Abstract
Purpose
We examined the relationship between travel burden for surgical cancer care and rurality, geographic bypass of the nearest surgical facility, cancer type, and mortality outcomes.
Methods
Using Medicare claims and enrollment data (2016–2018) from beneficiaries with cancer of the colon, rectum, lung, or pancreas, we measured travel times to: the nearest surgical facility and facility used. For those who bypassed the nearest, we examined travel time and rurality in relation to surgical rates. Using multivariable regression modeling, we estimated associations of bypass with 90‐day postoperative‐ and one‐year mortality; rurality was examined as an effect modifier.
Findings
Among 211,025 beneficiaries with cancer, 25.5% resided in non‐metropolitan areas. About 66% of metropolitan/micropolitan, and 78% of small town/rural patients bypassed their closest facility. Increasing rurality was significantly associated with increased likelihood of bypass (Referent = metropolitan, OR; 95%CI: micropolitan 1.10; 1.04–1.16, small town/rural 2.08; 1.96–2.20. Bypassing the nearest facility was associated with decreased likelihood of both 90‐day postoperative mortality (OR = 0.79; 95%CI 0.74–0.85) and 1‐year mortality (OR = 0.81; 95%CI 0.77–0.86). The greatest decrement in 1‐year mortality was for pancreatic cancer across all rural‐urban categories (OR; 95%CI: metropolitan 0.63; 0.53–0.76; micropolitan 0.53; 0.29–0.97); small town/rural 0.46; 0.25–0.86).
Conclusions
Most Medicare beneficiaries with lung, colon, rectal, or pancreatic cancer bypassed the closest facility providing surgical cancer care, especially rural patients. Bypassing was associated with a lower likelihood of 90‐day postoperative, and 1‐year mortality. Understanding determinants of bypassing, particularly among rural patients, may reveal potential mechanisms to improve cancer outcomes and reduce rural cancer disparities.
Keywords: bypass, cancer, rural, surgical oncology, travel burden
INTRODUCTION
Access to cancer care varies significantly by rurality, race/ethnicity, income, and other geographic and socio‐demographic factors. 1 , 2 , 3 , 4 Differences in access are particularly stark across rural settings, where patients travel five times longer to reach their nearest practicing oncologist compared to urban patients. Further exacerbating disparities, rural patients similarly lack access to highly specialized care, such as oncologic surgical care, which is associated with better outcomes, 5 , 6 with only 7% of rural residents living within 1 h of an NCI‐designated cancer center (compared with 24% of urban residents). 1 , 2 Our prior work suggests that cancer patients differentially use some of the most specialized cancer care settings, based on urban–rural residence, proximity to care, cancer type, and comorbidities, with race modifying some of these effects, with a higher proportion of urban Black patients using specialized settings and a lower proportion of rural Black patients doing so, relative to White counterparts. 3 Although increased travel time decreases use of specialized cancer centers 4 and sub‐specialist oncologists, 5 the phenomenon of patients bypassing their closest facility to access similar services at a more distant facility has been observed, but not well studied. 7 Specifically, although a few studies have described bypassing facilities, 8 , 9 , 10 such as for higher volume hospitals, there has not been a focus on understanding associations of bypassing with surgical outcomes for cancer patients. Ultimately, understanding bypassing patterns and associated cancer outcomes will provide insights into both health equity—who can access a range of facilities based on preference or clinical need—and inform effective health care delivery models, such as hub‐and‐spoke.
The objective of this study was to examine the role of rurality and cancer type on travel burden for surgical cancer care, and to evaluate potential implications for mortality outcomes. We specifically examined the effects of bypassing one's nearest surgical facility to seek care further afield to identify aspects of rurality and travel time that may influence surgical care and outcomes for cancer patients in the Medicare population.
METHODS
Study population and data
We used 100% Medicare enrollment and claims data from October 1, 2015, to December 31, 2018, to derive four cancer cohorts comprising beneficiaries diagnosed with incident cancer of the lung, colon, rectum, or pancreas, including those who did and did not undergo surgery. These cancers were selected as representative of varying degrees of high versus low incidence. Cancer‐directed surgeries for these cancers also differ in how surgical volume affects outcomes with well‐documented associations for pancreatic cancer surgeries. 11 , 12 , 13 We specifically used the beneficiary Denominator (enrollment), Medicare Provider Analysis and Review (MedPAR), outpatient, and 100% Carrier files. 14 Inclusion criteria were as follows: continuous enrollment in Parts A and B, age‐based eligibility, ages 65–99, and no HMO enrollment. We excluded beneficiaries with missing or invalid ZIP codes and those from outside of the contiguous United States.
Ascertainment of cancer cohorts
Incident cancers were ascertained using the following criteria: (1) at least one cancer diagnosis code (during a hospitalization or outpatient visit and a Current Procedural Terminology [CPT] code related to chemotherapy, radiation therapy, or cancer‐directed surgery); (2) at least two ICD‐10 (International Classification of Diseases, version 10) diagnosis codes on different dates within 12 months following a CPT code for cancer‐directed biopsy; or (3) one cancer ICD diagnosis code within 14 days of a diagnosis or procedure code for cancer‐related symptoms and another within the 12 months following the symptom. This ascertainment algorithm approximates incident cancers, particularly using a 6‐month look‐back period to exclude prevalent cancers. 15 , 16 We additionally excluded individuals with ICD‐10 codes for metastatic disease within 3 months of diagnosis. With this approach, we identified cancer‐specific claims indicating one of the four cancers of interest. We also used data from the 2018 American Hospital Association's Annual Hospital Survey 17 to characterize facilities.
Socio‐demographic and clinical characteristics
Attributes of beneficiaries were derived from the Denominator file, specifically: age (years), sex, race (White, non‐White), and dual‐eligibility status. A binary race categorization was necessary because of small numbers of observations and Centers for Medicare and Medicaid Services (CMS) data suppression rules. As recorded in CMS beneficiary data, we also categorized individuals as Hispanic and non‐Hispanic. We were also limited to the method of ascertainment of race and ethnicity in CMS data: specifically, self‐reported, hospital‐ascribed, surname‐based algorithms, or unknown. Additionally, we linked ZIP code level (ZIP Code Tabulation Areas) Rural Urban Commuting Area categorization (four‐tier: metropolitan, micropolitan, small town, rural, with small town and rural combined due to small cell sizes) to each beneficiary 18 , 19 and attributed each individual to their census region of residence 20 as well as to the area deprivation index (ADI) using nine‐digit ZIP code of beneficiaries available in Medicare enrollment data during our look‐back period. 21 We captured six significant comorbidities through a 1‐year look‐back of claims: diabetes, myocardial infarction, congestive heart failure, stroke/transient ischemic attack, chronic obstructive pulmonary disease, and end‐stage renal disease. We also calculated risk scores from the CMS–HCC (Hierarchical Condition Categories), per CMS. 22
Definition of key variables
Travel time
Travel time in minutes was estimated using the shortest travel path in minutes for road‐based travel between the population centroid of the Medicare beneficiary's ZIP code of residence (origin) and the population centroid of the billing ZIP code (destination) for that service (facility or physician) through linkages with data from the American Hospital Association. Travel time was calculated using ArcGIS Network Analyst and then refined by Google Maps API. 23 , 24 All travel time estimates were based on one‐way travel and irrespective of any specific travel restrictions, including, but not limited to, left‐turn restrictions and/or one‐way streets (which have the potential to add to overall travel time). Travel time was calculated from the population centroid of beneficiaries’ ZIP codes of residence to the nearest surgical facility that performed at least one cancer‐directed surgical resection for a given beneficiary's cancer type 25 in the last 3 years, based on Medicare claims. For those beneficiaries who received surgery, travel time was also calculated by the ZIP code of the facility used.
Surgical procedures
We identified beneficiaries treated with surgical procedures that are often used as definitive, curative‐intent therapy of lung cancer (e.g., lobectomy and segmentectomy), colon cancer (colectomy), rectal cancer (proctectomy), and pancreas cancer (pancreaticoduodenectomy and distal pancreatectomy). Procedure codes are enumerated in prior publications. 6 , 16 , 26
Bypass
For surgical patients, we assessed whether they bypassed their nearest surgical facility, as defined above, to receive the surgery at a facility that was not their nearest facility. We created a flag to denote bypass (yes/no) and also calculated additional travel time, defined as the difference in travel time from the used facility to the nearest facility. Because ZIP code was the smallest geographic unit captured, more than one hospital can be in the ZIP code, in which case the nearest and the used facility would be the same, as no difference in travel time would be calculable. Facility characteristics for the facility used among those bypassing their nearest were captured through the Provider of Services File and American Hospital Association data.
All study activities were approved through the Dartmouth Committee for the Protection of Human Subjects.
Analyses
Descriptive analyses included the following: (1) socio‐demographic and clinical study population characteristics in relation to receipt of surgery (yes/no); (2) one‐way travel time by patient characteristics; (3) closest versus actual travel time by rurality and cancer site; (4) receipt of surgery in relation to travel time, overall and by rurality; (5) bypass in relation to patient characteristics; and (6) bypass in relation to characteristics of the facility used.
We used multivariable logistic regression models to test the relation among patient characteristics, particularly rurality and bypass. The main outcomes of 90‐day surgical mortality and 1‐year mortality were examined in relation to bypass with multivariable logistic regression. Analyses were run in Stata, version 17.
RESULTS
The study population comprised 211,025 Medicare beneficiaries with lung (N = 124,522), colon (N = 53,854), rectal (N = 10,746), or pancreatic (N = 21,903) cancer. The majority of the study population resided in a metropolitan area (74.5%), with 13.0% in micropolitan areas and 12.5% in small towns and rural areas. Overall, the proportion receiving (32.5%; N = 68,537) versus not receiving (67.5%; N = 142,488) surgery was similar across rural–urban categories, but receipt of surgery was lower among older individuals, those with Medicare–Medicaid dual eligibility, and those with cardiovascular comorbidities (Table 1). Receipt of cancer‐directed surgery varied by cancer site and by rural–urban category, with higher proportions of surgery among metropolitan for lung and pancreatic cancers and higher among small town/rural for colon and rectal cancers, when not adjusting for any factors (Table S1).
TABLE 1.
Patient‐associated demographic characteristics for a cancer cohort with and without cancer‐directed surgery among Medicare beneficiaries from 2016 to 2018 (N = 211,025).
| Characteristics, N (%) | All cancers | Lung cancer | Colon cancer | Rectal cancer | Pancreatic cancer | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| No surgery | Surgery | No surgery | Surgery | No surgery | Surgery | No surgery | Surgery | No surgery | Surgery | |
| Demographics | N = 142,488 | N = 68,537 | N = 98,757 | N = 25,765 | N = 18,600 | N = 35,254 | N = 7847 | N = 2899 | N = 17,284 | N = 4619 |
| Patient rurality | ||||||||||
| Metropolitan | 105,860 (74.3%) | 51,322 (74.9%) | 72,292 (73.2%) | 20,014 (77.7%) | 14,279 (76.8%) | 25,544 (72.5%) | 5781 (73.7%) | 2047 (70.6%) | 13,508 (78.2%) | 3717 (80.5%) |
| Micropolitan | 18,706 (13.1%) | 8639 (12.6%) | 13,514 (13.7%) | 3027 (11.7%) | 2205 (11.9%) | 4763 (13.5%) | 1075 (13.7%) | 372 (12.8%) | 1912 (11.1%) | 477 (10.3%) |
| Small town/rural | 17,922 (12.6%) | 8576 (12.5%) | 12,951 (13.1%) | 2724 (10.6%) | 2116 (11.4%) | 4947 (14.0%) | 991 (12.6%) | 480 (16.6%) | 1864 (10.8%) | 425 (9.2%) |
| Age, mean years (SD) | 76.4 (7.2) | 74.6 (6.4) | 76.1 (7.0) | 72.6 (5.0) | 77.5 (7.6) | 76.3 (7.0) | 76.2 (7.5) | 73.9 (6.2) | 76.8 (7.5) | 72.9 (5.5) |
| Age categories (years) | ||||||||||
| 65–70 | 35,483 (24.9%) | 21,672 (31.6%) | 24,945 (25.3%) | 10,021 (38.9%) | 3999 (21.5%) | 8820 (25.0%) | 2227 (28.4%) | 1045 (36.0%) | 4314 (25.0%) | 1800 (39.0%) |
| 71–75 | 34,895 (24.5%) | 19,587 (28.6%) | 25,112 (25.4%) | 8762 (34.0%) | 4168 (22.4%) | 8610 (24.4%) | 1845 (23.5%) | 810 (27.9%) | 3772 (21.8%) | 1412 (30.6%) |
| 76–80 | 30,901 (21.7%) | 14,363 (21.0%) | 21,832 (22.1%) | 4973 (19.3%) | 3959 (21.3%) | 7859 (22.3%) | 1513 (19.3%) | 577 (19.9%) | 3598 (20.8%) | 960 (20.8%) |
| 81–85 | 23,549 (16.5%) | 8216 (12.0%) | 16,029 (16.2%) | 1726 (6.7%) | 3314 (17.8%) | 5834 (16.5%) | 1206 (15.4%) | 319 (11.0%) | 3000 (17.4%) | 344 (7.4%) |
| >85 | 17,660 (12.4%) | 4699 (6.9%) | 10,838 (11.0%) | 283 (1.1%) | 3166 (17.0%) | 4175 (11.8%) | 1056 (13.5%) | 148 (5.1%) | 2600 (15.0%) | 103 (2.2%) |
| Female sex | 72,071 (50.6%) | 37,142 (54.2%) | 49,865 (50.5%) | 14,172 (55.0%) | 9432 (50.7%) | 19,551 (55.4%) | 3799 (48.4%) | 1230 (42.4%) | 8977 (51.9%) | 2211 (47.9%) |
| Race/Ethnicity | ||||||||||
| White, non‐Hispanic | 124,242 (87.2%) | 59,903 (87.4%) | 87,053 (88.1%) | 23,025 (89.4%) | 15,877 (85.3%) | 30,553 (86.6%) | 6812 (86.8%) | 2508 (86.5%) | 14,504 (83.9%) | 3858 (83.5%) |
| Black, non‐Hispanic | 9156 (6.4%) | 3770 (5.5%) | 6171 (6.2%) | 1205 (4.7%) | 1267 (6.8%) | 2180 (6.2%) | 444 (5.7%) | 125 (4.3%) | 1275 (7.4%) | 263 (5.7%) |
| Hispanic | 4116 (2.9%) | 2161 (3.2%) | 2417 (2.4%) | 569 (2.2%) | 735 (4.0%) | 1249 (3.5%) | 287 (3.7%) | 116 (4.0%) | 677 (3.9%) | 227 (4.9%) |
| Other, non‐Hispanic | 4974 (3.5%) | 2703 (3.9%) | 3115 (3.2%) | 966 (3.7%) | 727 (3.9%) | 1316 (3.7%) | 304 (3.9%) | 150 (5.2%) | 828 (4.8%) | 271 (5.9%) |
| Medicare/Medicaid dual eligible | 10,633 (7.5%) | 3500 (5.1%) | 7510 (7.6%) | 1144 (4.4%) | 1295 (7.0%) | 2002 (5.7%) | 508 (6.5%) | 162 (5.6%) | 1320 (7.6%) | 194 (4.2%) |
| Area deprivation index, median (IQR) | 48 (26–70) | 45 (23–67) | 50 (28–72) | 42 (21–64) | 45.0 (23–68) | 48 (26–69) | 47 (24–68) | 48 (26–70) | 42 (21–65) | 37 (18–60) |
| Clinical characteristics | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Diabetes | 32,215 (22.6%) | 15,023 (21.9%) | 21,352 (21.6%) | 5057 (19.6%) | 4388 (23.6%) | 8211 (23.3%) | 1549 (19.7%) | 555 (19.1%) | 4927 (28.5%) | 1211 (26.2%) |
| Myocardial infarction | 1635 (1.1%) | 531 (0.8%) | 1191 (1.2%) | 192 (0.7%) | 186 (1.0%) | 307 (0.9%) | 94 (1.2%) | 13 (0.4%) | 164 (0.9%) | 19 (0.4%) |
| Congestive heart failure | 13,299 (9.3%) | 3932 (5.7%) | 9588 (9.7%) | 1164 (4.5%) | 1657 (8.9%) | 2489 (7.1%) | 576 (7.3%) | 96 (3.3%) | 1478 (8.6%) | 187 (4.0%) |
| Stroke/Transient ischemic attack | 3328 (2.3%) | 1116 (1.6%) | 2313 (2.3%) | 335 (1.3%) | 459 (2.5%) | 677 (1.9%) | 151 (1.9%) | 44 (1.5%) | 405 (2.3%) | 60 (1.3%) |
| Chronic obstructive pulmonary disease | 30,366 (21.3%) | 7489 (10.9%) | 26,341 (26.7%) | 4270 (16.6%) | 1787 (9.6%) | 2761 (7.8%) | 632 (8.1%) | 172 (5.9%) | 1607 (9.3%) | 288 (6.2%) |
| End‐stage renal disease | 1761 (1.2%) | 583 (0.9%) | 1208 (1.2%) | 179 (0.7%) | 261 (1.4%) | 352 (1.0%) | 79 (1.0%) | 19 (0.7%) | 213 (1.2%) | 33 (0.7%) |
| CMS–HCC risk score, median (IQR) | 0.7 (0.4–1.2) | 0.6 (0.4–0.9) | 0.7 (0.4–1.2) | 0.5 (0.3–0.9) | 0.7 (0.4–1.1) | 0.6 (0.4–1.0) | 0.6 (0.4–1.0) | 0.5 (0.3–0.7) | 0.8 (0.5–1.3) | 0.6 (0.4–0.9) |
Abbreviations: CMS, Medicare and Medicaid Services; HCC, Hierarchical Condition Categories.
Travel time
Travel time to the closest surgical facility for each cancer type varied across rural–urban categories. For example, 50% and 67% of rural lung and colon cancer patients, respectively, lived within 1‐h drive time to the nearest surgical facility, as compared to 12% and 26% for patients with pancreatic or rectal cancer (Figure 1A).
FIGURE 1.
(A) Estimated travel time (minutes) from ZIP code of patient residence to closest facility (N = 211,025). Gray shaded bars account for data suppression of <11 counts per Medicare and Medicaid Services (CMS) policy. (B) Estimated travel time (minutes) from ZIP code of patient residence to facility used for surgery, N = 68,537. Gray shaded bars account for data suppression of <11 counts per CMS policy.
Travel time to the facility where surgery was received was much longer than to the closest facility for all four cancer sites and rural–urban categories. For example, although 67% of small town/rural beneficiaries with colon cancer lived within 1 h from the closest facility performing colon cancer surgery, only 27% of these beneficiaries ultimately received surgery at a facility within 1‐h drive time. In fact, although only 6% of rural individuals with colon cancer were two or more hours away from their closest facility, 22% of rural patients receiving colon cancer surgery traveled more than 2 h for the surgery (Figure 1B). Among those beneficiaries receiving lung or pancreatic cancer surgery, median travel times were over twice as long for all non‐metropolitan areas and almost three times longer for the most rural areas (median travel time in minutes: lung; 37 metropolitan, 101 rural; pancreatic; 43 metropolitan, 119 rural).
Bypassing the closest surgical facility
We examined characteristics of patients who bypassed their closest surgical facility to receive cancer surgery at a more distant facility (N = 46,537; 68%). The proportion of cancer patients bypassing varied much more notably by cancer type than by urban–rural status (Figure 2). About 66% of both metropolitan and micropolitan residents bypassed their closest facility, as compared to 78% of small town/rural patients (Table 2). Patients with lung cancer had the lowest bypass of their closest facility (60.2%), followed by pancreatic (73.1%), colon (76.2%), and rectal (77.4%). As age increased, the proportion of patients who bypassed decreased. Bypassing did not vary notably by race or ethnicity, although non‐Hispanic patients had the highest rate of bypassing (Table 2). HCC risk score quartile and ADI quartile showed only modest differences in proportion bypassing. Among metropolitan patients who bypassed their closest facility (N = 34,082), the majority (73.9%) received their surgery at a facility that imposed less than 30 min of additional travel time. This is in contrast to micropolitan patients, of whom the majority (55.8%) went to a facility with >45 min of additional travel time; for small town/rural patients, it was 45.9% (Table 2).
FIGURE 2.
Proportion of Medicare beneficiaries with lung, colon, rectal, or pancreatic cancer without cancer‐directed surgery or with surgery, either bypassing one's closest facility for the patient cancer type or not, overall, and in relation to urban–rural area of residence.
TABLE 2.
Travel time burden measures for Medicare beneficiaries undergoing cancer‐directed surgery from 2016 to 2018 (N = 68,531).
| Characteristics | Bypassed nearest | Additional travel time one way a , min | |||||
|---|---|---|---|---|---|---|---|
| <30 | 30–44 | 45–59 | 60–89 | 90–119 | 120+ | ||
| Demographics | |||||||
| Patient rurality | |||||||
| Metropolitan | 34,082 (66.4%) | 21,277 (73.9%) | 3400 (11.8%) | 1485 (5.2%) | 1218 (4.2%) | 465 (1.6%) | 931 (3.2%) |
| Micropolitan | 5730 (66.3%) | 1358 (29.5%) | 671 (14.6%) | 827 (18.0%) | 866 (18.8%) | 393 (8.5%) | 483 (10.5%) |
| Small town/rural | 6725 (78.4%) | 2304 (40.4%) | 781 (13.7%) | 781 (13.7%) | 926 (16.2%) | 411 (7.2%) | 502 (8.8%) |
| Cancer type | |||||||
| Lung | 21,220 (60.2%) | 12,487 (68.1%) | 2065 (11.3%) | 1364 (7.4%) | 1192 (6.5%) | 482 (2.6%) | 757 (4.1%) |
| Colon | 19,620 (76.2%) | 9736 (59.8%) | 2232 (13.7%) | 1414 (8.7%) | 1433 (8.8%) | 598 (3.7%) | 859 (5.3%) |
| Rectal | 3577 (77.4%) | 1681 (62.6%) | 310 (11.5%) | 166 (6.2%) | 222 (8.3%) | 114 (4.2%) | 191 (7.1%) |
| Pancreatic | 2120 (73.1%) | 1035 (58.3%) | 245 (13.8%) | 149 (8.4%) | 163 (9.2%) | 75 (4.2%) | 109 (6.1%) |
| Age categories (years) | |||||||
| 65–70 | 15,448 (71.3%) | 7905 (61.7%) | 1656 (12.9%) | 1030 (8.0%) | 1041 (8.1%) | 472 (3.7%) | 718 (5.6%) |
| 71–75 | 13,707 (70.0%) | 7226 (63.0%) | 1446 (12.6%) | 954 (8.3%) | 924 (8.1%) | 356 (3.1%) | 566 (4.9%) |
| 76–80 | 9583 (66.7%) | 5197 (64.2%) | 1033 (12.8%) | 643 (7.9%) | 611 (7.6%) | 242 (3.0%) | 364 (4.5%) |
| 81–85 | 5126 (62.4%) | 2962 (67.9%) | 480 (11.0%) | 308 (7.1%) | 290 (6.6%) | 132 (3.0%) | 190 (4.4%) |
| >85 | 2673 (56.9%) | 1649 (70.7%) | 237 (10.2%) | 158 (6.8%) | 144 (6.2%) | 67 (2.9%) | 78 (3.3%) |
| Female sex | 25,002 (67.3%) | 13,528 (64.2%) | 2602 (12.3%) | 1605 (7.6%) | 1613 (7.7%) | 670 (3.2%) | 1060 (5.0%) |
| Race/Ethnicity | |||||||
| White, non‐Hispanic | 40,511 (67.6%) | 21,300 (62.8%) | 4242 (12.5%) | 2784 (8.2%) | 2727 (8.0%) | 1139 (3.4%) | 1703 (5.0%) |
| Black, non‐Hispanic | 2556 (67.8%) | 1592 (73.2%) | 230 (10.6%) | 141 (6.5%) | 120 (5.5%) | 42 (1.9%) | 51 (2.3%) |
| Hispanic | 1477 (68.3%) | 884 (67.9%) | 173 (13.3%) | 64 (4.9%) | 67 (5.1%) | 43 (3.3%) | 71 (5.5%) |
| Other, non‐Hispanic | 1993 (73.7%) | 1163 (68.2%) | 207 (12.1%) | 104 (6.1%) | 96 (5.6%) | 45 (2.6%) | 91 (5.3%) |
| Medicare/Medicaid dual eligible | 2347 (67.1%) | 1341 (65.0%) | 257 (12.5%) | 150 (7.3%) | 149 (7.2%) | 79 (3.8%) | 87 (4.2%) |
| HCC risk score, median (IQR) | 0.6 (0.4–0.9) | 0.6 (0.4–0.9) | 0.6 (0.4–0.9) | 0.5 (0.4–0.9) | 0.6 (0.4–0.9) | 0.5 (0.4–0.9) | 0.5 (0.3–0.8) |
| HCC risk score quartile | |||||||
| Quartile 1: <0.438 | 16,400 (69.9%) | 8543 (62.9%) | 1644 (12.1%) | 1114 (8.2%) | 1067 (7.9%) | 454 (3.3%) | 763 (5.6%) |
| Quartile 2: 0.438–0.678 | 11,641 (67.5%) | 6220 (63.6%) | 1269 (13.0%) | 789 (8.1%) | 733 (7.5%) | 318 (3.2%) | 457 (4.7%) |
| Quartile 3: 0.679–1.079 | 10,348 (66.5%) | 5652 (64.5%) | 1098 (12.5%) | 681 (7.8%) | 666 (7.6%) | 268 (3.1%) | 394 (4.5%) |
| Quartile 4: 1.080+ | 7957 (66.1%) | 4419 (65.0%) | 827 (12.2%) | 505 (7.4%) | 534 (7.9%) | 222 (3.3%) | 288 (4.2%) |
| ADI, median (IQR) | 44.0 (22.0–67.0) | 39.0 (19.0–63.0) | 45.0 (21.0–67.0) | 54.0 (30.0–73.0) | 57.1 (36.0–75.0) | 61.0 (39.0–78.0) | 52.0 (32.0–72.0) |
| ADI quartile | |||||||
| Quartile 1: <25 | 13,433 (71.1%) | 8435 (73.4%) | 1430 (12.4%) | 633 (5.5%) | 478 (4.2%) | 169 (1.5%) | 341 (3.0%) |
| Quartile 2: 25–47 | 11,585 (66.3%) | 6389 (66.5%) | 1118 (11.6%) | 695 (7.2%) | 632 (6.6%) | 253 (2.6%) | 517 (5.4%) |
| Quartile 3: 47–69 | 11,025 (66.1%) | 5296 (58.1%) | 1201 (13.2%) | 860 (9.4%) | 882 (9.7%) | 366 (4.0%) | 508 (5.6%) |
| Quartile 4: 69+ | 10,494 (67.9%) | 4819 (54.3%) | 1103 (12.4%) | 905 (10.2%) | 1018 (11.5%) | 481 (5.4%) | 550 (6.2%) |
Abbreviations: ADI, area deprivation index; HCC, Hierarchical Condition Categories.
aNB: additional travel time calculated only when 6+‐h travel bracket not used (n = 39,079) of total bypass = 46,537 (84%).
In multivariable logistic regression models, we assessed the independent association of patient rurality on likelihood of bypassing the closest facility overall and stratified by cancer type. When adjusting for other factors, increasing rurality was significantly associated with likelihood of bypassing (referent group = metropolitan, OR; 95% CI: micropolitan 1.10; 1.04–1.16, small town/rural 2.08 [1.96–2.20]) (Table 3). Effects of rurality on likelihood of bypassing the closest surgical facility were most notable for lung and rectal cancer patients, for whom small town and rural residence was associated with a >3‐fold increase in likelihood for bypassing for lung cancer surgery (referent group = metropolitan, OR; 95% CI: small town/rural 3.25; 2.87–3.68) and a 2.5‐fold increase for rectal cancer (referent group = metropolitan, OR; 95% CI: small town/rural 2.54; 1.90–3.38) (Table 3).
TABLE 3.
Adjusted odds ratios (ORs) for factors associated with bypass, by cancer type.
| Characteristics | All cancers | Lung cancer | Colon cancer | Rectal cancer | Pancreatic cancer | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p | |
| Demographics | ||||||||||
| Patient rurality | ||||||||||
| Metropolitan | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – |
| Micropolitan | 1.10 (1.04–1.16) | <0.001 | 2.38 (2.14–2.65) | <0.001 | 0.79 (0.74–0.84) | <0.001 | 1.20 (0.92–1.56) | 0.172 | 1.83 (1.39–2.40) | <0.001 |
| Small town/rural | 2.08 (1.96–2.20) | <0.001 | 3.25 (2.87–3.68) | <0.001 | 1.89 (1.76–2.02) | <0.001 | 2.54 (1.90–3.38) | <0.001 | 1.56 (1.18–2.06) | 0.002 |
| Age categories (years) | ||||||||||
| 65–70 | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – |
| 71–75 | 0.94 (0.90–0.98) | 0.003 | 0.97 (0.90–1.04) | 0.380 | 0.95 (0.89–1.01) | 0.080 | 1.04 (0.84–1.29) | 0.730 | 1.02 (0.86–1.22) | 0.799 |
| 76–80 | 0.80 (0.76–0.84) | <0.001 | 0.88 (0.81–0.95) | 0.002 | 0.88 (0.82–0.94) | <0.001 | 1.02 (0.80–1.31) | 0.863 | 0.89 (0.73–1.08) | 0.245 |
| 81–85 | 0.65 (0.61–0.69) | <0.001 | 0.89 (0.78–1.01) | 0.064 | 0.79 (0.73–0.85) | <0.001 | 0.75 (0.55–1.02) | 0.063 | 0.80 (0.60–1.06) | 0.118 |
| >85 | 0.52 (0.48–0.55) | <0.001 | 0.96 (0.72–1.27) | 0.763 | 0.69 (0.63–0.75) | <0.001 | 0.93 (0.61–1.42) | 0.753 | 0.66 (0.42–1.04) | 0.074 |
| Female sex | 0.97 (0.94–1.01) | 0.128 | 0.99 (0.94–1.06) | 0.863 | 0.98 (0.94–1.03) | 0.398 | 0.95 (0.80–1.12) | 0.529 | 0.92 (0.80–1.06) | 0.263 |
| Race/Ethnicity | ||||||||||
| Black, non‐Hispanic | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – |
| White, non‐Hispanic | 1.08 (1.00–1.16) | 0.044 | 1.14 (0.99–1.31) | 0.061 | 1.12 (1.03–1.23) | 0.013 | 1.04 (0.68–1.58) | 0.860 | 1.08 (0.79–1.47) | 0.630 |
| Hispanic | 1.10 (1.00–1.21) | 0.049 | 1.17 (0.96–1.44) | 0.121 | 1.18 (1.05–1.33) | 0.006 | 1.31 (0.83–2.06) | 0.242 | 0.80 (0.59–1.10) | 0.173 |
| Non‐Hispanic, non‐Black non‐White | 1.29 (1.17–1.41) | <0.001 | 1.15 (0.98–1.35) | 0.083 | 1.36 (1.21–1.54) | <0.001 | 1.32 (0.88–1.99) | 0.175 | 1.54 (1.08–2.20) | 0.017 |
| Medicare/Medicaid dual eligible | 0.92 (0.85–0.99) | <0.001 | 0.95 (0.82–1.10) | 0.525 | 0.97 (0.88–1.07) | 0.532 | 0.86 (0.59–1.25) | 0.421 | 0.88 (0.61–1.28) | 0.509 |
| ADI quartile | ||||||||||
| Quartile 1: <25 | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – |
| Quartile 2: 25–47 | 1.05 (1.00–1.10) | 0.060 | 1.02 (0.94–1.10) | 0.688 | 1.05 (0.98–1.12) | 0.162 | 0.81 (0.64–1.03) | 0.081 | 1.00 (0.82–1.23) | 0.963 |
| Quartile 3: 47–69 | 1.05 (1.00–1.11) | 0.064 | 1.00 (0.91–1.09) | 0.926 | 1.03 (0.95–1.10) | 0.483 | 0.89 (0.66–1.18) | 0.417 | 1.14 (0.91–1.43) | 0.243 |
| Quartile 4: 69+ | 1.05 (0.98–1.12) | 0.156 | 0.99 (0.88–1.12) | 0.876 | 1.07 (0.97–1.17) | 0.166 | 0.77 (0.52–1.14) | 0.189 | 0.96 (0.74–1.26) | 0.785 |
| Clinical characteristics | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Diabetes | 0.93 (0.89–0.97) | 0.001 | 0.97 (0.90–1.05) | 0.480 | 0.94 (0.89–1.00) | 0.046 | 1.12 (0.88–1.42) | 0.351 | 0.89 (0.75–1.07) | 0.216 |
| Myocardial infarction | 0.98 (0.81–1.18) | 0.810 | 1.17 (0.82–1.66) | 0.394 | 0.97 (0.76–1.22) | 0.768 | 0.80 (0.23–2.74) | 0.724 | 0.98 (0.32–3.05) | 0.975 |
| CHF | 1.07 (0.99–1.15) | 0.101 | 1.14 (0.98–1.34) | 0.091 | 1.12 (1.01–1.23) | 0.027 | 1.50 (0.87–2.58) | 0.142 | 1.28 (0.86–1.89) | 0.219 |
| Stroke/Transient ischemic attack | 0.90 (0.79–1.02) | 0.111 | 1.02 (0.78–1.32) | 0.894 | 0.88 (0.75–1.03) | 0.112 | 1.08 (0.53–2.19) | 0.838 | 1.54 (0.77–3.09) | 0.225 |
| COPD | 1.03 (0.97–1.09) | 0.355 | 0.90 (0.82–0.98) | 0.022 | 0.96 (0.88–1.05) | 0.383 | 0.75 (0.52–1.09) | 0.133 | 0.75 (0.56–1.00) | 0.050 |
| End‐stage renal disease | 1.01 (0.84–1.20) | 0.949 | 0.87 (0.62–1.22) | 0.424 | 1.07 (0.86–1.33) | 0.566 | 3.59 (0.82–15.81) | 0.090 | 0.75 (0.35–1.61) | 0.461 |
| HCC risk score quartile | ||||||||||
| Quartile 1: <0.438 | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – | 1.00 (ref) | – |
| Quartile 2: 0.438–0.678 | 0.77 (0.73–0.80) | <0.001 | 0.77 (0.71–0.83) | <0.001 | 0.78 (0.74–0.83) | <0.001 | 0.82 (0.65–1.03) | 0.087 | 0.94 (0.78–1.13) | 0.500 |
| Quartile 3: 0.679–1.079 | 0.70 (0.67–0.74) | <0.001 | 0.67 (0.62–0.73) | <0.001 | 0.74 (0.70–0.79) | <0.001 | 0.90 (0.70–1.15) | 0.393 | 1.04 (0.84–1.29) | 0.706 |
| Quartile 4: 1.080+ | 0.72 (0.68–0.76) | <0.001 | 0.69 (0.63–0.76) | <0.001 | 0.79 (0.74–0.85) | <0.001 | 1.04 (0.80–1.35) | 0.767 | 0.73 (0.59–0.91) | 0.005 |
With the high proportion of patients bypassing their closest facility, we sought to characterize the facilities that were used over their closer counterparts. In general, most facilities used by patients who bypassed their closest were in metropolitan areas (94.0%). Small town/rural residents who bypassed predominantly went to metropolitan (79.7%) or micropolitan (16.2%) facilities. For all patients who bypassed their closest facility, most went to an academic medical center (74.1%), and this was true across all rural–urban categories (metropolitan 76.0%, micropolitan 69.7%, small town/rural 68.2%) (Table S2). Between ∼13% (small town/rural) and ∼19% (metropolitan) of patients bypassing their closest facility received surgical treatment at NCI‐designated cancer centers.
Bypassing the closest surgical facility and clinical outcomes
In multivariable logistic regression models, we estimated the association of bypassing the nearest facility with 90‐day postoperative mortality and 1‐year mortality. Bypassing the nearest facility was associated with a decreased likelihood of both 90‐day postoperative mortality (OR = 0.79; 95% CI 0.74–0.85) and 1‐year mortality (OR = 0.81; 95% CI 0.77–0.86). For all cancers, 90‐day postoperative mortality was significantly lower for metropolitan (lung, colon, and pancreas), micropolitan (pancreas), and rural/small town (pancreas) residents who bypassed their closest facility, compared to those who did not (Table 4). The 90‐day postoperative mortality decrements ranged from 19% to 42%: lung/metropolitan OR = 0.81; 95% CI 0.67–0.98, colon/metropolitan OR = 0.75; 95% CI 0.68–0.83, pancreas/metropolitan OR = 0.53; 95% CI 0.43–0.79, and pancreas/micropolitan OR = 0.28; 95% CI 0.12–0.62 (Table 4).
TABLE 4.
Multivariable regression models*https://resdac.org/file‐availability for association between bypassing nearest facility and clinical outcomes among beneficiaries with cancer‐directed surgery, by cancer type and patient rurality.
| Cancer type and rurality | 90‐Day postoperative mortality* | 1‐Year mortality* | ||
|---|---|---|---|---|
| OR (95% CI) | p | OR (95% CI) | p | |
| All cancers | ||||
| Overall | 0.79 (0.74–0.85) | <0.001 | 0.81 (0.77–0.86) | <0.001 |
| Metropolitan | 0.75 (0.69–0.82) | <0.001 | 0.79 (0.74–0.84) | <0.001 |
| Micropolitan | 0.89 (0.73–1.10) | 0.286 | 0.90 (0.78–1.05) | 0.193 |
| Small town/rural | 0.98 (0.78–1.23) | 0.861 | 0.90 (0.76–1.06) | 0.217 |
| Lung cancer | ||||
| Overall | 0.82 (0.75–0.89) | <0.001 | 0.84 (0.79–0.90) | <0.001 |
| Metropolitan | 0.81 (0.67–0.98) | 0.034 | 0.83 (0.74–0.93) | 0.001 |
| Micropolitan | 1.16 (0.63–2.14) | 0.635 | 0.87 (0.62–1.22) | 0.417 |
| Small town/rural | 0.75 (0.41–1.38) | 0.357 | 0.59 (0.42–0.84) | 0.003 |
| Colon cancer | ||||
| Overall | 0.83 (0.70–0.99) | 0.037 | 0.81 (0.73–0.90) | <0.001 |
| Metropolitan | 0.75 (0.68–0.83) | <0.001 | 0.79 (0.73–0.86) | <0.001 |
| Micropolitan | 0.97 (0.77–1.22) | 0.796 | 0.96 (0.80–1.15) | 0.666 |
| Small town/rural | 1.10 (0.85–1.43) | 0.478 | 1.08 (0.88–1.32) | 0.451 |
| Rectal cancer | ||||
| Overall | 0.96 (0.70–1.33) | 0.817 | 0.57 (0.34–0.95) | 0.032 |
| Metropolitan | 0.66 (0.37–1.19) | 0.169 | 0.95 (0.66–1.36) | 0.767 |
| Micropolitan | 0.15 (0.02–1.09) | 0.061 | 0.94 (0.39–2.25) | 0.881 |
| Small town/rural | 0.48 (0.08–2.84) | 0.417 | 1.12 (0.31–4.09) | 0.860 |
| Pancreatic cancer | ||||
| Overall | 0.53 (0.40–0.69) | <0.001 | 0.62 (0.52–0.73) | <0.001 |
| Metropolitan | 0.58 (0.43–0.79) | 0.001 | 0.63 (0.53–0.76) | <0.001 |
| Micropolitan | 0.28 (0.12–0.62) | 0.002 | 0.53 (0.29–0.97) | 0.038 |
| Small town/rural | 0.30 (0.10–0.91) | 0.034 | 0.46 (0.25–0.86) | 0.015 |
*Models adjusted for age, sex, race, dual eligibility, beneficiary census region, HCC risk score, diabetes, myocardial infarction, stroke, and congestive heart failure, and ADI.
Similar patterns were seen for 1‐year mortality, with significantly lower likelihood of mortality among patients of all four cancer types who bypassed their closest surgical facility (OR = 0.81; 95% CI 0.77–0.84) (Table 4). The likelihood of 1‐year mortality was lower by between 17% and 54%, with the greatest decrement seen in pancreatic cancer across all rural–urban categories (OR; 95% CI: metropolitan 0.63; 0.53–0.76; micropolitan 0.53; 0.29–0.97; small town/rural 0.46; 0.25–0.86).
DISCUSSION
This large, population‐based study of Medicare beneficiaries with lung, colorectal, or pancreatic cancer provides novel insights into the well‐documented disparities in mortality among rural cancer patients, particularly related to bypassing their nearest facility for surgical care. Rates of surgery among beneficiaries with lung, colon, rectal, and pancreatic cancers ranged from 17% to 70%, varying mostly by cancer site but similar across all rural–urban categories. Residence in a rural community was associated with significantly longer travel times for all cancer types, and bypass travel times were over three times longer to facilities used than to the closest facility providing that care. More rural patients bypassed their closest facility more often than their non‐rural counterparts, whereas the majority of surgically treated patients bypassed their closest surgical facility for all rural–urban categories and for all cancers. The “destination” facilities used among those who bypassed were notably metropolitan, NCI‐designated cancer centers, and academic medical centers. Bypassing was significantly associated with lower 90‐day postoperative mortality for all cancers and all rural–urban categories. Bypassing was also significantly associated with lower 1‐year mortality—between 17% and 54%, with the greatest decrement among patients with pancreatic cancer.
The majority of geographic access studies have shown that longer travel is associated with better outcomes, 27 , 28 , 29 , 30 suggesting the possible selection bias of those able to travel farther, and/or facility‐level effects of the types of hospitals used as a result of farther travel. For example, in four separate studies using the National Cancer Database, focusing on pancreatic, rectal, colon, esophageal, and liver cancers, patients in the longest quartile of travel distance had improved mortality outcomes, measured as 30‐day mortality, 90‐day mortality, 5‐year mortality, and overall survival time. 27 , 28 , 29 , 30 Interestingly, three of the studies 28 , 29 , 30 examined the joint effect of travel distance and hospital volume and found that patients in the long‐distance, high‐volume hospital group had the most favorable mortality outcomes, even when surgical delays were evident. However, this positive association of travel distance with improved survival is not consistent throughout the literature. Some studies show that longer travel distance is associated with worse outcomes. In a review article of 27 studies, increasing travel distance or time was associated with surgical delay, more inappropriate care, and worse prognosis. 31 Of note, mortality as an outcome in relation to bypass could be confounded by individual's socioeconomic condition, physical capacity, network/support system, as well as characteristics of hospitals attended, such as surgical specialists, volume, and robust care delivery teams.
Our results, showing a lower likelihood of mortality for those who bypassed their nearest surgical facility, may partially obscure an overall higher risk of mortality for rural patients, by “mixing” outcomes of those who bypass and those who do not, particularly given the hospital characteristics for those who bypass. Thus, there could be “healthy bias” among those bypassing who are able to travel (physically and socioeconomically), whereas at the same time, higher quality health systems, which have volumes and better outcomes, such as lower mortality, are accessed by those bypassing. Of note, mortality as an outcome in relation to bypass could be confounded by individual's socioeconomic condition, physical capacity, network/support system, as well as characteristics of hospitals attended, such as surgical specialists, volume, and robust care delivery teams.
Bypassing one's nearest facility that offers the service needed is a measure that cannot distinguish long travel time due to the lack of proximal health care, from patient choice or referral factors. An early study of bypassing for inpatient care among rural patients found a 30% bypass rate, but that study did not examine whether the service needed was offered at the bypass facility. 29 Two subsequent studies examined bypassing a specific cancer service, with one reporting a bypass rate of 19%–26% for colon cancer care 24 and the other a bypass rate of 84% for pancreaticoduodenectomy (the most common form of pancreatectomy). 25 Our bypass rates were much higher for colon cancer (76%) and ranged from 60% for lung cancer to 77% for rectal cancer (pancreas was 73%).
Unlike the majority of evidence that farther travel time/distance is associated with reduced mortality, most evidence examining rurality among cancer patients shows worse mortality compared to less rural counterparts. Many studies demonstrate this relationship, exemplified by studies showing higher mortality rates among rural compared to urban individuals for all cancer sites combined and also specifically for colorectal, oropharyngeal, breast, cervical, and prostate cancers. 32 , 33 Although many factors are likely to contribute to this rural disparity, such as smoking rates, persistent poverty, health care resources, and others, the overall construct of rurality is complex, multifactorial, and difficult to fully understand in actionable ways.
Several limitations should be noted, including those typical of claims data, such as ascertainment of cancer based on billing codes, lack of cancer stage in Medicare claims, more elderly population, and restriction to fee‐for‐service beneficiaries. With regard to lack of cancer stage, although we excluded those with metastatic disease, we recognize that cancer stage can determine the suitability of surgical management. Thus, we interpret the seemingly high proportions of cancer patients without surgery quite cautiously, given the unknown clinical appropriateness. In addition, we used an empirical definition to capture surgical facilities for each cancer, based on the presence of such surgeries in claims over 3 years within a given facility. Moreover, we were unable to ascribe bypassing, or lack thereof, to a specific reason, such as patient preferences, referral networks, or physical, material, or financial ability/inability to do so. Finally, as with most observational studies based on administrative claims data, the threat of bias from unobserved factors must be considered when interpreting associations. Although a strength of this study is its national, population‐based sample through Medicare, studies with more granular data and/or mixed methods approaches may contribute to this evidence by overcoming some sources of potential bias; however, generalizability will need to be considered.
Our findings suggest that rurality is associated with a greater likelihood of bypassing the nearest surgical care facility to attend a farther one, which is likely to be more specialized. Because we defined bypass as the nearest facility providing any surgical care for the specific type of cancer for a given beneficiary, we know that the beneficiary was likely to have had access to surgical care closer but opted to travel farther for their surgery. Indeed, we defined “nearest facility” based on the facility having billed Medicare for at least one surgery for the cancer under consideration over 3 years, so these facilities are potentially very low volume hospitals. However, there are limitations in this definition, such as the possibility that the one surgery had been performed but only as an emergency or by a surgeon no longer at that facility. Moreover, by requiring only one surgery for a given cancer type, we are allowing for very low volume facilities, potentially, which is often correlated with quality. Speculatively, the reasons for bypassing are likely to be related to: socioeconomic resources, referral patterns, care quality or perceived quality, patient preference, and cultural or value‐based decisions. Notably, minoritized race/ethnicity was not a significant factor in bypassing.
Taken together with prior research, this study suggests that many rural cancer patients travel for surgery, either because of necessity or desire to access more specialized, high‐volume care. Further, outcomes seem to be better for patients who do bypass the closest surgical hospital, although causality cannot be inferred based on evidence to date. Health inequities must be considered, however, and barriers to traveling farther create disparities in both access to more specialized care and in outcomes. Disparities in cancer outcomes based on ability to access farther, more specialized facilities, require new strategies for equitable delivery of care. One such approach may be advanced through nonemergency medical transportation coverage in Medicaid, which could help improve transportation options but is being reduced/eliminated in many states.
In conclusion, this large, population‐based study of Medicare beneficiaries with cancer has helped to identify an important element of rurality associated with mortality, specifically bypass of the nearest surgical facility for a farther facility. Given the impact on outcomes for the rural population of cancer patients, understanding the heterogeneity of those who do and do not bypass facilities may point to a critical health inequity.
CONFLICT OF INTEREST STATEMENT
The authors report no conflicts of interest.
Supporting information
Supporting Information
ACKNOWLEDGMENTS
This work was funded by the National Cancer Institute, R01CA248470. We would like to thank Dr. Fahui Wang for use of national travel time calculations based on a ZIP–ZIP matrix developed by his team in ArcGIS.
Onega T, Ramkumar N, Brooks GA, et al. Travel burden and bypassing closest site for surgical cancer treatment for urban and rural oncology patients. J Rural Health. 2025;41:e12890. 10.1111/jrh.12890
REFERENCES
- 1. Onega T, Duell EJ, Shi X, Wang D, Demidenko E, Goodman D. Geographic access to cancer care in the U.S. Cancer. 2008;112(4):909‐918. [DOI] [PubMed] [Google Scholar]
- 2. Onega T, Alford‐Teaster J, Wang F. Population‐based geographic access to parent and satellite National Cancer Institute Cancer Center Facilities. Cancer. 2017;123(17):3305‐3311. [DOI] [PubMed] [Google Scholar]
- 3. Onega T, Duell EJ, Shi X, Demidenko E, Goodman DC. Race versus place of service in mortality among Medicare beneficiaries with cancer. Cancer. 2010;116(11):2698‐2706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Meilleur A, Subramanian SV, Plascak JJ, Fisher JL, Paskett ED, Lamont EB. Rural residence and cancer outcomes in the United States: issues and challenges. Cancer Epidemiol Biomarkers Prev. 2013;22(10):1657‐1667. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Tsuang WM, Arrigain S, Lopez R, Snair M, Budev M, Schold JD. Patient travel distance and post lung transplant survival in the United States: a cohort study. Transplantation. 2020;104(11):2365‐2372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Brooks GA, Tomaino MR, Ramkumar N, et al. Association of rurality, socioeconomic status, and race with pancreatic cancer surgical treatment and survival. J Natl Cancer Inst. 2023;115(10):1171‐1178. doi: 10.1093/jnci/djad102 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Alford‐Teaster J, Lange JM, Hubbard RA, et al. Is the closest facility the one actually used? An assessment of travel time estimation based on mammography facilities. Int J Health Geogr. 2016;15:8. doi: 10.1186/s12942-016-0039-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Diaz A, Burns S, Paredes AZ, Pawlik TM. Accessing surgical care for pancreaticoduodenectomy: patient variation in travel distance and choice to bypass hospitals to reach higher volume centers. J Surg Oncol. 2019;120(8):1318‐1326. [DOI] [PubMed] [Google Scholar]
- 9. Radcliff TA, Brasure M, Moscovice IS, Stensland JT. Understanding rural hospital bypass behavior. J Rural Health. 2003;19(3):252‐259. [DOI] [PubMed] [Google Scholar]
- 10. Weigel PAM, Ullrich F, Ward MM. Rural bypass of critical access hospitals in Iowa: do visiting surgical specialists make a difference? J Rural Health. 2018;34(1):s21‐s29. doi: 10.1111/jrh.12220 [DOI] [PubMed] [Google Scholar]
- 11. Hartwig W, Werner J, Jäger D, Debus J, Büchler MW. Improvement of surgical results for pancreatic cancer. Lancet Oncol. 2013;14(11):e476‐e485. [DOI] [PubMed] [Google Scholar]
- 12. Aquina CT, Probst CP, Becerra AZ, et al. High volume improves outcomes: the argument for centralization of rectal cancer surgery. Surgery. 2016;159(3):736‐748. [DOI] [PubMed] [Google Scholar]
- 13. Morche J, Mathes T, Pieper D. Relationship between surgeon volume and outcomes: a systematic review of systematic reviews. Syst Rev. 2016;5(1):1‐5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. ResDAC . Research identifiable file availability . ResDAC. Accessed October 7, 2024. https://resdac.org/file-availability [Google Scholar]
- 15. Bronson MR, Kapadia NS, Austin AM, et al. Leveraging linkage of cohort studies with administrative claims data to identify individuals with cancer. Med Care. 2018;56(12):e83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Ramkumar N, Colla CH, Wang Q, O'Malley AJ, Wong SL, Brooks GA. Association of rurality, race and ethnicity, and socioeconomic status with the surgical management of colon cancer and postoperative outcomes among medicare beneficiaries. JAMA Network Open. 2022;5(8):e2229247. doi: 10.1001/jamanetworkopen.2022.29247 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. American Hospital Association . American Hospital Association homepage . AHA. Accessed, October 7, 2024. https://www.aha.org/ [Google Scholar]
- 18. United States Census Bureau . ZIP code tabulation areas . United States Census Bureau. Accessed January 2019. https://www.census.gov/geo/maps‐data/data/cbf/cbf_zcta.html [Google Scholar]
- 19. Washington State Department of Health . Guidelines for using rural‐urban classification systems for public health assessment . Accessed, October 7, 2024. https://doh.wa.gov
- 20. United States Census Bureau . U.S. Census regions and divisions . United States Census Bureau. Accessed October 7, 2024. https://www2.census.gov/geo/pdfs/maps‐data/maps/reference/us_regdiv.pdf [Google Scholar]
- 21. Kind AJH, Buckingham W. Making neighborhood disadvantage metrics accessible: the neighborhood Atlas. New Engl J Med. 2018;378:2456‐2458. doi: 10.1056/NEJMp1802313 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Centers for Medicare and Medicaid Services . Program background . CMS. Accessed, October 7, 2024. https://www.cms.gov/data-research/monitoring-programs/improper-payment-measurement-programs/medicare-part-c-ipm/program-background [Google Scholar]
- 23. Wang F, Wang C, Hu Y, Weiss J, Alford‐Teaster J, Onega T. Automated delineation of cancer service areas in northeast region of the United States: a network optimization approach. Spat Spatio‐temporal Epidemiol. 2020;33:100338. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Hu Y, Wang C, Li R, Wang F. Estimating a large drive time matrix between zip codes in the United States: a differential sampling approach. J Transp Geogr. 2020;86:102770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Onega T, Alford‐Teaster J, Leggett C, et al. The interaction of rurality and rare cancers for travel time to cancer care. J Rural Health. 2023;39:426‐433. doi: 10.1111/jrh.12693 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Loehrer AP, Chen L, Wang Q, Colla CH, Wong SL. Rural disparities in lung cancer‐directed surgery: a Medicare cohort study. Ann Surg. 2023;277(3):e657‐e663. doi: 10.1097/SLA.0000000000005091 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Wasif N, Etzioni D, Habermann EB, Mathur A, Chang YH. Contemporary improvements in postoperative mortality after major cancer surgery are associated with weakening of the volume‐outcome association. Ann Surg Oncol. 2019;26:2348‐2356. [DOI] [PubMed] [Google Scholar]
- 28. Xu Z, Aquina CT, Justiniano CF, et al. Centralizing rectal cancer surgery: what is the impact of travel on patients. Dis Colon Rectum. 2020;63(3):319‐325. [DOI] [PubMed] [Google Scholar]
- 29. Jindal M, Zheng C, Quadri HS, et al. Why do long‐distance travelers have improved pancreatectomy outcomes? J Am Coll Surg. 2017;225(2):216‐225. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Lidsky ME, Sun Z, Nussbaum DP, Adam MA, Speicher PJ. Going the extra mile: improved survival for pancreatic cancer patients traveling to high‐volume centers. Ann Surg. 2017;266(2):333‐338. [DOI] [PubMed] [Google Scholar]
- 31. Ambroggi M, Biasini C, Del Giovane C, Fornari F, Cavanna L. Distance as a barrier to cancer diagnosis and treatment: review of the literature. Oncologist. 2015;20(12):1378‐1385. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Bhatia S, Landier W, Paskett ED, et al. Rural–urban disparities in cancer outcomes: opportunities for future research. JNCI: J Nat Cancer Inst. 2022;114(7):940‐952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Moss JL, Wang M, Liang M, Kameni A, Stoltzfus KC, Onega T. County‐level characteristics associated with incidence, late‐stage incidence, and mortality from screenable cancers. Cancer Epidemiol. 2021;75:102033. [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
Supporting Information