancer Registry (CCR) data from 1997 to 2014. California's sociodemographic diversity offers an opportunity to analyze disparities at multiple socioeconomic and racial/ethnic levels. Furthermore, the CCR is one of only five population-based cancer registries in the United States to have collected payer information since the 1990s, enabling an examination of trends in cancer survival disparities by health insurance status (18). These factors make California uniquely suited to provide insight into the effect of advances in science and policy on colorectal cancer survival disparities over time.

We used data from the CCR to estimate 5-year colorectal cancer–specific and overall (all causes of death) survival trends by race/ethnicity, insurance status, and neighborhood socioeconomic status (nSES) among colorectal cancer cases. Analyses included all patients diagnosed in California between January 1997 and December 2014 with CRC as a first primary malignancy with follow-up through December 2016. In order to evaluate disparities over time, three periods of six-year diagnoses were defined: January 1997 to December 2002 (1997–2002), January 2003 to December 2008 (2003–2008), and January 2009 to December 2014 (2009–2014). Of the 198,622 cases eligible for inclusion, we excluded 274 cases diagnosed at autopsy or from death certificate only.

This study received institutional review board approval as a part of the protocol for the Greater Bay Area Cancer Registry.

To ensure equal opportunity for follow-up, we right-truncated follow-up time at five years. Patient vital status was determined by routine linkage to state and national mortality and other follow-up files in addition to active follow-up efforts. For the analysis of colorectal cancer–specific survival, the underlying cause of death was obtained from death certificates, and follow-up time was censored at date of death for those who died from an underlying cause other than the primary cancer. There was a total of 58,414 colorectal cancer–specific deaths (29.4% of included cases) within 5 years of follow-up. Follow-up time for overall survival was computed as the number of days between date of diagnosis and the earliest of: date of death from any cause, date of last known contact, date 5 years after diagnosis, or December 31, 2016. In the 2009–2014 calendar period, only patients diagnosed in 2009–2011 were able to be followed for 5 years. Mean follow-up was 3.4 years for 1997–2002, 3.5 years for 2003–2008, and 3.0 for 2009–2014.

Race/ethnicity was defined as NHW, non-Hispanic Black (NHB), Hispanic, Asian/Pacific Islander (API), and unknown, primarily according to patient medical records and also with the classification system used by the CCR which employs the North American Association of Central Cancer Registries' identification algorithm for Hispanics based on surnames. For health insurance, the most extensive patient-level insurance status at the time of treatment and diagnosis was based on primary and secondary payer source and categorized as no insurance; private insurance only (no Medicare); Medicare only; Medicare plus private insurance; any public, military, or any Medicaid and/or Medi-Cal insurance; and unknown. The validity of health insurance status in the CCR has been verified with three other data sources, demonstrating an agreement of more than 80% (8).

For nSES, we employed an index that was developed for California using principal components analysis of 2000 Census (for cases diagnosed 1997–2005) or 2010 Census and 2007–2011 American Community Survey (for cases diagnosed 2006–2014) data on education, occupation, employment, household income, poverty, and rent and house values (19). Patients' addresses at diagnosis were geocoded and assigned to a census block group and then linked to the nSES index. This composite nSES score was categorized according to quintiles of the statewide distribution, with higher quintiles categories representing higher nSES.

HRs and 95% confidence intervals (CI) were calculated using multivariable Cox proportional hazard regression models to estimate the associations of race/ethnicity, insurance status, and nSES with 5-year colorectal cancer–specific and overall death (20). Models were adjusted for clustering by block group, using a sandwich estimator of the covariance structure that accounts for intracluster dependence. The proportional hazards assumption was tested by examining the correlation between time and scaled Schoenfeld residuals for all covariates. The assumption of proportional hazards was violated for chemotherapy and thus all models included this variable as a stratification factor to allow hazards to vary by chemotherapy. Model covariates included year of diagnosis, age, sex, marital status, AJCC stage, subsite, lymph nodes positive, tumor size, tumor grade, surgery, radiation, urbanization level, and whether or not patients were seen at an NCI-designated hospital. Sequential analyses were conducted adjusting for year and demographic characteristics (Model 1), Model 1 plus clinical and tumor characteristics (Model 2), Model 2 plus treatment (Model 3), and Model 3 plus neighborhood and hospital characteristics (Full Model). Wald global (and individual term) tests for interaction with time period were computed using cross-product terms in a fully-adjusted overall model additionally adjusted for all statistically significant (P < 0.05) interactions with time-period (year of diagnosis, age, AJCC stage, subsite, lymph nodes, tumor size, tumor grade, radiation, insurance type, and whether or not patients were seen at an NCI-designated hospital). All analyses were performed in SAS version 9.4 (SAS Institute, Inc).

A total of 197,060 patients diagnosed with first primary invasive colorectal cancer were included in the analysis and followed for a mean of 3.3 years (SD = 1.9). Mean age at diagnosis was 66.8 (SD = 14.1). From 1997–2002 to 2009–2014, there was an increase in the proportion of cases reported in Hispanics (12.2%–20.7%) and API (10.3%–14.7%), and a decrease in proportion of cases reported in NHW (69.7%–55.6%). Comparing the first and last time periods, the proportion of patients on Medicare only decreased from 14.9% to 7.7%, while the proportion of patients with any public insurance, Medicaid, or military insurance increased from 10.6% to 22.1%. The proportion of patients with private insurance only, Medicare plus private insurance, and the uninsured remained relatively stable. In the first time period, there was a lower proportion of colorectal cancer cases diagnosed in the lowest nSES quintile (14.0%) compared with the last time period (16.8%; Table 1). From the first two to the last time period, there was a general decrease in the proportion of uninsured Black patients and a general increase in the proportion of API and Hispanic patients with private insurance, Medicare only, Medicare plus private insurance, or any public, Medicaid, or military insurance (Supplementary Table S1).

Results of the multivariable models by periods of diagnosis are shown in Table 2. A nonsignificant decreasing trend in racial/ethnic survival disparities was observed for the study period (P_interaction = 0.559). In 1997–2002, compared with NHW, the colorectal cancer–specific HR was higher for NHB (HR, 1.12; 95% CI, 1.06–1.19) and lower for API (HR, 0.92; 95% CI, 0.87–0.96) and Hispanics (HR, 0.94; 95% CI, 0.90–0.99). In 2009–2014, however, colorectal cancer–specific hazard rate for NHB was not significantly different to the rate observed for NHW (HR, 1.03; 95% CI, 0.97–1.10). There were no significant changes in racial/ethnic disparities observed for API and Hispanics. HRs for all causes of death in the study period also reflected decreasing disparities for NHB and no significant changes for API and Hispanics, as compared with NHW (Supplementary Table S2). Sequential models indicate that disparities for NHB relative to NHW are largely due to clinical and tumor characteristics (subsite, AJCC stage, lymph nodes, tumor size, and tumor grade; Tables 3A–D).

Disparities in colorectal cancer–specific hazard rates by health insurance status persisted throughout the study period. Compared to patients with private insurance, patients in all other insurance categories had a higher colorectal cancer–specific hazard rate, and these differences were not homogeneous across the three study periods (P_interaction = 0.003). In 1997–2002, patients with no insurance had a higher colorectal cancer–specific hazard rate than patients with private insurance (HR, 1.12; 95% CI, 1.01–1.25), and based on effect estimates these differences increased in 2009–2014 (HR, 1.24; 95% CI, 1.11–1.37). Similarly, the HR in patients with Medicare only increased from the first (HR 1.09; 95% CI, 1.04–1.13) to the last time period (HR 1.26; 95% CI, 1.18–1.34). The HR in patients with Medicare plus private insurance increased from nonsignificantly different from private insurance only in the first period (HR 1.03; 95% CI, 0.99–1.07) to significantly higher in the last period (HR 1.11; 95% CI, 1.06–1.16). Relative to the hazard rate among patients with private insurance, the hazard rate in patients with any public insurance, Medicaid, or military insurance also increased from the first (HR 1.06; 95% CI, 1.01–1.12) to last time period (HR 1.20; 95% CI, 1.16–1.26; Table 2). Trends for overall survival also reflected increasing disparities for all nonprivate insurance groups (Supplementary Table S2). Sequential models indicate that while a considerable portion of the survival differences between uninsured, public, and Medicare only groups compared with private insurance were accounted for by differences in clinical and tumor characteristics, disparities remained constant over time. Treatment and neighborhood and institutional factors did not further explain the differences among the insurance groups (Tables 3A–D). An inverse association between nSES and CRC-specific hazard rates was found, with patients in the lower nSES quintiles having a higher hazard rate than those in the highest nSES quintile. This disparity did not change by period of diagnosis (P_interaction = 0.652). Trends for overall survival by nSES were similar to colorectal cancer–specific survival (Supplementary Table S2).

Using a population-based sample of nearly 200,000 individuals with incident colorectal cancer, we evaluated changes over three time periods from 1997 to 2014 in the relationship between sociodemographic and socioeconomic characteristics and colorectal cancer–specific survival. We found persistent disparities by nSES and insurance status over time but decreased colorectal cancer–specific survival differences between NHB and NHW, after adjusting for demographic, clinical, and treatment characteristics.

Historically, racial/ethnic survival disparities have been observed among patients with colorectal cancer, with minorities and especially NHB having poorer outcomes than NHW (2, 14, 21–23). However, our study shows a lack of significant difference in short-term survival (colorectal cancer and overall) between NHB and NHW in the most recent period of our study, after multivariable adjustment. This implies that survival disparities between NHB and NHW are in part due to known demographic, clinical, and treatment factors. A recent study by Sineshaw and colleagues found that most of the difference in colorectal cancer survival between NHB and NHW was explained by insurance coverage (54%) and tumor characteristics (27%; ref. 24). In addition, this pattern could reflect recent progress in California related to the uptake of colorectal cancer screening. Screening rates in California have improved for all racial/ethnic groups, but particularly in NHB. According to data from the Behavioral Risk Factor Surveillance System, adherence to screening recommendations of the United States Preventive Services Task Force have continuously increased, and indicate a very similar screening rate for NHB and NHW in recent years (67.5% vs. 68.9% in 2012 and 77.5% vs. 77.3% in 2016; refs. 25, 26).

Short-term survival disparities by health insurance status were consistently observed during the whole study period. In fact, colorectal cancer–specific and overall survival effect estimates for the uninsured and those with Medicare only, Medicare plus private insurance, or any public, Medicaid, or Military insurance, increased from one time period to another compared to patients with private coverage. This finding of higher mortality hazard for all insurance groups as compared to private insurance aligns with previous research using national SEER data that report that the uninsured and those with Medicaid are diagnosed at more advanced stages of disease, have less access to treatment, and have more postoperative complications than patients with private insurance (5, 6, 9). These findings highlight the role of health insurance and type of coverage as a critical aspect for accessing care and facilitating both early diagnosis and optimal management of the disease.

In the context of California, we would have expected a reduction in disparities during the 2009–2014 time period due to the implementation of the Affordable Care Act. However, we instead observed an increase in insurance-related disparities. This may be due to improvements in survival among privately insured patients and/or due to early enrollment of cases with more advanced disease. California opted for early expansion of Medicaid to people with incomes as high as 200% of poverty level starting in 2011. Full implementation of the ACA began in January 2014 (27); thus, to the extent that lower SES is associated with less colorectal cancer screening and more advanced stage, the colorectal cancer cases included in the early expansion time period likely included more advanced cases who were not previously eligible for Medicaid. A previous study using CCR data up to 2014 also reflected this persistent disparity across multiple cancer sites (8). Reasons for poor outcomes in people with Medicaid coverage in California may include limited physician access, unavailability of new therapies due a limited medications formulary and high costs, and a system requiring prior treatment authorizations. To eliminate these disparities, it has been proposed that a redesign of cancer care delivery is needed beyond extending health coverage for people of low income (8, 28).

Interpretation of results on race/ethnicity from the 2009–2014 time period in our study may be challenging as the effects of the ACA differed by race/ethnicity (29). While Hispanic and API women experienced a decrease in uninsured rates from 2012 to 2014, Hispanic and API men experienced little change. Uninsured rates among NHB women remained low at 8%–9%, and NHB men were the only group of men in California who experienced a large decrease in their uninsured rate (23% in 2012 to 13% in 2014). Both NHB women and men experienced an increase in coverage though employer-based insurance (45%–64% among men, 45%–53% among women; ref. 29). This may partially explain why our results showed a decrease in disparities for NHB over time, but a persistence of disparities by insurance type.

Socioeconomic status is another important factor linked to colorectal cancer patients' survival (10, 15, 30). Low SES has been linked to no access to care, late stage of disease at time of diagnosis, comorbidities, individual stressors, and poorer survival (31). In this study, we found that patients living in areas with lower nSES at diagnosis had lower colorectal cancer–specific and overall short-term survival than patients residing in places with higher nSES. A dose–response association was observed across all three periods of diagnosis. An SES gradient in cancer survival has been reported before for colorectal cancer, other cancer types, and other health outcomes (16, 30, 32). The diverse socioeconomic range among populations residing in California and access to patients' data with sociodemographic and tumor characteristics, allowed us to confirm this finding independently of other prognostic factors of survival (e.g., race/ethnicity, stage at diagnosis, health insurance coverage, and treatment).

Limitations to this study include factors inherent to cancer registry data. Patient health insurance status is determined by primary and secondary payer source which may change over time; in addition, cancer registry data capture the most extensive health insurance coverage during both diagnosis and initial treatment period, and thus may not reflect coverage at the time of diagnosis. Not knowing Medicaid enrollment information at diagnosis might have biased our results towards poorer survival in the “Any Public/Medicaid/Military” group if those patients enrolled in Medicaid after diagnosis (33). The survival experience of these patients resembles that of the uninsured since they may have been without continuous coverage until receiving a late-stage diagnosis. Previous research has shown differences in stage at diagnosis of colorectal cancer for patients with different versions of Medicaid coverage, with increasing odds of later stages (compared with private coverage) for continuous Medicaid, discontinuous Medicaid, and Medicaid at diagnosis (34). Our study was also unable to account for other factors such as patient sociodemographic factors, comorbidity, receipt of detailed guideline concordant treatment, or managed and/or fee-for-service health care systems which may have influenced survival disparities in our study population (8). CCR data on surgery, chemotherapy, and radiation is limited to first course and thus does not provide a full account of all treatment received. Furthermore, complete cancer registry data reporting and vital status determination is delayed and we were only able to obtain follow-up data through 2016, resulting in incomplete 5-year follow-up for part of our study population. Other limitations to our study include potential lead-time bias due to differential uptake of screening and the restriction to short-term 5-year survival due to the structure of the analysis.

In summary, using population-based cancer registry data spanning two decades from a large, diverse state, we found a decrease in survival disparities over time by race/ethnicity but a persistence of disparities by nSES and health insurance status. As more years of cancer data and other types of data on underlying causes are available, further investigation into the drivers for these disparities can help direct policy and practice toward health equity for all groups.

No potential conflicts of interest were disclosed.

The ideas and opinions expressed herein are those of the author(s) and do not necessarily reflect the opinions of the NIH, State of California, Department of Public Health, the NCI, and the Centers for Disease Control and Prevention or their Contractors and Subcontractors.

Conception and design: D.L. Oh, L. Ellis, S.L. Gomez

Development of methodology: D.L. Oh, L. Ellis, L. Tao, S.L. Gomez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.L. Oh, S.L. Gomez

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.L. Oh, E.J. Santiago-Rodriguez, A.J. Canchola, L. Tao

Writing, review, and/or revision of the manuscript: D.L. Oh, E.J. Santiago-Rodriguez, A.J. Canchola, L. Ellis, S.L. Gomez

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.L. Oh

Study supervision: D.L. Oh, S.L. Gomez

Research reported in this publication was supported by National Institutes of Aging of the NIH under award number T32-AG-049663. The collection of cancer incidence data used in this study was supported by the California Department of Public Health pursuant to California Health and Safety Code Section 103885 (UCSF principal investigator: S.L. Gomez); Centers for Disease Control and Prevention's (CDC) National Program of Cancer Registries, under cooperative agreement 5NU58DP006344 (UCSF principal investigator: S.L. Gomez); the National Cancer Institute's Surveillance, Epidemiology, and End Results Program under contract HHSN261201800032I awarded to the University of California, San Francisco (UCSF principal investigator: S.L. Gomez), contract HHSN261201800015I awarded to the University of Southern California, and contract HHSN261201800009I awarded to the Public Health Institute, Cancer Registry of Greater California. E.J. Santiago-Rodriguez's role on this manuscript is supported by T32-AG-049663; all other authors' roles are supported by HHSN261201800032I.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.