Abstract
Background
Lower neighborhood socioeconomic status (SES) is associated with suboptimal cancer care and reduced survival. Most studies examining cancer inequities across area-level socioeconomic status tend to use less granular or unidimensional measures and pre-date the COVID-19 pandemic. Here, we examined the association of area-level socioeconomic status on real-world treatment initiation and overall survival among adults with 20 common cancers.
Methods
This retrospective cohort study used electronic health record–derived deidentified data (Flatiron Health Research Database, 2011-2022) linked to US Census Bureau data from the American Community Survey (2015-2019). Area-level socioeconomic status quintiles (based on a measure incorporating income, home values, rental costs, poverty, blue-collar employment, unemployment, and education information) were computed from the US population and applied to patients based on their mailing address. Associations were examined using Cox proportional hazards models adjusted for diagnosis year, age, sex, performance status, stage, and cancer type.
Results
This cohort included 291 419 patients (47.7% female; median age = 68 years). Patients from low–SES areas were younger and more likely to be Black (21.9% vs 3.3%) or Latinx (8.4% vs 3.0%) than those in high–SES areas. Living in low–SES areas (vs high) was associated with lower treatment rates (hazard ratio = 0.94 [95% confidence interval = 0.93 to 0.95]) and reduced survival (median real-world overall survival = 21.4 vs 29.5 months, hazard ratio = 1.20 [95% confidence interval = 1.18 to 1.22]). Treatment and survival inequities were observed in 9 and 19 cancer types, respectively. Area-level socioeconomic inequities in treatment and survival remained statistically significant in the COVID-19 era (after March 2020).
Conclusion
To reduce inequities in cancer outcomes, efforts that target marginalized, low–socioeconomic status neighborhoods are necessary.
Socioeconomic status describes the position of individuals or areas on a hierarchical social structure based on wealth, prestige, and power; it is conceptualized using income, wealth, education, occupation, and living conditions (1). Lower individual-level and area-level socioeconomic status is associated with suboptimal cancer care and reduced survival (2-8). Because of structural injustices (eg, underinvestment in low–socioeconomic status neighborhoods leading to fewer health-care resources), area-level socioeconomic status influences cancer mortality, even after accounting for individual-level socioeconomic status (5,7).
Although previous research has demonstrated that lower area-level socioeconomic status adversely influences cancer outcomes, knowledge gaps remain. For example, prior studies tended to use less granular (eg, counties or zip codes) or overly simplistic socioeconomic status measures (eg, income or education alone) that fail to capture heterogeneity across large geographic areas (median population sizes range from 25 736 in counties to 1270 in block groups) or the multiple domains that contribute to area-level socioeconomic status (6,8). More recent studies have captured area-level socioeconomic status holistically (ie, composite variables that capture several socioeconomic status domains), but they tended to focus on limited subpopulations (eg, specific regions, cancer types, or age groups) (4,7). Moreover, to our knowledge, none of these studies evaluated inequities experienced by patients with cancer during the COVID-19 pandemic, which has exacerbated cancer inequities (9-11). Therefore, an up-to-date evaluation of area-level socioeconomic status and its potential influence on treatment and survival inequities among patients diagnosed with cancer is necessary.
Here, we provide an up-to-date, comprehensive, and critically needed evaluation of area-level socioeconomic status (measured using US Census Bureau block group data incorporating multiple socioeconomic status dimensions) and its potential influence on treatment and survival inequities among adult patients who received care for 20 solid and blood cancers. In the United States, the types of cancers examined were common and constituted approximately 75% of cancer-related deaths in 2020 (12).
Methods
Design and participants
This retrospective study used the nationwide Flatiron Health Research Database, which comprises deidentified longitudinal, patient-level structured and unstructured electronic health record (EHR)–derived data from approximately 280 US cancer clinics (approximately 800 sites), curated by technology-enabled abstraction (13,14). Patients were required to have clinical activity in the EHR between January 2011 and December 2022 (inclusion criteria are available in Supplementary Table 1, available online).
Patient characteristics, exposure, and outcomes
Patient characteristics were extracted from the EHR and included age (at index diagnosis), sex, race and ethnicity (non-Latinx White, non-Latinx Black, Latinx, non-Latinx Asian, or other/not documented), region, practice type (academic or community-based cancer center), insurance (insurer at any point during the 2 years before index diagnosis to 1 year following diagnosis), stage (at initial diagnosis), Eastern Cooperative Oncology Group (ECOG) performance status (maximum value within 30 days of index diagnosis), and index diagnosis year. Practices affiliated with a university or medical school were considered academic cancer centers; all other practices were considered community-based cancer centers. Because of differences in reporting practices at clinical sites, it is unknown whether sex and race and ethnicity variables are self-report or clinician reported. Sex values likely represent a patient’s sex assigned at birth.
The primary exposure, area-level socioeconomic status, was calculated using Census block group (ie, neighborhood) data from the American Community Survey (2015-2019) per the Yost Index (incorporating income, home values, rental costs, poverty, blue-collar employment, unemployment, and education information) (2). The Yost Index outperforms other socioeconomic status indices in terms of stratifying areas and detecting cancer inequities (3). Population-standardized socioeconomic status quintiles were applied to patients based on their most recent documented residential address. The distribution of area-level socioeconomic status in the Flatiron Health Research Database is similar to that of the US population, suggesting that the Flatiron Health Research Database is representative of area-level socioeconomic status in the United States (Supplementary Figure 1, available online). Comparisons focused on differences between low– socioeconomic status (Q1) and high– socioeconomic status (Q5 [Referent]) areas.
Patients were followed from index diagnosis date (initial, advanced, or metastatic diagnosis, depending on cancer type; Supplementary Table 2, available online) until the event of interest (treatment initiation or death), last confirmed activity (recording of vital information, medication administration, a noncanceled drug order, or a reported laboratory test), or end of study period. Treatment initiation was defined as the first documented receipt of systemic antineoplastic therapy or stem cell transplantation (for blood cancers). Death was determined using structured and unstructured EHR-derived data linked to a commercial death data source and US Social Security Death Index (15).
Statistical analysis
We used Χ2 tests, Wilcoxon rank-sum tests, and log-rank tests to evaluate differences. To examine associations, we used Cox proportional hazards models adjusted for clinical characteristics (ie, age, sex, performance status, stage, and cancer type) and index year (informed by the National Academy of Medicine “healthcare disparities” definition) (16). Explicitly, to assess the association of area-level socioeconomic status with treatment initiation and survival, we did not adjust for other social determinants of health (eg, insurance) because these factors are likely on the causal pathway (17-19). We examined these associations among all patients diagnosed with cancer and conducted stratified analyses by cancer type, race and ethnicity, and period of diagnosis (before or during the COVID-19 pandemic) (17).
As a sensitivity analysis, we also examined the associations of area-level socioeconomic status with treatment initiation and overall survival in fully adjusted models. These estimates should be interpreted with caution because this approach (“residual direct effect”) underestimates inequities by ignoring distinctions between variables that may be legitimate sources of health differences and those that may represent systemic disadvantages (17,18).
Statistical analyses were conducted using R, version 4.2.2. All P values were 2-sided, P < .05 was considered statistically significant, and estimates were presented with 95% confidence intervals (CIs). The institutional review board of WCG (reference No. IRB00000533) gave ethical approval for the study protocol before study conduct that included a waiver of informed consent.
Results
This study included 291 419 patients (47.7% female; median age = 68 years; 43.7% had a de novo metastatic diagnosis; 11.4% had a performance score ≥2) (Table 1). Patients living in low–socioeconomic status areas (Q1) were younger, more likely to be non-Latinx Black (21.9% vs 3.3%) or Latinx (8.4% vs 3.0%), live in the South (61.7% vs 29.9%), be seen at community-based centers (85.0% vs 70.3%), and be Medicaid insured (20.9% vs 7.3%) than those in high–socioeconomic status areas (Q5) (all P < .001).
Table 1.
Sociodemographic and health characteristics of patients with cancer, by area-level socioeconomic status, 2011-2022
|
Area-level socioeconomic status,
a
,
b
No. (%)
|
||||||
|---|---|---|---|---|---|---|
| Characteristicsc | Overall | 5 (high) | 4 | 3 | 2 | 1 (low) |
| Total | 291 419 (100.0) | 58 636 (20.1) | 67 097 (23.0) | 62 067 (21.3) | 56 552 (19.4) | 47 067 (16.2) |
| Age, No. (%), y | ||||||
| Median (IQR) | 68 (60-76) | 69 (60-77) | 69 (60-76) | 68 (60-76) | 67 (59-75) | 67 (58-75) |
| 18-49 | 23 085 (7.9) | 4579 (7.8) | 5158 (7.7) | 4817 (7.8) | 4454 (7.9) | 4077 (8.7) |
| 50-64 | 88 721 (30.4) | 16 371 (27.9) | 19 298 (28.8) | 18 657 (30.1) | 18 313 (32.4) | 16 082 (34.2) |
| 65-75 | 93 429 (32.1) | 18 314 (31.2) | 21 789 (32.5) | 20 049 (32.3) | 18 146 (32.1) | 15 131 (32.1) |
| >75 | 86 184 (29.6) | 19 372 (33.0) | 20 852 (31.1) | 18 544 (29.9) | 15 639 (27.7) | 11 777 (25.0) |
| Sex, No. (%) | ||||||
| Female | 139 060 (47.7) | 28 346 (48.3) | 31 647 (47.2) | 29 405 (47.4) | 26 848 (47.5) | 22 814 (48.5) |
| Male | 152 359 (52.3) | 30 290 (51.7) | 35 450 (52.8) | 32 662 (52.6) | 29 704 (52.5) | 24 253 (51.5) |
| Race and ethnicity, No. (%) | ||||||
| Latinx | 14 095 (4.8) | 1735 (3.0) | 2492 (3.7) | 2754 (4.4) | 3154 (5.6) | 3960 (8.4) |
| Non-Latinx Asian | 6334 (2.2) | 2211 (3.8) | 1547 (2.3) | 1107 (1.8) | 881 (1.6) | 588 (1.2) |
| Non-Latinx Black | 25 626 (8.8) | 1949 (3.3) | 3626 (5.4) | 4260 (6.9) | 5487 (9.7) | 10 304 (21.9) |
| Non-Latinx White | 191 572 (65.7) | 41 548 (70.9) | 47 041 (70.1) | 42 358 (68.2) | 36 668 (64.8) | 23 957 (50.9) |
| Other/not documented | 53 792 (18.5) | 11 193 (19.1) | 12 391 (18.5) | 11 588 (18.7) | 10 362 (18.3) | 8258 (17.5) |
| Region, No. (%) | ||||||
| Northeast | 77 622 (26.6) | 23 708 (40.4) | 21 740 (32.4) | 15 271 (24.6) | 9817 (17.4) | 7086 (15.1) |
| Midwest | 32 666 (11.2) | 4019 (6.9) | 7030 (10.5) | 8203 (13.2) | 7871 (13.9) | 5543 (11.8) |
| South | 132 174 (45.4) | 17 516 (29.9) | 26 000 (38.7) | 28778 (46.4) | 30841 (54.5) | 29039 (61.7) |
| West | 48 957 (16.8) | 13 393 (22.8) | 12 327 (18.4) | 9815 (15.8) | 8023 (14.2) | 5399 (11.5) |
| Community-based center, No. (%) | 229 977 (78.9) | 41 242 (70.3) | 50 882 (75.8) | 50030 (80.6) | 47808 (84.5) | 40015 (85.0) |
| Insurance,d No. (%) | ||||||
| Yes | 252 556 (86.7) | 51 258 (87.4) | 58 657 (87.4) | 53 898 (86.8) | 48 657 (86.0) | 40 086 (85.2) |
| Commerciale | 131 010 (45.0) | 26 513 (45.2) | 30 693 (45.7) | 27 916 (45.0) | 25 182 (44.5) | 20 706 (44.0) |
| Medicaree | 189 813 (65.1) | 38 927 (66.4) | 44 576 (66.4) | 40 753 (65.7) | 36 196 (64.0) | 29 361 (62.4) |
| Medicaid/other public payere | 37 238 (12.8) | 4291 (7.3) | 6737 (10.0) | 7567 (12.2) | 8795 (15.6) | 9848 (20.9) |
| No/not documented | 38 849 (13.3) | 7375 (12.6) | 8437 (12.6) | 8165 (13.2) | 7893 (14.0) | 6979 (14.8) |
| Stage,f No. (%) | ||||||
| <II | 53 041 (18.2) | 11 651 (19.9) | 12 447 (18.6) | 11 426 (18.4) | 9779 (17.3) | 7738 (16.4) |
| III | 59 117 (20.3) | 11 475 (19.6) | 13 526 (20.2) | 12 607 (20.3) | 11 740 (20.8) | 9769 (20.8) |
| IV (de novo metastatic) | 127 476 (43.7) | 24 501 (41.8) | 29 060 (43.3) | 27 064 (43.6) | 25 292 (44.7) | 21 559 (45.8) |
| Not documented | 51 785 (17.8) | 11 009 (18.8) | 12 064 (18.0) | 10 970 (17.7) | 9741 (17.2) | 8001 (17.0) |
| ECOG performance status,g No. (%) | ||||||
| 0 | 48 961 (16.8) | 11 009 (18.8) | 11 880 (17.7) | 10 360 (16.7) | 8881 (15.7) | 6831 (14.5) |
| 1 | 59 710 (20.5) | 10 631 (18.1) | 13 618 (20.3) | 12 999 (20.9) | 12 330 (21.8) | 10 132 (21.5) |
| ≥2 | 33 155 (11.4) | 5514 (9.4) | 7318 (10.9) | 7315 (11.8) | 6992 (12.4) | 6016 (12.8) |
| Not documented | 149 579 (51.3) | 31 479 (53.7) | 34 278 (51.1) | 31 389 (50.6) | 28 347 (50.1) | 24 086 (51.2) |
| Diagnosis, No. (%), y | ||||||
| 2011-2014 | 74 334 (25.5) | 15 317 (26.1) | 17 133 (25.5) | 15 894 (25.6) | 14 330 (25.3) | 11 660 (24.8) |
| 2015-2018 | 114 649 (39.3) | 23 041 (39.3) | 26 510 (39.5) | 24 424 (39.4) | 22 226 (39.3) | 18 448 (39.2) |
| 2019-2022 | 102 436 (35.2) | 20 278 (34.6) | 23 454 (35.0) | 21 749 (35.0) | 19 996 (35.4) | 16 959 (36.0) |
US Census Bureau block group data from the American Community Survey (2015-2019) were used to measure area-level socioeconomic status per the Yost Index (incorporating income, home values, rental costs, poverty, blue-collar employment, unemployment, and education information). Socioeconomic status quintiles were determined from the US population, and then applied to patients based on their latest residential address. ECOG = Eastern Cooperative Oncology Group; IQR = interquartile range.
All differences by area-level socioeconomic status were statistically significant (P < .05).
Characteristics were measured at index diagnosis unless otherwise stated. Index diagnosis was defined as initial, advanced, or metastatic diagnosis, depending on cancer type (as listed in Supplementary Table 2, available online).
Insurer at any point during the 2 years before index cancer diagnosis to 1 year following diagnosis.
Not mutually exclusive categories.
Stage at initial cancer diagnosis. Staging system varies by cancer diagnosis. Staging was not documented for patients with acute myeloid leukemia.
Highest (worse) ECOG performance status value measured 30 days before or after index cancer diagnosis.
In this cohort of 20 common cancers, lower area-level socioeconomic status was adversely associated with treatment initiation (Figure 1, A), and median time to treatment initiation was 3 days longer among patients in low–socioeconomic status neighborhoods than for their counterparts in high–socioeconomic status neighborhoods (48 vs 45 days). Patients in low–socioeconomic status areas had 6% lower rates of treatment initiation than those in high–socioeconomic status areas (adjusted hazard ratio [HR] = 0.94, 95% CI = 0.93 to 0.95, P < .001) (Table 2). In analyses stratified by cancer type, treatment inequities between low– and high–socioeconomic status neighborhoods were found in 9 cancer diagnoses (all P < .05). The largest treatment inequities were observed among patients with advanced endometrial cancer (time to treatment initiation = 91 vs 76 days; HR = 0.77, 95% CI = 0.70 to 0.86), advanced gastric/esophageal cancer (time to treatment initiation = 52 vs 42 days, HR = 0.87, 95% CI = 0.81 to 0.94), and ovarian cancer (time to treatment initiation = 68 vs 56 days, HR = 0.88, 95% CI = 0.81 to 0.95) (all P < .001). (For ease of clinical interpretation, Supplementary Table 3, available online, presents time to treatment initiation within 30, 60, and 180 days).
Figure 1.
Unadjusted Kaplan-Meier estimates of treatment initiation and real-world overall survival, by area-level socioeconomic status among patients receiving care for 20 common cancers. US Census Bureau block group data from the American Community Survey (2015-2019) were used to measure area-level socioeconomic status per the Yost Index (incorporating income, home values, rental costs, poverty, blue-collar worker, unemployment, and education information). Socioeconomic status quintiles were determined from the US population, and then applied to patients based on their latest residential address. The index date was defined as initial, advanced, or metastatic diagnosis (depending on specific cancer diagnosis; see Supplementary Table 2, available online). Patients were followed from index diagnosis until the event of interest (treatment initiation or death), last confirmed activity, or end of the study period (December 2022). SES = socioeconomic status.
Table 2.
Association of area-level socioeconomic status with treatment initiation and real-world overall survival
| Total, No. | Treatment initiation |
Real-world overall survival |
|||||
|---|---|---|---|---|---|---|---|
| Area-level socioeconomic status a | Treated, No. | Median time to treatment initiation (95% CI)b | Adjusted HR (95% CI) b , c | Deaths, No. | Median survival (95% CI), b mo | Adjusted HR (95% CI) b , c | |
| All cancers | 291 419 | 228 219 | — | — | 174 304 | — | — |
| 5 (high) | 58 636 | 46 203 | 45 (44 to 46) | 1.00 (Referent) | 33 008 | 29.5 (29.0 to 30.1) | 1.00 (Referent) |
| 4 | 67 097 | 52 795 | 45 (45 to 46) | 0.98 (0.97 to 1.00)f | 39 669 | 26.1 (25.6 to 26.5) | 1.06 (1.04 to 1.07)g |
| 3 | 62 067 | 48 737 | 46 (45 to 47) | 0.98 (0.97 to 0.99)g | 37 410 | 24.4 (24.0 to 24.8) | 1.10 (1.08 to 1.12)g |
| 2 | 56 552 | 44 120 | 46 (46 to 47) | 0.96 (0.95 to 0.98)g | 34 894 | 22.4 (22.0 to 22.8) | 1.16 (1.14 to 1.17)g |
| 1 (low) | 47 067 | 36 364 | 48 (48 to 49) | 0.94 (0.93 to 0.95)g | 29 323 | 21.4 (20.9 to 21.8) | 1.20 (1.18 to 1.22)g |
| Solid cancers | |||||||
| Advanced non-small cell lung cancer | 74 265 | 54 652 | — | — | 53 215 | — | — |
| 5 (high) | 13 607 | 10 202 | 44 (43 to 45) | 1.00 (Referent) | 9571 | 15.4 (14.9 to 15.9) | 1.00 (Referent) |
| 4 | 16 627 | 12 332 | 46 (45 to 47) | 0.96 (0.94 to 0.99)g | 11 870 | 14.2 (13.7 to 14.6) | 1.04 (1.02 to 1.07)g |
| 3 | 15 987 | 11 873 | 46 (45 to 47) | 0.96 (0.93 to 0.98)g | 11 555 | 13.4 (12.9 to 13.7) | 1.09 (1.06 to 1.12)g |
| 2 | 15 195 | 10 990 | 48 (47 to 49) | 0.91 (0.88 to 0.93)g | 11 015 | 12.9 (12.5 to 13.3) | 1.14 (1.11 to 1.18)g |
| 1 (low) | 12 849 | 9255 | 49 (48 to 50) | 0.90 (0.87 to 0.92)g | 9204 | 12.7 (12.3 to 13.1) | 1.15 (1.11 to 1.18)g |
| Metastatic colorectal cancer | 30 342 | 24 308 | — | — | 19 008 | — | — |
| 5 (high) | 5799 | 4653 | 47 (45 to 49) | 1.00 (Referent) | 3508 | 27.4 (26.3 to 28.6) | 1.00 (Referent) |
| 4 | 6723 | 5351 | 48 (46 to 49) | 0.98 (0.94 to 1.02) | 4172 | 25.9 (24.9 to 26.9) | 1.06 (1.01 to 1.11)g |
| 3 | 6204 | 4994 | 48 (46 to 49) | 1.00 (0.96 to 1.04) | 3858 | 25.4 (24.5 to 26.3) | 1.07 (1.03 to 1.12)g |
| 2 | 6106 | 4902 | 49 (48 to 50) | 0.98 (0.94 to 1.02) | 3883 | 24.4 (23.4 to 25.5) | 1.12 (1.07 to 1.18)g |
| 1 (low) | 5510 | 4408 | 53 (51 to 55) | 0.96 (0.92 to 1.00) | 3587 | 23.5 (22.8 to 24.3) | 1.19 (1.14 to 1.25)g |
| Metastatic breast cancer | 28 269 | 24 945 | — | — | 16 489 | — | — |
| 5 (high) | 6131 | 5451 | 26 (25 to 27) | 1.00 (Referent) | 3416 | 40.4 (39.1 to 42.0) | 1.00 (Referent) |
| 4 | 6459 | 5737 | 27 (26 to 28) | 0.94 (0.91 to 0.98)g | 3720 | 37.0 (35.8 to 38.4) | 1.07 (1.02 to 1.12)g |
| 3 | 5902 | 5208 | 28 (27 to 29) | 0.94 (0.91 to 0.98)g | 3496 | 35.2 (33.5 to 36.5) | 1.13 (1.08 to 1.18)g |
| 2 | 5229 | 4582 | 29 (28 to 30) | 0.92 (0.88 to 0.95)g | 3077 | 33.8 (32.3 to 35.3) | 1.14 (1.09 to 1.20)g |
| 1 (low) | 4548 | 3967 | 31 (30 to 33) | 0.91 (0.87 to 0.94)g | 2780 | 30.5 (29.2 to 32.2) | 1.25 (1.18 to 1.31)g |
| Metastatic prostate cancer | 15 964 | 11 469 | — | — | 8062 | — | — |
| 5 (high) | 3445 | 2435 | 255 (233 to 277) | 1.00 (Referent) | 1705 | 41.3 (38.8 to 43.1) | 1.00 (Referent) |
| 4 | 3817 | 2772 | 224 (207 to 252) | 1.03 (0.97 to 1.08) | 1905 | 40.3 (38.5 to 42.0) | 1.04 (0.98 to 1.11) |
| 3 | 3422 | 2471 | 234 (213 to 256) | 1.02 (0.96 to 1.07) | 1783 | 39.2 (37.2 to 40.8) | 1.09 (1.02 to 1.17)g |
| 2 | 2939 | 2130 | 237 (217 to 270) | 1.00 (0.94 to 1.06) | 1482 | 40.9 (38.4 to 43.3) | 1.06 (0.99 to 1.13) |
| 1 (low) | 2341 | 1661 | 238 (212 to 276) | 0.96 (0.90 to 1.02) | 1187 | 39.9 (37.8 to 41.6) | 1.14 (1.06 to 1.23)g |
| Advanced melanoma | 13 791 | 9779 | — | — | 5774 | — | — |
| 5 (high) | 3338 | 2392 | 114 (106 to 122) | 1.00 (Referent) | 1270 | 74.8 (64.5 to 84.0) | 1.00 (Referent) |
| 4 | 3405 | 2411 | 114 (107 to 124) | 0.99 (0.93 to 1.04) | 1407 | 59.1 (53.6 to 66.7) | 1.15 (1.07 to 1.24)g |
| 3 | 3034 | 2170 | 119 (110 to 126) | 0.99 (0.93 to 1.05) | 1288 | 54.6 (51.4 to 59.1) | 1.18 (1.10 to 1.28)g |
| 2 | 2507 | 1760 | 117 (110 to 129) | 0.96 (0.90 to 1.02) | 1115 | 46.0 (40.5 to 52.0) | 1.34 (1.23 to 1.45)g |
| 1 (low) | 1507 | 1046 | 124 (114 to 137) | 0.92 (0.86 to 0.99)f | 694 | 44.7 (39.7 to 52.2) | 1.43 (1.30 to 1.57)g |
| Metastatic pancreatic cancer | 12 803 | 9468 | — | — | 10 149 | — | — |
| 5 (high) | 2889 | 2137 | 28 (27 to 29) | 1.00 (Referent) | 2276 | 8.6 (8.2 to 9.0) | 1.00 (Referent) |
| 4 | 3146 | 2366 | 26 (25 to 27) | 1.02 (0.97 to 1.09) | 2486 | 8.0 (7.6 to 8.4) | 1.05 (1.00 to 1.12) |
| 3 | 2670 | 1955 | 28 (26 to 28) | 0.97 (0.91 to 1.03) | 2097 | 7.5 (7.0 to 7.9) | 1.10 (1.04 to 1.17)g |
| 2 | 2299 | 1693 | 27 (26 to 28) | 0.98 (0.92 to 1.05) | 1877 | 6.2 (5.8 to 6.6) | 1.24 (1.17 to 1.32)g |
| 1 (low) | 1799 | 1317 | 28 (27 to 30) | 0.94 (0.87 to 1.00) | 1413 | 6.4 (6.0 to 6.9) | 1.21 (1.13 to 1.29)g |
| Advanced gastric cancer | 11 679 | 8655 | — | — | 8387 | — | — |
| 5 (high) | 2298 | 1702 | 42 (40 to 44) | 1.00 (Referent) | 1605 | 13.3 (12.6 to 14.2) | 1.00 (Referent) |
| 4 | 2717 | 2007 | 43 (41 to 45) | 0.95 (0.89 to 1.01) | 1958 | 13.5 (12.7 to 14.3) | 0.99 (0.93 to 1.06) |
| 3 | 2413 | 1804 | 45 (42 to 47) | 0.94 (0.88 to 1.01) | 1742 | 13.6 (12.9 to 14.5) | 1.02 (0.95 to 1.09) |
| 2 | 2332 | 1758 | 46 (43 to 48) | 0.92 (0.87 to 0.99)f | 1715 | 12.6 (12.1 to 13.2) | 1.09 (1.02 to 1.17)f |
| 1 (low) | 1919 | 1384 | 52 (49 to 54) | 0.87 (0.81 to 0.94)g | 1367 | 12.8 (12.0 to 13.8) | 1.12 (1.04 to 1.21)g |
| Metastatic renal cell carcinoma | 10 763 | 8334 | — | — | 6346 | — | — |
| 5 (high) | 1977 | 1522 | 69 (63 to 77) | 1.00 (Referent) | 1096 | 35.1 (31.3 to 39.0) | 1.00 (Referent) |
| 4 | 2492 | 1938 | 64 (60 to 70) | 0.97 (0.90 to 1.03) | 1432 | 32.3 (30.3 to 34.4) | 1.03 (0.95 to 1.12) |
| 3 | 2421 | 1873 | 60 (55 to 65) | 1.0 (0.93 to 1.06) | 1435 | 30.2 (27.7 to 32.7) | 1.06 (0.98 to 1.15) |
| 2 | 2059 | 1604 | 59 (54 to 64) | 1.04 (0.97 to 1.12) | 1263 | 27.0 (24.7 to 29.1) | 1.15 (1.06 to 1.25)g |
| 1 (low) | 1814 | 1397 | 59 (53 to 63) | 1.03 (0.96 to 1.11) | 1120 | 25.7 (24.3 to 27.9) | 1.20 (1.10 to 1.30)g |
| Advanced urothelial carcinoma | 10 352 | 7532 | 7187 | ||||
| 5 (high) | 2070 | 1526 | 56 (52 to 61) | 1.00 (Referent) | 1429 | 14.9 (13.6 to 15.9) | 1.00 (Referent) |
| 4 | 2536 | 1852 | 56 (53 to 61) | 1.00 (0.94 to 1.07) | 1758 | 13.9 (12.8 to 14.7) | 1.01 (0.95 to 1.09) |
| 3 | 2251 | 1625 | 61 (57 to 64) | 1.00 (0.93 to 1.08) | 1576 | 14.2 (13.1 to 15.2) | 1.06 (0.98 to 1.13) |
| 2 | 2008 | 1432 | 61 (56 to 64) | 0.98 (0.92 to 1.06) | 1399 | 12.7 (11.9 to 13.5) | 1.09 (1.02 to 1.18)f |
| 1 (low) | 1487 | 1097 | 62 (56 to 65) | 1.01 (0.94 to 1.10) | 1025 | 13.0 (12.0 to 13.9) | 1.15 (1.06 to 1.25)g |
| Ovarian cancer | 8760 | 6448 | — | — | 3714 | — | — |
| 5 (high) | 2079 | 1542 | 56 (50 to 63) | 1.00 (Referent) | 832 | 63.5 (60.3 to 70.3) | 1.00 (Referent) |
| 4 | 2053 | 1556 | 52 (49 to 56) | 0.98 (0.92 to 1.05) | 883 | 60.6 (55.1 to 64.1) | 1.03 (0.93 to 1.13) |
| 3 | 1835 | 1358 | 68 (60 to 80) | 0.89 (0.83 to 0.96)g | 775 | 58.5 (53.5 to 66.5) | 1.09 (0.99 to 1.20) |
| 2 | 1555 | 1118 | 63 (55 to 71) | 0.91 (0.84 to 0.98)f | 688 | 57.0 (52.9 to 61.7) | 1.08 (0.98 to 1.20) |
| 1 (low) | 1238 | 874 | 68 (59 to 84) | 0.88 (0.81 to 0.95)g | 536 | 56.3 (51.4 to 63.2) | 1.18 (1.06 to 1.32)g |
| Advanced head and neck cancer | 8722 | 6884 | — | — | 6203 | — | — |
| 5 (high) | 1353 | 1030 | 76 (67 to 87) | 1.00 (Referent) | 913 | 19.0 (17.3 to 20.5) | 1.00 (Referent) |
| 4 | 1862 | 1432 | 73 (65 to 82) | 1.00 (0.93 to 1.09) | 1315 | 16.6 (15.6 to 18.0) | 1.07 (0.98 to 1.16) |
| 3 | 1930 | 1539 | 65 (61 to 73) | 1.06 (0.98 to 1.15) | 1374 | 16.4 (15.6 to 17.6) | 1.12 (1.03 to 1.22)g |
| 2 | 1891 | 1556 | 61 (56 to 63) | 1.13 (1.04 to 1.22)g | 1391 | 15.4 (14.6 to 16.4) | 1.21 (1.11 to 1.31)g |
| 1 (low) | 1686 | 1327 | 67 (63 to 73) | 1.09 (1.01 to 1.19)f | 1210 | 15.5 (15.0 to 16.5) | 1.23 (1.12 to 1.34)g |
| Small cell lung cancer | 8456 | 7015 | — | — | 6174 | — | — |
| 5 (high) | 1104 | 916 | 24 (22 to 26) | 1.00 (Referent) | 805 | 12.0 (10.9 to 12.9) | 1.00 (Referent) |
| 4 | 1825 | 1512 | 25 (23 to 26) | 1.06 (0.98 to 1.15) | 1346 | 11.2 (10.7 to 11.9) | 1.11 (1.02 to 1.21)f |
| 3 | 1859 | 1546 | 25 (24 to 26) | 1.01 (0.93 to 1.10) | 1348 | 11.6 (10.9 to 12.4) | 1.07 (0.98 to 1.16) |
| 2 | 1967 | 1649 | 25 (23 to 26) | 1.03 (0.95 to 1.12) | 1416 | 11.1 (10.5 to 11.9) | 1.11 (1.02 to 1.21)f |
| 1 (low) | 1701 | 1392 | 26 (25 to 27) | 0.97 (0.89 to 1.06) | 1259 | 10.5 (9.9 to 11.1) | 1.16 (1.06 to 1.27)g |
| Advanced hepatocellular carcinoma | 5407 | 3303 | — | — | 3672 | — | — |
| 5 (high) | 793 | 450 | 199 (142 to 334) | 1.00 (Referent) | 513 | 14.9 (13.2 to 17.0) | 1.00 (Referent) |
| 4 | 1102 | 662 | 141 (104 to 212) | 1.04 (0.92 to 1.17) | 746 | 14.4 (13.1 to 15.4) | 1.09 (0.97 to 1.22) |
| 3 | 1149 | 694 | 143 (107 to 182) | 1.04 (0.93 to 1.18) | 791 | 13.5 (12.0 to 15.4) | 1.05 (0.94 to 1.18) |
| 2 | 1139 | 729 | 98 (76 to 133) | 1.08 (0.96 to 1.22) | 776 | 13.3 (12.3 to 14.7) | 1.07 (0.95 to 1.19) |
| 1 (low) | 1224 | 768 | 107 (79 to 140) | 1.11 (0.98 to 1.24) | 846 | 13.7 (12.6 to 15.3) | 1.10 (0.99 to 1.23) |
| Advanced endometrial cancer | 4821 | 3551 | — | — | 2070 | — | — |
| 5 (high) | 1012 | 765 | 76 (70 to 82) | 1.00 (Referent) | 403 | 56.0 (49.9 to 67.7) | 1.00 (Referent) |
| 4 | 1065 | 805 | 78 (74 to 84) | 0.93 (0.85 to 1.03) | 456 | 49.2 (44.2 to 54.0) | 1.14 (1.00 to 1.31)f |
| 3 | 943 | 676 | 86 (82 to 94) | 0.84 (0.76 to 0.93)g | 400 | 48.1 (43.1 to 57.6) | 1.22 (1.06 to 1.40)g |
| 2 | 902 | 664 | 80 (74 to 90) | 0.87 (0.78 to 0.96)g | 385 | 49.8 (42.9 to 59.5) | 1.21 (1.05 to 1.40)g |
| 1 (low) | 899 | 641 | 91 (83 to 98) | 0.77 (0.70 to 0.86)g | 426 | 37.4 (33.0 to 44.0) | 1.44 (1.26 to 1.66)g |
| Blood cancers | |||||||
| Multiple myeloma | 11 760 | 10 622 | — | — | 4681 | — | — |
| 5 (high) | 2732 | 2486 | 32 (31 to 33) | 1.00 (Referent) | 958 | 77.4 (72.5 to 83.4) | 1.00 (Referent) |
| 4 | 2718 | 2476 | 33 (32 to 34) | 0.95 (0.90 to 1.01) | 1059 | 70.4 (66.5 to 74.5) | 1.12 (1.02 to 1.22)f |
| 3 | 2427 | 2196 | 32 (31 to 33) | 0.96 (0.90 to 1.01) | 995 | 64.4 (60.5 to 69.4) | 1.19 (1.09 to 1.30)g |
| 2 | 2108 | 1874 | 33 (31 to 34) | 0.92 (0.87 to 0.98)g | 905 | 58.4 (53.5 to 62.4) | 1.26 (1.15 to 1.38)g |
| 1 (low) | 1775 | 1590 | 32 (31 to 34) | 0.93 (0.87 to 0.99)f | 764 | 59.5 (55.5 to 64.4) | 1.37 (1.25 to 1.51)g |
| Acute myeloid leukemia | 8609 | 7659 | — | — | 5056 | — | — |
| 5 (high) | 1908 | 1700 | 7 (7 to 8) | 1.00 (Referent) | 1107 | 14.9 (13.8 to 16.1) | 1.00 (Referent) |
| 4 | 2043 | 1811 | 8 (7 to 9) | 0.96 (0.89 to 1.02) | 1211 | 14.9 (13.7 to 16.3) | 1.01 (0.93 to 1.09) |
| 3 | 1936 | 1710 | 8 (7 to 9) | 0.91 (0.85 to 0.97)g | 1154 | 14.0 (12.7 to 15.0) | 1.11 (1.02 to 1.21)f |
| 2 | 1571 | 1404 | 7 (6 to 8) | 0.93 (0.86 to 1.00)f | 910 | 15.7 (14.2 to 17.5) | 1.13 (1.03 to 1.23)g |
| 1 (low) | 1151 | 1034 | 7 (7 to 8) | 0.87 (0.81 to 0.94)g | 674 | 14.7 (13.3 to 17.1) | 1.24 (1.12 to 1.36)g |
| Diffuse large B-cell lymphoma | 8514 | 7906 | — | — | 2926 | — | — |
| 5 (high) | 1881 | 1736 | 21 (20 to 22) | 1.00 (Referent) | 589 | 101.9 (90.4 to 110.5) | 1.00 (Referent) |
| 4 | 2032 | 1901 | 22 (21 to 23) | 0.97 (0.91 to 1.04) | 674 | 96.5 (89.3 to 103.7) | 1.07 (0.95 to 1.19) |
| 3 | 1806 | 1684 | 23 (22 to 24) | 0.93 (0.87 to 0.99)f | 613 | 100.1 (89.4 to 111.2) | 1.05 (0.94 to 1.18) |
| 2 | 1581 | 1474 | 23 (22 to 24) | 0.91 (0.85 to 0.98)g | 592 | 85.1 (78.7 to 95.5) | 1.20 (1.07 to 1.34)g |
| 1 (low) | 1214 | 1111 | 23 (22 to 25) | 0.84 (0.78 to 0.90)g | 458 | 88.4 (69.7 to 101.1) | 1.33 (1.18 to 1.50)g |
| Chronic lymphocytic leukemia | 7945 | 7665 | — | — | 2346 | — | — |
| 5 (high) | 1855 | 1780 | 552 (499 to 606) | 1.00 (Referent) | 478 | —d | 1.00 (Referent) |
| 4 | 1944 | 1880 | 508 (465 to 561) | 1.04 (0.98 to 1.11) | 571 | 124.5 (116.5 to 136.4) | 1.12 (0.99 to 1.27) |
| 3 | 1678 | 1620 | 403 (352 to 455) | 1.09 (1.02 to 1.17)g | 498 | 118.5 (108.5 to 135.5) | 1.25 (1.10 to 1.42)g |
| 2 | 1389 | 1346 | 354 (305 to 407) | 1.07 (0.99 to 1.15) | 452 | 109.5 (103.4 to 122.5) | 1.33 (1.17 to 1.51)g |
| 1 (low) | 1079 | 1039 | 342 (271 to 413) | 1.10 (1.02 to 1.19)f | 347 | 113.5 (104.5 to 126.5) | 1.42 (1.23 to 1.63)g |
| Follicular lymphoma | 5465 | 3944 | — | — | 1066 | — | — |
| 5 (high) | 1280 | 856 | 93 (79 to 121) | 1.00 (Referent) | 172 | —dd | 1.00 (Referent) |
| 4 | 1342 | 966 | 66 (61 to 77) | 1.10 (1.01 to 1.21)f | 250 | 139.4 (132.7-e) | 1.29 (1.06 to 1.57)f |
| 3 | 1167 | 846 | 58 (54 to 63) | 1.09 (0.99 to 1.20) | 242 | 131.4 (121.7-e) | 1.50 (1.24 to 1.83)g |
| 2 | 944 | 725 | 56 (53 to 64) | 1.15 (1.04 to 1.27)g | 218 | 128.7 (115.6-e) | 1.54 (1.26 to 1.88)g |
| 1 (low) | 732 | 551 | 57 (50 to 69) | 1.14 (1.03 to 1.28)f | 184 | 130.3 (111.2-e) | 1.84 (1.49 to 2.27)g |
| Mantle cell lymphoma | 4732 | 4080 | — | — | 1779 | — | — |
| 5 (high) | 1085 | 922 | 36 (34 to 42) | 1.00 (Referent) | 362 | 99.0 (88.8 to 108.5) | 1.00 (Referent) |
| 4 | 1189 | 1028 | 34 (33 to 38) | 1.03 (0.94 to 1.13) | 450 | 83.4 (79.0 to 92.7) | 1.22 (1.06 to 1.40)g |
| 3 | 1033 | 895 | 34 (32 to 36) | 1.09 (0.99 to 1.19) | 390 | 80.1 (71.5 to 85.9) | 1.24 (1.08 to 1.43)g |
| 2 | 831 | 730 | 36 (33 to 39) | 1.07 (0.97 to 1.18) | 335 | 74.5 (67.9 to 85.2) | 1.33 (1.15 to 1.55)g |
| 1 (low) | 594 | 505 | 38 (34 to 41) | 1.00 (0.90 to 1.12) | 242 | 72.9 (62.4 to 79.5) | 1.41 (1.20 to 1.67)g |
US Census Bureau block group data from the American Community Survey (2015-2019) were used to measure area-level socioeconomic status per the Yost Index (incorporating income, home values, rental costs, poverty, blue-collar employment, unemployment, and education information). Socioeconomic status quintiles were determined from the US population, and then applied to patients based on their latest residential address. CI = confidence interval; HR = hazard ratio.
The index date was defined as initial, advanced, or metastatic diagnosis (depending on specific cancer diagnosis; see Supplementary Table 2, available online). Patients were followed from index diagnosis until the event of interest (treatment initiation or death), last confirmed activity, or end of the study period (December 2022).
Models are adjusted for clinical characteristics (ie, age, sex, Eastern Cooperative Oncology Group [ECOG] performance status, stage, and cancer diagnosis) and year of diagnosis, as appropriate.
Median survival cannot be computed because the probability of survival exceeds 50% at the longest time point.
Could not estimate the upper bound of the confidence interval.
P <.05;
P <.01.
Lower area-level socioeconomic status was associated with reduced survival (Figure 1, B). Median real-world overall survival was 8.1 months lower among patients living in low– vs high–socioeconomic status areas (21.4 vs 29.5 months, HR = 1.20, 95 CI = 1.18 to 1.22, P < .001) (Table 2). In analyses stratified by cancer type, survival inequities between low– and high–socioeconomic status neighborhoods were observed in 19 cancer types. Survival inequities were most pronounced among patients with follicular lymphoma (130.3 months vs not reached, HR = 1.84, 95% CI = 1.49 to 2.27), advanced endometrial cancer (37.4 vs 56.0 months HR = 1.44, 95% CI = 1.26 to 1.66), and advanced melanoma (44.7 vs 74.8 months, HR = 1.43, 95% CI = 1.30 to 1.57]) (all P < .001).
Treatment and survival inequities associated with area-level socioeconomic status persisted in the COVID-19 era (Figure 2). Specifically, among patients diagnosed with cancer on or after March 2020, living in low– vs high–socioeconomic status areas was associated with lower treatment initiation rates (HR = 0.88, 95% CI = 0.86 to 0.91, P < .001) and worse survival (HR = 1.18, 95% CI = 1.14 to 1.23, P < .001) (Table 3).
Figure 2.
Unadjusted Kaplan-Meier estimates of treatment initiation and real-world overall survival, by area-level socioeconomic status among patients receiving care for 20 common cancers during the COVID-19 pandemic, March 2020 to December 2022. US Census Bureau block group data from the American Community Survey (2015-2019) were used to measure area-level socioeconomic status per the Yost Index (incorporating income, home values, rental costs, poverty, blue-collar worker, unemployment, and education information). Socioeconomic status quintiles were determined from the US population, and then applied to patients based on their latest residential address. The index date was defined as initial, advanced, or metastatic diagnosis (depending on specific cancer diagnosis; see Supplementary Table 2, available online). Patients were followed from index diagnosis until the event of interest (treatment initiation or death), last confirmed activity, or end of the study period (December 2022). SES = socioeconomic status.
Table 3.
Association of area-level socioeconomic status with treatment initiation and real-world overall survival during the pre– and post–COVID-19 era
| Total, No. | Treatment initiation |
Real-world overall survival |
|||||
|---|---|---|---|---|---|---|---|
| Area-level socioeconomic status a | Treated, No. | Median time to treatment initiation (95% CI) b | HR (95% CI) b , c | Deaths, No. | Median (95% CI) survival, mo b | Adjusted HR (95% CI) b , c | |
| Before COVID-19d | 218 061 | 170 261 | — | — | 146 655 | — | — |
| 5 (high) | 44 183 | 34 791 | 49 (49 to 50) | 1.00 (Referent) | 28 005 | 30.4 (29.7 to 31.0) | 1.00 (Referent) |
| 4 | 50 352 | 39 486 | 49 (48 to 49) | 0.99 (0.97 to 1.00)f | 33 497 | 26.6 (26.2 to 27.1) | 1.06 (1.04 to 1.08)g |
| 3 | 46 425 | 36 358 | 49 (48 to 50) | 0.98 (0.97 to 1.00)f | 31 543 | 24.8 (24.4 to 25.3) | 1.11 (1.09 to 1.12)g |
| 2 | 42 176 | 32 707 | 50 (49 to 50) | 0.97 (0.96 to 0.99)g | 29 209 | 22.8 (22.4 to 23.3) | 1.16 (1.14 to 1.18)g |
| 1 (low) | 34 925 | 26 919 | 51 (50 to 52) | 0.96 (0.95 to 0.98)g | 24 401 | 21.8 (21.4 to 22.3) | 1.20 (1.18 to 1.22)g |
| After COVID-19e | 68 424 | 54 006 | — | — | 25 030 | — | — |
| 5 (high) | 13 466 | 10 627 | 37 (36 to 38) | 1.00 (Referent) | 4520 | 25.4 (24.4 to 26.4) | 1.00 (Referent) |
| 4 | 15 648 | 12 414 | 38 (37 to 39) | 0.98 (0.95 to 1.00) | 5589 | 23.5 (22.5 to 24.7) | 1.04 (1.00 to 1.08) |
| 3 | 14 622 | 11 545 | 39 (38 to 40) | 0.95 (0.93 to 0.98)g | 5326 | 22.4 (21.5 to 23.4) | 1.06 (1.02 to 1.11)g |
| 2 | 13 385 | 10 631 | 40 (39 to 41) | 0.94 (0.91 to 0.96)g | 5142 | 20.7 (19.9 to 21.4) | 1.16 (1.12 to 1.21)g |
| 1 (low) | 11 303 | 8789 | 42 (41 to 43) | 0.88 (0.86 to 0.91)g | 4453 | 19.8 (19.0 to 20.6) | 1.18 (1.14 to 1.23)g |
US Census Bureau block group data from the American Community Survey (2015-2019) were used to measure area-level socioeconomic status per the Yost Index (incorporating income, home values, rental costs, poverty, blue-collar employment, unemployment, and education information). Socioeconomic status quintiles were determined from the US population, and then applied to patients based on their latest residential address. CI = confidence interval; HR = hazard ratio.
Index diagnosis was defined as initial, advanced, or metastatic diagnosis, depending on cancer type (as listed in Supplementary Table 2, available online). Patients were followed from index diagnosis until the event of interest (treatment initiation or death), last confirmed activity, or end of the study period (December 2022).
Models are adjusted for clinical characteristics (ie, age, sex, Eastern Cooperative Oncology Group [ECOG] performance status, stage, and cancer diagnosis) and year of diagnosis, as appropriate.
Diagnosed with cancer between January 2011 and December 2019.
Diagnosed with cancer on or after March 2020.
P < .05;
P < .01.
We also examined associations stratified by race and ethnicity (Supplementary Table 4, available online). Although area-level socioeconomic status was associated with survival inequities among non-Latinx White and non-Latinx Black patients, these inequities were not observed in other racial and ethnic groups. In our sensitivity analysis, observed associations of area-level socioeconomic status on treatment initiation and overall survival remained similar in fully adjusted models (Supplementary Table 5, available online).
Discussion
In this up-to-date evaluation using a nationwide EHR-derived database of patients receiving care for 20 common cancers, lower area-level socioeconomic status was statistically significantly associated with lower rates of treatment initiation in 9 cancers and poorer survival in 19 cancers. To our knowledge, this is the first study to examine inequities in area-level socioeconomic status and cancer outcomes during the COVID-19 pandemic. Among patients diagnosed after March 2020, amidst the pandemic, the magnitude of area-level socioeconomic status inequities persisted, where patients in low–socioeconomic status areas were 14% less likely to initiate treatment and 18% more likely to die than those in high–socioeconomic status areas. These findings lend credence to previous reports of delayed cancer diagnosis and treatment amidst the pandemic, especially in marginalized, low–socioeconomic status neighborhoods (10,11), whose residents experienced more barriers to cancer care before the pandemic (20) and increased exposure to infection and death during the pandemic (21,22).
Area-level socioeconomic status is an important driver of cancer inequities in the United States. For example, among patients diagnosed with the 4 leading causes of US cancer death (lung, colon, pancreatic, and breast cancer) (1,2), we found inequities: Patients diagnosed with these cancers living in low–socioeconomic status neighborhoods were 6% to 10% less likely to receive treatment and 15% to 25% more likely to die than those living in high–socioeconomic status areas. We observed more modest inequities in treatment initiation, but in every cancer type where treatment inequities existed, there were also statistically significant inequities in survival, suggesting that more patients are dying in low–socioeconomic status neighborhoods partially because of inequitable and delayed access to cancer treatment (23).
In 11 cancer types, we did not observe inequities in treatment initiation. Yet, in 10 of these cancers and in grouped analyses across all studied cancer types, there were statistically significant mortality inequities. Therefore, evaluating other health-care inequities in these cancer types, such as differences in guideline-concordant care or delays in genetic testing or subsequent lines of therapy (20), may elucidate additional health system factors that may mediate mortality inequities associated with area-level socioeconomic status.
Beyond treatment inequities, area-level socioeconomic status influences cancer outcomes through multiple structural mechanisms. In marginalized, low–socioeconomic status neighborhoods, for example, fewer health, financial, and supportive resources and lower trust in health-care institutions (because of historical/contemporary medical mistreatment) may result in worse access to cancer treatment and reduced survival (7,20). Therefore, to improve cancer care in low–socioeconomic status neighborhoods, policymakers should consider multilevel interventions, ranging from point-of-care improvements (eg, reimbursement for navigation services) to federal interventions that address persistent underinvestments in health-care systems serving historically marginalized areas (eg, financial incentives to provide high-quality care in these neighborhoods).
In our evaluation of race and ethnicity-specific cohorts, we found that lower area-level socioeconomic status was associated with reduced survival among non-Latinx White and non-Latinx Black patients but not among Latinx or non-Latinx Asian patients. Our findings contrast with previous research that has demonstrated associations between area-level socioeconomic status and cancer survival in all these racial and ethnic groups (24). Our inability to detect inequities in certain minoritized groups may stem from our sample’s limited statistical power and generalizability. For instance, Asian and Latinx patients make up 12.8% and 5.8% of all patients diagnosed with cancer in the United States, respectively (12), but they represent just 4.8% and 2.2%, respectively, of our sample. Furthermore, Latinx and Asian patients are extremely heterogeneous in terms of barriers to health and health care (25-27). Latinx and Asian patients represent people whose families descend from dozens of countries with distinct historical trajectories—and subsequent structural barriers—in the United States (25). Aggregating these heterogeneous populations may obfuscate inequities (26,27). Nonetheless, to inform efforts to reduce persistent racial/ethnic inequities (20), future research should also evaluate to what extent area-level socioeconomic status mediates the association between race and ethnicity and cancer outcomes, including access to costly treatments indicated for later lines of therapy (28,29).
This study has limitations. First, findings may not be generalizable to other cancer types and to earlier stages of the 12 cancers for which the dataset included only patients with advanced or metastatic disease. Second, we are unable to capture nonsystemic treatment (eg, surgery or radiation), which likely attenuates the inequities observed. This limitation is most relevant to evaluations of patients with solid cancers, but for 12 of 14 solid cancers examined in this study, we followed patients beginning at their advanced/metastatic diagnosis, when surgery or radiation is less likely to be optimal first-line therapy, limiting the risk of attenuation. Third, lower statistical power and inappropriate aggregations (26,27) may limit the ability to detect inequities among Latinx and non-Latinx Asian patients. Fourth, we are unable to measure individual-level socioeconomic status. Fifth, area-level socioeconomic status was based on the most recent residential address in the EHR and may not represent area-level socioeconomic status at index. Sixth, unadjusted estimates of pooled analyses (all cancer types) should be interpreted with caution, given the wide range of expected values by cancer type. This being said, all measures of association from pooled analyses account for cancer type as a confounder. Finally, we were unable to examine several important factors that may explain observed cancer inequities by area-level socioeconomic status, including financial toxicity (ie, a function of individual-level income, wealth, and health insurance coverage), provider/health system indicators (eg, patient–health-care professional racial and ethnic and language concordance or access to cancer navigation services), and structural racism (eg, residential segregation). Missing data and unmeasured confounders may also affect observed inequities.
Evaluating inequities in cancer outcomes by area-level socioeconomic status is 1 way to represent the culmination of adverse social determinants of health that result from societal injustices in marginalized neighborhoods. Lower area-level socioeconomic status was associated with lower treatment rates and worse survival among patients with cancer—inequities that persisted amidst the COVID-19 pandemic. To reduce inequities in cancer care and survival, efforts that target marginalized, lower–socioeconomic status neighborhoods are necessary.
Supplementary Material
Acknowledgements
This research was presented, in part, at the annual meeting of the American Association for Cancer Research in April 2022. The study sponsor had a role in the study design in the collection, analysis, and interpretation of data, the writing of the manuscript, and the decision to submit the manuscript for publication. We would like to thank Hannah Gilham and Darren Johnson (Flatiron Health) for editorial assistance.
Contributor Information
Jenny S Guadamuz, Flatiron Health, New York, NY, USA; Division of Health Policy and Management, University of California, Berkeley, School of Public Health, Berkeley, CA, USA; Program on Medicines and Public Health, University of Southern California School of Pharmacy, Los Angeles, CA, USA.
Xiaoliang Wang, Flatiron Health, New York, NY, USA.
Cleo A Ryals, Flatiron Health, New York, NY, USA.
Rebecca A Miksad, Flatiron Health, New York, NY, USA; Department of Hematology and Oncology, Boston University School of Medicine, Boston, MA, USA.
Jeremy Snider, Flatiron Health, New York, NY, USA.
James Walters, Flatiron Health, New York, NY, USA.
Gregory S Calip, Flatiron Health, New York, NY, USA; Program on Medicines and Public Health, University of Southern California School of Pharmacy, Los Angeles, CA, USA.
Data availability
The deidentified patient-level EHR-derived data that support the findings of this study have been originated by Flatiron Health, Inc. Requests for data sharing by license or by permission for the specific purpose of replicating results in this manuscript can be submitted to dataaccess@flatiron.com. The Census block group data that support the findings of this study are openly available at https://www.census.gov/programs-surveys/acs.
Author contributions
Jenny S. Guadamuz, PhD (Formal analysis; Methodology; Project administration; Visualization; Writing—original draft; Writing—review & editing), Xiaoliang Wang, MPH, PhD (Conceptualization; Formal analysis; Methodology; Writing—review & editing), Cleo A. Ryals, PhD (Conceptualization; Writing—review & editing), Rebecca A. Miksad, MD, MPH (Writing—review & editing), Jeremy Snider, MPH, PhD (Writing—review & editing), James Walters, BA (Conceptualization; Writing—review & editing), Gregory S. Calip, PharmD, MPH, PhD (Conceptualization; Methodology; Supervision; Writing—review & editing).
Funding
This work was supported by Flatiron Health, Inc, which is an independent member of the Roche Holding AG.
Conflicts of interest
All authors report current or previous employment with Flatiron Health, Inc, which is an independent member of the Roche Holding AG, and stock ownership in Roche. Dr Calip reports research grants from Pfizer. Dr Miksad and Dr Snider report equity ownership in Flatiron Health, Inc (initiated before acquisition by Roche in 2018).
References
- 1. Shavers VL. Measurement of socioeconomic status in health disparities research. J Natl Med Assoc. 2007;99(9):1013-1023. [PMC free article] [PubMed] [Google Scholar]
- 2. Yost K, Perkins C, Cohen R, Morris C, Wright W.. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes Control. 2001;12(8):703-711. doi: 10.1023/a:1011240019516 [DOI] [PubMed] [Google Scholar]
- 3. Boscoe FP, Liu B, Lee F.. A comparison of two neighborhood-level socioeconomic indexes in the United States. Spat Spatiotemporal Epidemiol. 2021;37:100412. doi: 10.1016/j.sste.2021.100412 [DOI] [PubMed] [Google Scholar]
- 4. Ellis L, Canchola AJ, Spiegel D, Ladabaum U, Haile R, Gomez SL.. Racial and ethnic disparities in cancer survival: the contribution of tumor, sociodemographic, institutional, and neighborhood characteristics. J Clin Oncol. 2018;36(1):25-33. doi: 10.1200/JClinOncol.2017.74.2049 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Shariff-Marco S, Yang J, John EM, et al. Impact of neighborhood and individual socioeconomic status on survival after breast cancer varies by race/ethnicity: the Neighborhood and Breast Cancer Study. Cancer Epidemiol Biomarkers Prev. 2014;23(5):793-811. doi: 10.1158/1055-9965.EPI-13-0924 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. 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(6):e183146. doi: 10.1001/jamanetworkopen.2018.3146 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Cheng E, Soulos PR, Irwin ML, et al. Neighborhood and individual socioeconomic disadvantage and survival among patients with nonmetastatic common cancers. JAMA Netw Open 2021;4(12):e2139593. doi: 10.1001/jamanetworkopen.2021.39593 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Dong W, Bensken WP, Kim U, et al. Variation in and factors associated with US county-level cancer mortality, 2008-2019. JAMA Netw Open. 2022;5(9):e2230925. doi: 10.1001/jamanetworkopen.2022.30925 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Patel MI, Ferguson JM, Castro E, et al. Racial and ethnic disparities in cancer care during the COVID-19 pandemic. JAMA Netw Open. 2022;5(7):e2222009. doi: 10.1001/jamanetworkopen.2022.22009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Mullangi S, Aviki EM, Chen Y, Robson M, Hershman DL.. Factors associated with cancer treatment delay among patients diagnosed with COVID-19. JAMA Netw Open. 2022;5(7):e2224296. doi: 10.1001/jamanetworkopen.2022.24296 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Amram O, Robison J, Amiri S, Pflugeisen B, Roll J, Monsivais P.. Socioeconomic and racial inequities in breast cancer screening during the COVID-19 pandemic in Washington State. JAMA Netw Open. 2021;4(5):e2110946. doi: 10.1001/jamanetworkopen.2021.10946 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. National Cancer Institute . Cancer Statistics Explorer Network. https://seer.cancer.gov/statistics-network/explorer/application.html. Updated 2023. Accessed January 5, 2023.
- 13. Birnbaum B, Nussbaum N, Seidl-Rathkopf K, et al. Model-assisted cohort selection with bias analysis for generating large-scale cohorts from the EHR for oncology research. arXiv. Preprint 2020. doi:arXiv:2001.09765.
- 14. Ma X, Long L, Moon S, Adamson Blythe JS, Baxi SS.. Comparison of population characteristics in real-world clinical oncology databases in the US: Flatiron Health, SEER, and NPCR . medRxiv. Preprint 2020. doi: 10.1101/2020.03.16.20037143. [DOI] [Google Scholar]
- 15. Zhang Q, Gossai A, Monroe S, Nussbaum NC, Parrinello CM.. Validation analysis of a composite real-world mortality endpoint for patients with cancer in the United States. Health Serv Res. 2021;56(6):1281-1287. doi: 10.1111/1475-6773.13669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Institute of Medicine (US) Committee on Understanding and Eliminating Racial and Ethnic Disparities in Health Care, Smedley BD, Stith AY, Nelson AR, eds.. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, DC: National Academies Press (US); 2003. [PubMed]
- 17. Lê Cook B, McGuire TG, Zaslavsky AM.. Measuring racial/ethnic disparities in health care: methods and practical issues. Health Serv Res. 2012;47(3 Pt 2):1232-1254. doi: 10.1111/j.1475-6773.2012.01387.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. McGuire TG, Alegria M, Cook BL, Wells KB, Zaslavsky AM.. Implementing the Institute of Medicine definition of disparities: an application to mental health care. Health Serv Res. 2006;41(5):1979-2005. doi: 10.1111/j.1475-6773.2006.00583.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. VanderWeele TJ, Robinson WR.. On causal interpretation of race in regressions adjusting for confounding and mediating variables. Epidemiology. 2014;25(4):473-484. doi: 10.1097/EDE.0000000000000105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. American Association for Cancer Research. AACR Cancer Disparities Progress Report. 2020. http://www.CancerDisparitiesProgressReport.org/. Accessed January 5, 2023.
- 21. Feldman JM, Bassett MT.. Variation in COVID-19 Mortality in the US by Race and Ethnicity and Educational Attainment. JAMA Netw Open. 2021;4(11):e2135967. doi: 10.1001/jamanetworkopen.2021.35967 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Henley SJ, Dowling NF, Ahmad FB, Ellington TD, Wu M, Richardson LC.. COVID-19 and other underlying causes of cancer deaths — United States, January 2018–July 2022. MMWR Morb Mortal Wkly Rep 2022;71(50):1583-1588. doi: 10.15585/mmwr.mm7150a3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Cone EB, Marchese M, Paciotti M, et al. Assessment of time-to-treatment initiation and survival in a cohort of patients with common cancers. JAMA Netw Open. 2020;3(12):e2030072. doi: 10.1001/jamanetworkopen.2020.30072 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. 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: 10.1155/2017/2819372 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Guadamuz JS, Kapoor K, Lazo M, et al. Understanding immigration as a social determinant of health: cardiovascular disease in Hispanics/Latinos and South Asians in the United States. Curr Atheroscler Rep. 2021;23(6):25. doi: 10.1007/s11883-021-00920-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Zamora SM, Pinheiro PS, Gomez SL, et al. Disaggregating Hispanic American cancer mortality burden by detailed ethnicity. Cancer Epidemiol Biomarkers Prev. 2019;28(8):1353-1363. doi: 10.1158/1055-9965.EPI-18-0872 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Yi SS, Kwon SC, Suss R, et al. The mutually reinforcing cycle of poor data quality and racialized stereotypes that shapes Asian American health. Health Aff (Millwood). 2022;41(2):296-303. doi: 10.1377/hlthaff.2021.01417 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Reyes C, Engel‐Nitz NM, DaCosta Byfield S, et al. Cost of disease progression in patients with metastatic breast, lung, and colorectal cancer. Oncologist. 2019;24(9):1209-1218. doi: 10.1634/theoncologist.2018-0018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Leighl NB, Nirmalakumar S, Ezeife DA, Gyawali B.. An arm and a leg: the rising cost of cancer drugs and impact on access. Am Soc Clin Oncol Educ Book. 2021;41:e1-e12. doi: 10.1200/EDBK_100028 [DOI] [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
The deidentified patient-level EHR-derived data that support the findings of this study have been originated by Flatiron Health, Inc. Requests for data sharing by license or by permission for the specific purpose of replicating results in this manuscript can be submitted to dataaccess@flatiron.com. The Census block group data that support the findings of this study are openly available at https://www.census.gov/programs-surveys/acs.