Abstract
Background:
Social determinants of health (SDOH) cluster together and can have deleterious impacts on health outcomes. Individually, SDOH increase risk of cancer mortality but their cumulative burden is not well understood. We sought to determine the combined effect of SDOH on cancer mortality.
Methods:
Using the REasons for Geographic and Racial differences in Stroke (REGARDS) cohort, we studied 29,766 participants aged 45+ years and followed them 10+ years. We considered 8 potential SDOH and retained SDOH that were associated with cancer mortality (p<0.10) to create a count (0, 1, 2, 3+). Cox Proportional Hazard models estimated associations between the SDOH count and cancer mortality through 12/31/2017, adjusting for confounders. Models were age-stratified (45-64 vs. 65+ years).
Results:
Participants were followed for a median of 10.6 years (IQR 6.5, 12.7). Low education, low income, zip code poverty, poor public health infrastructure, lack of health insurance, and social isolation were significantly associated with cancer mortality. In adjusted models, among those <65 years, compared to no SDOH, having 1 SDOH (aHR 1.39; 95% CI 1.11-1.75), 2 SDOH (aHR 1.61; 95% CI 1.26-2.07) and 3+ SDOHs (aHR 2.09; 95% CI 1.58-2.75) was associated with cancer mortality (p for trend <.0001). Among individuals 65+ years, compared to no SDOH, having 1 SDOH (aHR 1.16; 95% CI 1.00-1.35) and 3+ SDOHs (aHR 1.26; 95% CI 1.04-1.52) was associated with cancer mortality (p for trend 0.032).
Conclusions:
A greater number of SDOH was significantly associated with increased risk of cancer mortality, which persisted after adjustment for confounders.
Keywords: social determinants of health, cancer mortality, health equity
Introduction
Racial and ethnic minorities bear a disproportionately higher burden of cancer, as they have similar incidence yet higher mortality rates across cancer types compared to Non-Hispanic Whites.1 Given that racial/ethnic minority patients with cancer are often diagnosed with later stage disease2 and more aggressive cancer types3, under-utilization of cancer therapies is of great concern.4 However, race is not the only factor that contributes to observed cancer mortality disparities.5, 6 Numerous social determinant of health (SDOH) such as low educational attainment7, having low income8, 9, living in an impoverished area8, living in an area with relatively few healthcare services10, living in rural areas11, social isolation12, and lacking health insurance13 may worsen cancer disparities.5, 6, 14 Although many of these SDOH may have been studied separately, little has been described about the cumulative effects of these sources of health disparities on cancer mortality.
Previous studies have documented that multiple SDOH often co-exist within the same individual.15–19 A recent study using the REasons for Geographic and Racial Differences in Stroke (REGARDS) study data found that individuals with two SDOH or with three or more SDOH were at 38% and 51% increased risk of incident stroke, respectively, even after adjusting for a host of physiologic, behavioral, and medical factors.15 This was consistent with another REGARDS study that observed a graded increase in the risk of fatal CHD as the number of SDOH increased in the same individual.19 Another REGARDS study reported that individuals with two or more SDOH had twice the risk of having an incident heart failure hospitalization compared to those without any SDOH.16 Finally, among individuals hospitalized for heart failure, having two or more SDOH increased the risk of 90-day mortality by 157%.18 Although these studies have each documented the increased risk of multiple SDOH for various cardiovascular disease endpoints, to our knowledge, no studies have examined the cumulative burden of SDOH in the same individual on risk of cancer mortality.
With its rich self-reported and rigorously assessed clinical data, the large, prospective biracial REGARDS cohort is well suited for a study to fill this important gap in the literature. The first objective of this study was to determine the independent effects of individual SDOH on cancer mortality and then to determine the cumulative burden of SDOH on cancer mortality. We hypothesized that a greater number of SDOH would be significantly associated with higher risk for cancer mortality, and that associations would be greater in magnitude than any of the SDOH examined alone.
Methods
REGARDS Study:
REGARDS is a national, prospective, longitudinal cohort study evaluating racial and geographic disparities in stroke mortality.20 From 2003-2007, REGARDS recruited 30,239 community-dwelling, English-speaking individuals ≥45 years of age and continues to follow participants today.20 Upon enrollment in REGARDS, participants completed a baseline computer assisted telephone interview, which collected information about socio-demographics and medical history. At this time, participants underwent an in-home physical exam, and a medication inventory was completed. This study was approved by the University of Alabama at Birmingham and Weill Cornell Medical College’s Institutional Review Board. All participants provided written informed consent.
Study cohort:
We included all REGARDS participants who completed the baseline survey and received the in-home exam.
Cancer death:
Determining cancer mortality in REGARDS has been previously described.21 All REGARDS participants were contacted by phone to ascertain vital status every six-months.20 Deaths were identified from the medical records and from reports from participants’ proxies.21 Participants were also linked to the Social Security Death Index and the National Index.21 Proxy-reported death and cause of death in REGARDS has been found to be highly concordant with clinical adjudication.22 We calculated time to cancer death as the number of days from the baseline REGARDS survey to the date of death listed on their death certificate, the Social Security Death Index, the National Death Index, or a published obituary. The cause of death (cancer or not) was determined by expert adjudication using all available information, including medical history, hospitalizations, death certificates, autopsy reports, and proxy interviews.22 Our data included follow-up through December, 31, 2017.
Social determinants of health (SDOH):
Consistent with our prior work,15, 23 this study was guided by the Healthy People 2020 framework of SDOH.24 Overall, we considered SDOH from five primary domains of this framework: 1) education (<high school); 2) economic stability (<$35,000 annual household income); 3) neighborhood/built environment (living in a zip code with >25% of residents living below the Federal poverty line, and living in a rural area as defined by rural urban commuting area codes 9 and 10); 4) health and healthcare (living in a Health Professional Shortage Area [HPSA], lacking health insurance, and living in a US state with poor public health infrastructure); and 5) social context and community (social isolation). We determined social isolation using two questions from the baseline REGARDS survey (not seeing friends or family members at least once a month; having nobody to care for you if you become seriously ill or disabled). U.S. states that were considered to have poor public health infrastructure were identified using data from America’s Health Ranking (AHR)25 and represented states that were in the lowest decile >80% of the decade before enrollment in REGARDS (1993-2002).
Covariates:
To understand the mechanisms leading to associations between SDOH and cancer mortality, we adjusted for variables reflecting socio-demographics, medical conditions, quality of life, and lifestyle. Socio-demographics included age at baseline, race (Black vs. White), sex, and Southeastern region (stroke belt/buckle, defined as North Carolina, South Carolina, Georgia, Tennessee, Mississippi, Alabama, Louisiana and Arkansas; or non-stroke belt). Medical conditions included history of hypertension (self-report of hypertension diagnosis, use of antihypertensive medications, or blood pressure ≥140/90 mm Hg at the baseline in-home visit reflecting hypertension guidelines at the time of the observation period26), high cholesterol (self-reported diagnosis, total cholesterol >240 or low density lipoprotein (LDL) cholesterol >160 mg/dL or high density lipoprotein (HDL) <40), diabetes (use of diabetes medications or insulin, or fasting blood glucose >126 mg/dL, or non-fasting glucose >200 mg/dL), cancer (self-reported history of cancer without treatment for 2+ years on the baseline survey27), history of heart disease (self-reported MI, CABG, bypass, angioplasty, or stenting or evidence of MI via ECG and self-reported history of stroke on the baseline survey. Quality of life was measured with the Physical Component Summary (PCS) and the Mental Component Summary (MCS) scores. Behavioral measures that have been associated with cancer mortality included smoking (currently vs. not), alcohol use (risky drinking based on sex-specific National Institute on Drug Abuse cut points vs. others), and physical activity (enough activity to work up a sweat on most days of the week vs. others).
Statistical Analyses:
We examined bivariate associations between the eight candidate SDOH and cancer mortality, adjusting for age and gender. The SDOH that had statistically significant associations (p<0.10) with cancer mortality were retained for further analysis. Using the retained six SDOH (p<0.10), we created a count variable (0, 1, 2, 3+) and described characteristics of our cohort within each of the SDOH count categories. We assessed multicollinearity among SDOH using variance inflation factors (VIF).18
We then sought to determine if there were differences in the effect of the SDOH count on cancer mortality by race (Black vs. White). As such, we tested an interaction between the SDOH count and race. To assess for effect modification by age, we tested interactions between the SDOH count and two age subgroups in an overall model: 45-64 years and 65+ years.
Using Kaplan Meier plots, we depicted the cumulative risk of cancer mortality by SDOH count for each age group, separately. Using the log-rank test, we tested the equal cancer age-adjusted mortality rates by SDOH count for the two age groups, separately. We also estimated age-adjusted Cumulative Incidence Functions (CIF) curves for incidence of cancer death by SDOH count for the two age groups, separately.
We then used Cox Proportional Hazard models to determine the association between the cumulative SDOH variable and cancer mortality, by age group. First, we estimated a crude model. Then, we estimated a minimally-adjusted model that adjusted for age and sex. Finally, we added possible confounders including demographics, medical conditions, quality of life, and lifestyle. From our Cox models, we calculated adjusted hazard ratios (aHR) and 95% confidence intervals (95% CI). Using Fine and Gray’s sub-distribution hazard models, all of our models considered death from any other cause as a competing risk.28 We performed multiple imputation by chained equations on covariates that were missing. History of cancer (40%) and annual household income (12%) were the two covariates with the largest percentage of missing values. All other covariates had less than 5% missing. Analyses were conducted in SAS version 9.4 and R 3.4.1.
Results
SDOH Selection:
As shown in Table 1, out of the eight possible SDOH, low educational attainment, low annual household income, zip code poverty, poor public health infrastructure, lack of health insurance, and social isolation were statistically significantly associated with cancer mortality (p<0.10). As such, we retained these six SDOH to calculate the SDOH count variable (0. 1, 2, and 3+), as done in our prior work.15, 19, 23 In Supplementary Figure 1, we show the absolute value correlation coefficients ( coefficients) for the six SDOH. There was no evidence of multicollinearity that would impact our results.
Table 1.
Social determinants of health and cancer mortality adjusted for age and gender
| SDOHa | HR | 95 % CI | p-value | |
|---|---|---|---|---|
| Low education | 1.56 | 1.39 | 1.74 | <.0001 |
| Low income | 1.66 | 1.50 | 1.82 | <.0001 |
| Zip poverty | 1.23 | 1.11 | 1.36 | <.0001 |
| HPSAb status | 1.06 | 0.97 | 1.15 | 0.22 |
| Lack of insurance | 1.58 | 1.31 | 1.90 | <.0001 |
| Social isolation (sickness) | 1.09 | 0.96 | 1.24 | 0.18 |
| Social isolation (friends) | 1.20 | 0.99 | 1.44 | 0.06 |
| Public health infrastructure | 1.09 | 0.99 | 1.18 | 0.07 |
SDOH (Social determinants of health)
HPSA (health professional shortage area)
Race and Age Interactions:
The interaction term between race (Black vs. White) and the SDOH count variable was not statistically significant (joint test p-value=0.07). However, given that the p-value was below 0.10, we decided to estimate race-stratified models in order to visually examine possible differences between Blacks and Whites. However, did not find notable differences between Blacks and Whites in terms of which SDOH were associated with an increased risk of cancer mortality. The only difference was observed for the SDOH count level of 1 (vs. 0) so we decided not to race-stratify our analyses for increased generalizability to both Black and White adults. However, the age and SDOH interaction term was significant (joint test p<0.0001). As such, we present age-stratified results going forward.
Cohort Characteristics:
We included 29,766 REGARDS participants who completed the baseline survey. Characteristics of these participants stratified by age (<65 years and 65+ years) are shown in Tables 2 and 3. Among both participants <65 years and 65+ years, compared to those without any SDOH, those with a greater number of SDOH were more likely to be female, Black, reside in the stroke belt/buckle region, have a higher comorbidity burden (hypertension, diabetes, history of heart disease and stroke), be a smoker, and have worse physical quality of life.
Table 2.
Baseline participants’ characteristics of individuals <65 years old
| Number of SDOH | |||||
|---|---|---|---|---|---|
|
| |||||
| Variable N (%) | 0 | 1 | 2 | 3+ | p-value |
|
| |||||
| N | 4481 | 4304 | 2642 | 1849 | |
| SDOH | |||||
| < High school education | 0 (0.0%) | 97 (2.3%) | 324 (12.3%) | 730 (39.5%) | <0.001 |
| Income <$25,000 | 0 (0.0%) | 1551 (36.0%) | 1892 (71.6%) | 1675 (90.6%) | <0.001 |
| Zip with poverty >25% | 0 (0.0%) | 470 (10.9%) | 944 (35.7%) | 1186 (64.1%) | <0.001 |
| No health insurance | 0 (0.0%) | 152 (3.5%) | 499 (18.9%) | 931 (50.4%) | <0.001 |
| Social isolation | 0 (0.0%) | 191 (4.4%) | 188 (7.1%) | 211 (11.4%) | <0.001 |
| Residence in the states with poor public health infrastructure | 0 (0.0%) | 1843 (42.8%) | 1437 (54.4%) | 1419 (76.7%) | <0.001 |
|
| |||||
| Demographics | |||||
| Age, mean (SD) | 56.89 (4.88) | 57.30 (4.93) | 57.34 (4.97) | 57.48 (5.01) | <0.001 |
| Male | 2191 (48.9%) | 1863 (43.3%) | 1021 (38.6%) | 649 (35.1%) | <0.001 |
| Black | 1452 (32.4%) | 1709 (39.7%) | 1425 (53.9%) | 1281 (69.3%) | <0.001 |
| Residence in Belt/Buckle Region | 1768 (39.5%) | 2670 (62.0%) | 1802 (68.2%) | 1411 (76.3%) | <0.001 |
|
| |||||
| Comorbidities | |||||
| Hypertension | 2222 (49.6%) | 2433 (56.5%) | 1700 (64.3%) | 1316 (71.2%) | <0.001 |
| High cholesterol | 2764 (61.7%) | 2685 (62.4%) | 1685 (63.8%) | 1191 (64.4%) | 0.13 |
| Diabetes | 590 (13.2%) | 850 (19.7%) | 647 (24.5%) | 527 (28.5%) | <0.001 |
| History of CAD | 435 (9.7%) | 531 (12.3%) | 347 (13.1%) | 299 (16.2%) | <0.001 |
| History of stroke | 109 (2.4%) | 218 (5.1%) | 145 (5.5%) | 145 (7.8%) | <0.001 |
| History of cancer | 242 (5.4%) | 255 (5.9%) | 103 (3.9%) | 67 (3.6%) | <0.001 |
|
| |||||
| Health behaviors | |||||
| Current smoking | 578 (12.9%) | 750 (17.4%) | 644 (24.4%) | 565 (30.6%) | <0.001 |
| Risky alcohol consumption | 219 (4.9%) | 189 (4.4%) | 90 (3.4%) | 72 (3.9%) | 0.025 |
| Physical activity | 1299 (29.0%) | 1179 (27.4%) | 742 (28.1%) | 562 (30.4%) | 0.075 |
|
| |||||
| Health behaviors | |||||
| PCS-12, median (IQR) | 52.77 (46.95, 55.89) | 51.09 (41.62, 55.37) | 47.99 (35.64, 53.88) | 44.96 (33.53, 52.70) | <0.001 |
| MCS-12, median (IQR) | 56.64 (52.99, 58.75) | 55.93 (50.41, 58.75) | 55.48 (47.94, 58.75) | 53.24 (41.68, 57.92) | <0.001 |
Table 3.
Baseline participants’ characteristics of individuals 65+ years old
| Number of SDOH | |||||
|---|---|---|---|---|---|
|
| |||||
| Variable N (%) | 0 | 1 | 2 | 3+ | p-value |
|
| |||||
| N | 2836 | 4447 | 3122 | 1927 | |
| SDOH | |||||
| < High school education | 0 (0.0%) | 85 (1.9%) | 690 (22.1%) | 1222 (63.4%) | <0.001 |
| Income <$25,000 | 0 (0.0%) | 2632 (59.2%) | 2719 (87.1%) | 1782 (92.5%) | <0.001 |
| Zip with poverty >25% | 0 (0.0%) | 303 (6.8%) | 876 (28.1%) | 1370 (71.1%) | <0.001 |
| No health insurance | 0 (0.0%) | 10 (0.2%) | 43 (1.4%) | 101 (5.2%) | <0.001 |
| Social isolation | 0 (0.0%) | 138 (3.1%) | 229 (7.3%) | 264 (13.7%) | <0.001 |
| Residence in the states with poor public health infrastructure | 0 (0.0%) | 1279 (28.8%) | 1687 (54.0%) | 1455 (75.5%) | <0.001 |
|
| |||||
| Demographics | |||||
| Age, mean (SD) | 71.68 (5.49) | 72.63 (5.81) | 73.00 (5.94) | 73.09 (6.02) | <0.001 |
| Male | 1766 (62.3%) | 2180 (49.0%) | 1349 (43.2%) | 818 (42.4%) | <0.001 |
| Black | 754 (26.6%) | 1398 (31.4%) | 1343 (43.0%) | 1297 (67.3%) | <0.001 |
| Residence in Belt/Buckle Region | 1031 (36.4%) | 2207 (49.6%) | 1941 (62.2%) | 1341 (69.6%) | <0.001 |
|
| |||||
| Comorbidities | |||||
| Hypertension | 1807 (63.7%) | 3082 (69.3%) | 2236 (71.6%) | 1515 (78.6%) | <0.001 |
| High cholesterol | 1830 (64.5%) | 2899 (65.2%) | 2042 (65.4%) | 1274 (66.1%) | 0.69 |
| Diabetes | 496 (17.5%) | 920 (20.7%) | 787 (25.2%) | 634 (32.9%) | <0.001 |
| History of CAD | 631 (22.2%) | 1043 (23.5%) | 785 (25.1%) | 468 (24.3%) | 0.046 |
| History of stroke | 163 (5.7%) | 314 (7.1%) | 279 (8.9%) | 245 (12.7%) | <0.001 |
| History of cancer | 367 (12.9%) | 576 (13.0%) | 414 (13.3%) | 218 (11.3%) | 0.011 |
|
| |||||
| Health behaviors | |||||
| Current smoking | 165 (5.8%) | 427 (9.6%) | 377 (12.1%) | 289 (15.0%) | <0.001 |
| Risky alcohol consumption | 145 (5.1%) | 166 (3.7%) | 93 (3.0%) | 38 (2.0%) | <0.001 |
| Physical activity | 959 (33.8%) | 1315 (29.6%) | 877 (28.1%) | 513 (26.6%) | <0.001 |
|
| |||||
| Health behaviors | |||||
| PCS-12, median (IQR) | 51.84 (44.31, 55.50) | 49.36 (40.14, 54.56) | 47.15 (37.84, 53.38) | 45.71 (35.61, 52.34) | <0.001 |
| MCS-12, median (IQR) | 57.88 (55.20, 59.84) | 57.72 (53.55, 59.87) | 56.94 (52.02, 59.84) | 55.76 (48.04, 59.74) | <0.001 |
Cancer mortality:
With an overall median follow-up of 10.6 years (SD 3.9), there were 2,169 cancer-related deaths observed in our sample. Among those aged 45-64 years, the median follow-up was 11.0 years (SD 3.9), and 651 deaths were observed. Among those who were 65+ years, we had a median follow-up of 10.2 years (SD 3.9) and 1,518 deaths were observed. We report age-adjusted CIF for each of the four SDOH groups and age in Figure 1. Cancer mortality rates were lowest among those <65 years. Among both age groups, cancer mortality rates rose with the addition of each SDOH. In Figure 2, the Kaplan Meier curves are shown for the two age groups, separately. The log rank test p-value was <0.0001 for differences in survival among SDOH groups for those aged 45-64 years and for those aged 65+ (p=0.0001).
Figure 1.
Age-adjusted Cumulative Incidence Function per age strata
Figure 2.
Kaplan-meier survival curves by age strata
Participants aged <65 years:
In crude models that considered death from any cause as a competing risk, we observed a statistically significant association between increased SDOH and greater risk of cancer mortality (Table 4). Once age and sex were added to these models, observed associations were sustained. Compared to individuals with no SDOH, each additional SDOH conferred additional risk of cancer mortality (1 SDOH: HR 1.54, 95% CI 1.23-1.92; 2 SDOH: HR 2.02, 95% CI 1.60-2.55; 3+ SDOH: 2.95, 95% CI 2.23-3.75, p for trend <.0001). In fully adjusted models, we continued to observe HRs that increased in a graded fashion with p for trend <.0001. The HRs attenuated slightly from the crude and minimally adjusted models but remained statistically significant. Compare to those without any SDOH, having 1 SDOH had aHR 1.39; 95% CI 1.11-1.75; having 2 SDOH had aHR 1.61; 95% CI 1.26-2.07, and having 3 or more SDOH had aHR 2.09; 95% CI 1.58-2.75.
Table 4.
Associations between SDOH count and cancer mortality with death from other causes as a competing risk
| Crude | Minimally adjusteda | Fully adjustedb | |||||
|---|---|---|---|---|---|---|---|
| HR (95% CI) | p-value | HR (95% CI) | p-value | HR (95% CI) | p-value | ||
| SDOH 1 | 1.55 (1.24-1.93) | <.0001 | 1.54 (1.23-1.92) | <.0001 | 1.39 (1.11-1.75) | 0.005 | |
| <65 | SDOH 2 | 2.01 (1.59-2.54) | <.0001 | 2.02 (1.60-2.55) | <.0001 | 1.61 (1.26-2.07) | 0.000 |
| SDOH 3+ | 2.90 (2.29-3.68) | <.0001 | 2.95 (2.32-3.75) | <.0001 | 2.09 (1.58-2.75) | <.0001 | |
| p for trend | <.0001 | <.0001 | <.0001 | ||||
| SDOH 1 | 1.15 (1.00-1.33) | 0.0012 | 1.20 (1.04-1.39) | 0.015 | 1.16 (1.00-1.35) | 0.045 | |
| 65+ | SDOH 2 | 1.15 (0.99-1.34) | 0.0002 | 1.21 (1.04-1.42) | 0.014 | 1.16 (0.98-1.36) | 0.084 |
| SDOH 3+ | 1.24 (1.05-1.47) | <.0001 | 1.32 (1.11-1.57) | 0.002 | 1.26 (1.04-1.52) | 0.019 | |
| p for trend | 0.0173 | 0.0023 | 0.032 |
Cox Proportional Hazard Models. Subdistribution Hazard Ratios (HRs)
Multiple imputations results; SDOH contain education, income, zip poverty, insurance, social isolation, public health infrastructure
Adjusted for age, gender
Adjusted for age, gender, race, region, hypertension, high cholesterol, diabetes, history of CAD, history of stroke, smoking, alcohol consumption, physical activity, cancer, PCS, MCS
Participants aged 65+ years:
In crude and minimally adjusted models (Tables 4), we also observed statistically significant HRs in a graded fashion. In age and sex adjusted models, compared to those with 0 SDOH, having 1 SDOH (HR 1.20, 95% CI 1.04-1.39), having 2 SDOH (HR 1.21, 95% CI 1.04-1.42), and having 3+ SDOH (HR 1.31, 95% CI 1.11-1.57) significantly increased the risk of cancer mortality (p for trend 0.0023). In fully adjusted models, the HRs attenuated a bit and the HR for the 2 SDOH group became non-significant (aHR 1.16, 95% CI 0.98-1.36). However, compared to those without any SDOH, having 1 SDOH (aHR 1.16; 95% CI 1.00-1.25) and 3+ SDOH (aHR 1.26, 95% CI 1.04-1.52) was significantly associated with increased risk of death from cancer. The p for trend was 0.032.
Discussion
In our longitudinal cohort study, we observed that an increasing number of SDOH in the same individual was associated with increased risk of dying from cancer, even after adjusting for demographic factors, health status, and comorbid conditions. Although the significant relationship between a greater number of SDOH and risk of cancer mortality was larger in magnitude among individuals aged 45-65 years, it also existed among those who were 65+ years. Although there were many more deaths observed among the older group, SDOH appear to have a larger impact on cancer mortality among younger adults. We believe that these findings shed light on the cumulative burden of SDOH on cancer mortality, which can be used to inform ongoing and future efforts to reduce cancer mortality among vulnerable groups including racial/ethnic minorities and individuals with low socioeconomic status.
To our knowledge, ours is the first study to examine the effects of multiple SDOH on an individual’s risk of cancer mortality over time. To date, prior studies have focused on population trends in cancer mortality, which use area level measures for SDOH.9 For example, Singh and Jemal published a study using census-based deprivation indices from 1950-2014 that were linked to both national mortality and cancer data and found higher mortality among Blacks compared to Whites, and among individuals residing in more deprived areas.9 Although these population level findings are incredibly informative, we are also left to wonder the extent to which findings translate to the individual level due ecological fallacy. The authors suggest that observed disparities in cancer mortality may reflect differences in individual-level factors including smoking, obesity, physical inactivity, and alcohol consumption.9 However, using population-level data does not allow us to consider the role that these lifestyle behaviors play. As such, our study makes an important contribution, as we examined individual and area level measures of SDOH and assessed their relationship with cancer mortality, adjusting for a variety of demographic, clinical, and lifestyle factors.
We know that SDOH often cluster together in the same individual, as 32% of participants had 2 or more SDOH. For example, Black adults in the general US population have double the poverty and unemployment rates compared to White adults.29 Each SDOH considered in our study has been found to independently increase the risk of cancer mortality 5–14 Taken together, these SDOH appear to play a synergistic role in increasing an individual’s risk of cancer mortality, which may widen existing disparities in cancer outcomes. Interestingly, we observed that relationships between SDOH and cancer mortality, although attenuated, were sustained among individuals older than 65 years of age. This observation is distinct from the other REGARDS SDOH studies,15–19 which found that the burden of SDOH on cardiovascular outcomes lessened among older adults. Our findings suggest that SDOH continue to play an important role in cancer mortality even after adults are eligible for Medicare.
It is important to note that we did not consider race as a SDOH. After examining a possible interaction between race and SDOH on cancer mortality, we found that associations between SDOH and cancer mortality were consistent for Blacks and Whites. As such, it appears that an increased number of SDOH increases the risk of cancer mortality similarly across both racial groups. However, given that Blacks are more likely to have multiple SDOH (i.e., 43.6% of Blacks had 2+ SDOH compared to 23.9% of Whites), Blacks may experience greater adverse impacts on cancer mortality compared to Whites. Therefore, successful cancer prevention and treatment efforts should consider multiple SDOHs instead of focusing only on one (e.g., insurance status). For example, in primary care practices, individuals with several SDOH could be targeted to receive cancer screening support through interventions like patient navigation30, 31 when they become eligible for certain screenings. Comprehensive strategies that account for multiple SDOH are needed to effectively mitigate existing disparities that are prevalent across the cancer care continuum. As SDOH negatively impact cancer mortality across racial and age groups, strategies targeting multiple SDOH may benefit a broad range of cancer patients with SDOH.
Limitations:
Our study has some limitations. First, we were unable to determine the type of cancer that led to a cancer-related death. We also did not have access to receipt of cancer treatments or severity that undoubtedly influence mortality. Additionally, we did not have access to individual cancer screening patterns, which undoubtedly influence stage at diagnosis and overall prognosis. Finally, we assessed SDOH at baseline, but some factors may change over time (e.g., insurance status, social isolation). Future studies should consider time-varying effects of multiple SDOH and how they may influence risk of cancer mortality over time.
A greater number of SDOH considerably increased an individual’s risk of cancer mortality, even after adjustment for confounders. This relationship was more pronounced among individuals aged 45-64 years but was still observed among those 65+ years. Our findings highlight the profound influence that multiple SDOH can have on an individual’s risk of cancer mortality across age and race groups.
Supplementary Material
Funding:
This work was supported by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Service (U01 NS041588). This work was also supported by the National Heart Lung and Blood Institute (NHLBI) (RO1 HL80477) and by the National Cancer Institute at the National Institutes of Health (NCI) (K01 CA251645). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS, NIA, NHLBI, or NCI. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis or interpretation of the data.
Conflict of Interest:
Dr. Safford receives salary support from Amgen for investigator-initiated research. My other co-authors and I have no conflicts of interest or financial disclosures.
Footnotes
This study was approved by the participating institutions’ Institutional Review Boards. All participants provided written informed consent. All authors have read and approved the manuscript for submission to Cancer. This manuscript has not been published elsewhere and is not under consideration by another journal.
The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at: https://www.uab.edu/soph/regardsstudy/
Data availability:
Because of the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in human subject confidentiality protocols may be sent to Monika Safford, MD (mms9024@med.cornell.edu) at Weill Cornell Medicine.
References
- 1.ACS. Cancer Facts & Figures 2020. Atlanta, Georgia: American Cancer Society, 2020. [Google Scholar]
- 2.Virnig BA, Baxter NN, Habermann EB, Feldman RD, Bradley CJ. A matter of race: early-versus late-stage cancer diagnosis. Health Aff (Millwood). 2009;28: 160–168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Özdemir BC, Dotto GP. Racial Differences in Cancer Susceptibility and Survival: More Than the Color of the Skin? Trends Cancer. 2017;3: 181–197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Esnaola NF, Ford ME. Racial differences and disparities in cancer care and outcomes: where’s the rub? Surg Oncol Clin N Am. 2012;21: 417–437, viii. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Singh GK, Jemal A. Socioeconomic and Racial/Ethnic Disparities in Cancer Mortality, Incidence, and Survival in the United States, 1950-2014: Over Six Decades of Changing Patterns and Widening Inequalities. J Environ Public Health. 2017;2017: 2819372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Institute of Medicine Committee on U, Eliminating R, Ethnic Disparities in Health C. In: Smedley BD, Stith AY, Nelson AR, editors. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington (DC): National Academies Press (US) Copyright 2002 by the National Academy of Sciences. All rights reserved., 2003. [PubMed] [Google Scholar]
- 7.Albano JD, Ward E, Jemal A, et al. Cancer Mortality in the United States by Education Level and Race. JNCI: Journal of the National Cancer Institute. 2007;99: 1384–1394. [DOI] [PubMed] [Google Scholar]
- 8.O’Connor JM, Sedghi T, Dhodapkar M, Kane MJ, Gross CP. Factors Associated With Cancer Disparities Among Low-, Medium-, and High-Income US Counties. JAMA Netw Open. 2018;1: e183146. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Singh GK, Jemal A. Socioeconomic and Racial/Ethnic Disparities in Cancer Mortality, Incidence, and Survival in the United States, 1950-2014: Over Six Decades of Changing Patterns and Widening Inequalities. J Environ Public Health. 2017;2017: 2819372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Brown S, Castelli M, Hunter DJ, et al. How might healthcare systems influence speed of cancer diagnosis: a narrative review. Soc Sci Med. 2014;116: 56–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hashibe M, Kirchhoff AC, Kepka D, et al. Disparities in cancer survival and incidence by metropolitan versus rural residence in Utah. Cancer Med. 2018;7: 1490–1497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Alcaraz KI, Eddens KS, Blase JL, et al. Social Isolation and Mortality in US Black and White Men and Women. Am J Epidemiol. 2019;188: 102–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Marlow NM, Pavluck AL, Bian J, Ward EM, Halpern MT. RTI Press Research Report Series. The Relationship Between Insurance Coverage and Cancer Care: A Literature Synthesis. Research Triangle Park (NC): RTI Press; © 2009 Research Triangle Institute. All rights reserved., 2009. [PubMed] [Google Scholar]
- 14.Rehkopf DH, Haughton LT, Chen JT, Waterman PD, Subramanian SV, Krieger N. Monitoring socioeconomic disparities in death: comparing individual-level education and area-based socioeconomic measures. Am J Public Health. 2006;96: 2135–2138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Reshetnyak E, Ntamatungiro M, Pinheiro LC, et al. Impact of Multiple Social Determinants of Health on Incident Stroke. Stroke. 2020;51: 2445–2453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Pinheiro LC, Reshetnyak E, Sterling MR, Levitan EB, Safford MM, Goyal P. Multiple Vulnerabilities to Health Disparities and Incident Heart Failure Hospitalization in the REGARDS Study. Circ Cardiovasc Qual Outcomes. 2020: Circoutcomes119006438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Sterling MR, Ringel JB, Pinheiro LC, et al. Social Determinants of Health and 90-Day Mortality After Hospitalization for Heart Failure in the REGARDS Study. J Am Heart Assoc. 2020;9: e014836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Schroff P, Gamboa CM, Durant RW, Oikeh A, Richman JS, Safford MM. Vulnerabilities to Health Disparities and Statin Use in the REGARDS (Reasons for Geographic and Racial Differences in Stroke) Study. J Am Heart Assoc. 2017;6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Safford MM, Reshetnyak E, Sterling MR, et al. Number of Social Determinants of Health and Fatal and Nonfatal Incident Coronary Heart Disease in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study. Circulation. 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Howard VJ, Cushman M, Pulley L, et al. The reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology. 2005;25: 135–143. [DOI] [PubMed] [Google Scholar]
- 21.Akinyemiju T, Wilson LE, Deveaux A, et al. Association of Allostatic Load with All-Cause andCancer Mortality by Race and Body Mass Index in theREGARDS Cohort. Cancers (Basel). 2020;12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Halanych JH, Shuaib F, Parmar G, et al. Agreement on cause of death between proxies, death certificates, and clinician adjudicators in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study. Am J Epidemiol. 2011;173: 1319–1326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Pinheiro LC, Reshetnyak E, Sterling MR, Levitan EB, Safford MM, Goyal P. Multiple Vulnerabilities to Health Disparities and Incident Heart Failure Hospitalization in the REGARDS Study. Circ Cardiovasc Qual Outcomes. 2020;13: e006438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.HealhyPeople2020. Healthy People 2020: An Opportunity to Address the Societal Determinants of Health in the United States. Available from URL: https://www.healthypeople.gov/2020/topics-objectives/topic/social-determinants-of-health [accessed 1/5/2019, 2019].
- 25.United Health Foundation U. America’s Health Rankings. Available from URL: https://www.americashealthrankings.org. [accessed 12/17/18, 2018].
- 26.Qaseem A, Wilt TJ, Rich R, Humphrey LL, Frost J, Forciea MA. Pharmacologic Treatment of Hypertension in Adults Aged 60 Years or Older to Higher Versus Lower Blood Pressure Targets: A Clinical Practice Guideline From the American College of Physicians and the American Academy of Family Physicians. Ann Intern Med. 2017;166: 430–437. [DOI] [PubMed] [Google Scholar]
- 27.Pinheiro LC, Reshetnyak E, Safford MM, Nanus D, Kern LM. Healthcare fragmentation and cardiovascular risk control among older cancer survivors in the Reasons for Geographic And Racial Differences in Stroke (REGARDS) study. J Cancer Surviv. 2021;15: 325–332. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Fine JP, Gray RJ. A Proportional Hazards Model for the Subdistribution of a Competing Risk. Journal of the American Statistical Association. 1999;94: 496–509. [Google Scholar]
- 29.Singh GK, Daus GP, Allender M, et al. Social Determinants of Health in the United States: Addressing Major Health Inequality Trends for the Nation, 1935-2016. Int J MCH AIDS. 2017;6: 139–164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Doubeni CA, Selby K, Gupta S. Framework and Strategies to Eliminate Disparities in Colorectal Cancer Screening Outcomes. Annu Rev Med. 2021;72: 383–398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Dickerson JC, Ragavan MV, Parikh DA, Patel MI. Healthcare delivery interventions to reduce cancer disparities worldwide. World J Clin Oncol. 2020;11: 705–722. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Because of the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in human subject confidentiality protocols may be sent to Monika Safford, MD (mms9024@med.cornell.edu) at Weill Cornell Medicine.