Abstract
Purpose:
Oncology outreach is a common strategy for extending cancer care to rural patients. However, a nationwide characterization of the traveling workforce that enables this outreach is lacking, and the extent to which outreach minimizes travel burden for rural patients is unknown.
Methods:
This cross-sectional study analyzed a rural (non-urban) subset of a 100% fee-for-service sample of 355,139 Medicare beneficiaries with incident breast, colorectal, and lung cancer. Surgical, medical, and radiation oncologists were linked to patients using Part B claims, and traveling oncologists were identified by observing hospital service area transition patterns. We defined oncology outreach as the provision of cancer care by a traveling oncologist outside of their primary hospital service area. We used hierarchical Gamma regression models to examine the separate associations between patient receipt of oncology outreach and one-way patient travel times to chemotherapy, radiotherapy, and surgery.
Results:
On average, 9,934 of 39,960 oncologists conducted annual outreach, where 57.8% traveled with low frequency (0–1 outreach visits/month), 21.1% with medium frequency, (1–3 outreach visits/month), and 21.1% with high frequency (>3 outreach visits/month). Oncologists delivered surgery, radiotherapy, and chemotherapy to 51,715, 27,120, and 5,874 rural beneficiaries, respectively, of whom 2.5%, 6.9%, and 3.6% received oncology outreach. Rural patients who received oncology outreach traveled 16% (95% CI: 11–21%) and 11% (95% CI: 9–13%) less minutes to chemotherapy and radiotherapy than those who did not receive oncology outreach, corresponding to expected one-way savings of 15.9 (95% CI: 15.5–16.4) and 11.9 (95% CI: 11.7–12.2) minutes, respectively.
Conclusion:
Our study introduces a novel claims-based approach for tracking the nationwide traveling oncology workforce and supports oncology outreach as an effective means for improving rural access to cancer care.
INTRODUCTION
Rural-urban cancer disparities are prevalent in the U.S. population. They are largely a function of differential access to care, as rural patients have significantly reduced proximity to cancer specialists than urban patients.1–5 Like many medical specialists, surgical, radiation, and medical oncologists (herein referred to as “oncologists”) tend to cluster in metropolitan areas with high population density.6–8 With only 7.4% of oncologists practicing in Rural America, which is home to 20% of the population, rural patients often experience disproportionate travel times to cancer care.9–11 This increased travel burden can lead to delayed diagnosis, untimely or inappropriate treatment assignment, worse disease outcomes, and overall lower quality of life.12–14
Given the poorer outcomes associated with geographic barriers to cancer care, a key priority for policymakers and providers is minimizing rural patient travel burden to cancer care.15 One strategy for doing so involves establishing oncology outreach arrangements, where oncologists commute to secondary rural clinics to extend specialized care on a recurring basis.16,17 While oncology outreach is a seemingly prevalent and growing strategy,18,19 a nationwide characterization of the traveling oncology workforce is currently lacking. And despite prior work that has measured potential access to the nearest outreach/satellite location for rural populations,20,21 no research has quantified the realized access for patients who received care from traveling oncologists at such locations.
In this study, we aimed to develop a claims-based approach for identifying traveling oncologists, characterize the attributes of the traveling oncology workforce, and investigate the influence of oncology outreach on rural patient travel time to chemotherapy, radiotherapy, and surgery. We limited the scope of this study to rural patients, as oncology outreach has historically focused on addressing the geographic barriers affecting rural populations.
METHODS
Data and approvals
The data used in this cross-sectional study are not publicly available due to a data use agreement with the Centers for Medicare and Medicaid Services (CMS). Dartmouth College’s institutional review board approved all study protocols.
Study cohort
We identified patients with a biopsy for breast, colorectal, or lung cancer followed by 2 cancer diagnosis codes in the following 12 months from a 100% sample of 2015–20 Medicare claims.22 We excluded patients with a cancer diagnosis code in the 12 months prior to biopsy to enrich for incident cases. Further, we excluded patients who were younger than 66 years or older than 99 years at biopsy, were not continuously enrolled in Medicare Parts A and B in the 12 months prior and following biopsy, received multiple cancer diagnoses, or had a missing or non-U.S. ZIP code. Breast, colorectal, and lung cancers were selected due to their prevalence in the population.
Traveling oncology workforce
We linked the study cohort to oncologists who provided care within 12 months following and 3 months preceding biopsy using Part B claims. We sorted each physician’s claims by encounter date and used the Dartmouth Atlas to join hospital service areas (HSAs), or local health care markets for community-based care, to billing ZIP codes.23 To identify traveling oncologists who extend cancer care across communities, we tallied the number of times a physician billed to one HSA followed by another HSA in each year. These HSA “transitions” were used to assign physicians to the following annual travel frequency bins: none (0 transitions), low (1–24 transitions), medium (25–72 transitions), and high (>72 transitions). Assuming a pair of transitions approximates a back-and-forth physician outreach visit and that transitions are uniformly distributed across months, the bins correspond to an average of 0, 0–1, 1–3, and >3 outreach visits/month, respectively. This categorization approach aligns with prior research that examined similar oncologists across several rural states, who typically averaged 1–3 outreach visits/month.20,24–26
Study variables
Patient-level variables
Patient age, sex, and race were identified in the Master Beneficiary Summary file. Race was dichotomized into White vs. other to adhere to CMS data suppression rules. Cancer type was determined using ICD-10 and CPT codes (Appendix Table A1). Charlson comorbidities were assigned by the method of Klabunde using claims over the 12 months preceding biopsy and discretized into the following bins: 0, 1, and 2+ comorbidities.27 Patients were labeled as having metastatic disease if they had a diagnosis of a secondary malignant site within 90 days post-biopsy.28 National Cancer Institute (NCI) affiliation was assigned if a patient had ≥1 encounter at an NCI cancer center within 12 months post-biopsy. Annual residential ZIP code was used to determine a 4-tiered rural categorization (urban, large rural, small rural, and isolated rural) according to the University of Washington’s Categorization of secondary Rural-Urban Commuting Area codes29 and whether the patient lived in a disadvantaged area, defined as a ZIP code with >20% of individuals below the poverty line.30
Physician-level variables
Physician gender and specialty were identified using information from the National Plan and Provider Enumeration System (Appendix Table A2).31 Annual patient volume was estimated by tallying the number of study cohort patients that a physician delivered care to over a year and categorized into low, medium, and high tertiles.
Encounter-level variables
Encounter year and location were identified using the respective date and billing ZIP code in each claim. Chemotherapy, radiotherapy, and surgery procedures were determined for each claim by requiring a procedure code (Appendix Table A3) and the presence of an oncologist with the expected specialty (e.g., medical oncologist for chemotherapy).
Exposure
The primary exposure was patient receipt of oncology outreach, where oncology outreach was defined as the provision of cancer care by a traveling oncologist at a location outside of their primary HSA. Primary HSA assignments were determined annually based on the plurality of encounters.
Outcome
The primary outcome was one-way patient travel time to cancer care, defined as the shortest path in minutes via road networks from the population-weighted centroid of the beneficiary’s residential ZIP code to the population-weighted centroid of the billing ZIP code in each claim. Travel times were extracted from a nationwide ZIP-to-ZIP drive time matrix that was estimated using ArcGIS Network Analyst and refined by Google Maps.32
Procedure types
We identified the first chemotherapy, radiotherapy, and surgery procedures for each patient and apportioned the observations into separate procedure cohorts. The resulting cohorts contained one observation per patient and allowed for patients to span multiple cohorts. We restricted each cohort to patients who lived in rural (non-urban) areas, traveled ≤240 minutes to care, and received their first treatment before the onset of the COVID-19 pandemic. We chose a ≤240 minute restriction, as prior work showed that patient travel times >240 minutes represent extreme/outlier cases.10 We also excluded encounters during COVID-19 to account for the irregularities of care delivery during this period, considering March 15, 2020 as the U.S. pandemic start date.33
Statistical analysis
We characterized the traveling oncology workforce by summarizing physician-level attributes by travel frequency. We assessed bivariate associations between patient-level characteristics and receipt of oncology outreach by procedure cohort. We used hierarchical Gamma regression models with a log-link to investigate the primary associations between patient receipt of oncology outreach and right-skewed travel time estimates for each procedure cohort. In these models, we adjusted for all patient-, physician-, and encounter-level study variables to control for potential confounding and included random intercepts for physician identifier and beneficiary HSA to account for within-cluster correlation. These models were used to compute expected marginal means for outreach and non-outreach encounters, and differences between expected marginal means were calculated to quantify one-way savings attributable to oncology outreach. Further, we performed sensitivity analysis to test the robustness of primary associations when flexing the travel time restriction to 210, 180, and 150 minutes.21,34 In supplementary analysis, we explored if physician travel frequency, patient cancer type, or patient rurality modified the primary associations, accounting for multiple testing with a Bonferroni adjustment. For each statistical model, we reported point estimates, 95% confidence intervals, and P-values.
RESULTS
We linked 39,960 oncologists to the 355,159 patients in the study cohort and identified 9,935 (24.9%) oncologists who engaged in annual outreach. Of these traveling oncologists, 57.8% traveled with low frequency, 21.1% with medium frequency, and 21.1% with high frequency. The proportion of male and high-volume oncologists increased as a function of travel frequency, while the proportion of female and medium/low-volume oncologists decreased. In addition, the proportion of oncologists who delivered most of their care in rural HSAs rose with increasing travel frequency. Oncologists who traveled with high frequency spanned a greater number of HSAs relative to those who traveled with low frequency. Further, the distribution of cancer specialists varied substantially by travel frequency (Table 1). Specifically, of the 22,731 surgeons, 79.4% did not travel while 14.1% traveled with low frequency, 4.0% with medium frequency, and 2.5% with high frequency; of the 12,454 medical oncologists, 70.3% did not travel while 14.6% traveled with low frequency, 7.2% with medium frequency, and 7.9% with high frequency; and of the 4,775 radiation oncologists, 67.3% did not travel while 15.0% traveled with low frequency, 6.3% with medium frequency, and 11.5% with high frequency (Figure 1).
Table 1.
Physician attributes stratified by travel frequency
| Travel Frequencya | ||||
|---|---|---|---|---|
| None | Low | Medium | High | |
| N=30025 | N=5742 | N=2097 | N=2096 | |
| Gender | ||||
| Men | 22013 (73.3%) | 4173 (72.7%) | 1575 (75.1%) | 1655 (79.0%) |
| Women | 8012 (26.7%) | 1569 (27.3%) | 522 (24.9%) | 441 (21.0%) |
| Patient Volume | ||||
| Low | 13318 (44.4%) | 2228 (38.8%) | 588 (28.0%) | 382 (18.2%) |
| Medium | 8058 (26.8%) | 1664 (29.0%) | 544 (26.0%) | 435 (20.8%) |
| High | 8649 (28.8%) | 1850 (32.2%) | 964 (46.0%) | 1279 (61.0%) |
| Primary HSA Setting | ||||
| Ruralb | 2738 (9.1%) | 827 (14.4%) | 309 (14.7%) | 342 (16.3%) |
| Urban | 27288 (90.9%) | 4915 (85.6%) | 1787 (85.3%) | 1754 (83.7%) |
| Specialty | ||||
| Surgery | 18053 (60.1%) | 3207 (55.8%) | 904 (43.1%) | 567 (27.0%) |
| Medical Oncology | 8760 (29.2%) | 1821 (31.7%) | 892 (42.6%) | 981 (46.8%) |
| Radiation Oncology | 3212 (10.7%) | 714 (12.4%) | 301 (14.3%) | 548 (26.2%) |
| Unique HSAs (SD) | 1.0 (0.0) | 2.2 (0.02) | 2.4 (0.04) | 2.6 (0.04) |
Physician attributes were averaged across annual travel frequency assignments, using the number of physicians in each annual travel frequency bin as a weight.
Large rural, small rural, and isolated rural areas grouped together.
Note: Column-wise proportions may not sum to 1 due to rounding.
Figure 1.
Distribution of travel frequency by oncology specialty
Of the original study cohort, 36,118, 147,751, and 263,646 patients received chemotherapy, radiotherapy, and surgery, respectively. Upon restriction to rural patients, those who traveled ≤240 minutes to care, and those who received treatment before COVID-19, 5,874, 27,120, and 51,715 remained and were respectively included in the final chemotherapy, radiotherapy, and surgery cohorts (Figure 2). In total, 61,704 unique patients spanned these cohorts, and 3,299 (5.3%) received at least one treatment at an outreach visit. Of the 21,320 (34.6%) patients who received two or more treatment modalities, 1,925 (9.0%) received one modality at an outreach visit, while 77 (<1%) received two or more modalities at an outreach visit (Appendix Table A4).
Figure 2.
Study cohort flow diagram
In the chemotherapy cohort, 213 (3.6%) patients received oncology outreach while 5,661 (96.4%) did not; the distribution of patient characteristics in the chemotherapy cohort was balanced among those who received oncology outreach versus those who did not. In the radiotherapy cohort, 1,861 (6.9%) patients received oncology outreach while 25,259 (93.1%) did not; recipients of oncology outreach were more likely to be female (vs. male, P<0.01), live in disadvantaged areas (vs. advantaged, P<0.01), have breast cancer (vs. colorectal/lung, P<0.01), and have non-metastatic disease (vs. metastatic, P<0.01) compared to those who did not receive oncology outreach. In the surgery cohort, 1,302 (2.5%) patients received oncology outreach while 50,413 (97.5%) did not; recipients of oncology outreach were more likely to have lung cancer (vs. breast/colorectal, P=0.01) and greater comorbidity (P<0.01) compared to those who did not receive oncology outreach (Table 2).
Table 2.
Patient characteristics among those who received oncology outreach versus those who did not, stratified by procedure cohort
| Chemotherapy | Radiotherapy | Surgery | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Outreach | Non-outreach | P | Outreach | Non-outreach | P | Outreach | Non-outreach | P | |
| (N=213) | (N=5661) | (N=1861) | (N=25259) | (N=1302) | (N=50413) | ||||
| Age at Diagnosis | |||||||||
| 66–70 | 83 (39.0%) | 2233 (39.4%) | 0.95 | 688 (37.0%) | 9483 (37.5%) | 0.49 | 441 (33.9%) | 16178 (32.1%) | 0.40 |
| 71–75 | 67 (31.5%) | 1831 (32.3%) | 556 (29.9%) | 7725 (30.6%) | 356 (27.3%) | 14764 (29.3%) | |||
| 76–80 | 40 (18.8%) | 1052 (18.6%) | 388 (20.8%) | 4896 (19.4%) | 275 (21.1%) | 10535 (20.9%) | |||
| >80 | 23 (10.8%) | 545 (9.6%) | 229 (12.3%) | 3155 (12.5%) | 230 (17.7%) | 8936 (17.7%) | |||
| Cancer Type | |||||||||
| Breast | 103 (48.4%) | 2439 (43.1%) | 0.26 | 1340 (72.0%) | 16540 (65.5%) | <0.01 | 798 (61.3%) | 32169 (63.8%) | 0.01 |
| Colorectal | 46 (21.6%) | 1444 (25.5%) | 157 (8.4%) | 2179 (8.6%) | 341 (26.2%) | 13204 (26.2%) | |||
| Lung | 64 (30.0%) | 1778 (31.4%) | 364 (19.6%) | 6540 (25.9%) | 163 (12.5%) | 5040 (10.0%) | |||
| Comorbidities | |||||||||
| 0 | 104 (48.8%) | 2591 (45.8%) | 0.17 | 910 (48.9%) | 12188 (48.3%) | 0.76 | 598 (45.9%) | 24770 (49.1%) | <0.01 |
| 1 | 63 (29.6%) | 1520 (26.9%) | 462 (24.8%) | 6236 (24.7%) | 306 (23.5%) | 12212 (24.2%) | |||
| 2+ | 46 (21.6%) | 1550 (27.4%) | 489 (26.3%) | 6835 (27.1%) | 398 (30.6%) | 13431 (26.6%) | |||
| Living in Disadvantaged Area | |||||||||
| No | 166 (77.9%) | 4333 (76.2%) | 0.61 | 1352 (72.6%) | 20059 (79.4%) | <0.01 | 1037 (79.6%) | 39383 (78.1%) | 0.20 |
| Yes | 47 (22.1%) | 1348 (23.8%) | 509 (27.4%) | 5200 (20.6%) | 265 (20.4%) | 11030 (21.9%) | |||
| Metastatic Disease | |||||||||
| No | 172 (80.8%) | 4437 (78.4%) | 0.46 | 1722 (92.5%) | 22824 (90.4%) | <0.01 | 1210 (92.9%) | 47132 (93.5%) | 0.45 |
| Yes | 41 (19.2%) | 1224 (21.6%) | 139 (7.5%) | 2435 (9.6%) | 92 (7.1%) | 3281 (6.5%) | |||
| NCI Affiliation | |||||||||
| No | 186 (87.3%) | 4675 (82.6%) | 0.08 | 1508 (81.0%) | 20559 (81.4%) | 0.72 | 1077 (82.7%) | 41859 (83.0%) | 0.79 |
| Yes | 27 (12.7%) | 986 (17.4%) | 353 (19.0%) | 4700 (18.6%) | 225 (17.3%) | 8554 (17.0%) | |||
| Race | |||||||||
| White | 200 (93.9%) | 5294 (93.5%) | 0.94 | 1733 (93.1%) | 23800 (94.2%) | 0.06 | 1239 (95.2%) | 47338 (93.9%) | 0.07 |
| Other | 13 (6.1%) | 367 (6.5%) | 128 (6.9%) | 1459 (5.8%) | 63 (4.8%) | 3075 (6.1%) | |||
| Rurality | |||||||||
| Isolated | 46 (21.6%) | 1071 (18.9%) | 0.33 | 338 (18.2%) | 5098 (20.2%) | 0.11 | 247 (19.0%) | 10143 (20.1%) | 0.09 |
| Small Rural | 49 (23.0%) | 1539 (27.2%) | 513 (27.6%) | 6858 (27.2%) | 333 (25.6%) | 13856 (27.5%) | |||
| Large Rural | 118 (55.4%) | 3051 (53.9%) | 1010 (54.3%) | 13303 (52.7%) | 722 (55.5%) | 26414 (52.4%) | |||
| Sex | |||||||||
| Male | 57 (26.8%) | 1686 (29.8%) | 0.38 | 267 (14.3%) | 4437 (17.6%) | <0.01 | 226 (17.4%) | 8713 (17.3%) | 0.97 |
| Female | 156 (73.2%) | 3975 (70.2%) | 1594 (85.7%) | 20822 (82.4%) | 1076 (82.6%) | 41700 (82.7%) | |||
Note: Column-wise proportions may not sum to 1 due to rounding.
Rural patients had median one-way travel times of 83.3 minutes (IQR: 62.9–116.0) to chemotherapy, 76.4 minutes (IQR: 57.0–105.0) to radiotherapy, and 75.8 minutes (IQR: 54.2–105.5) to surgery. In adjusted models, rural patients who received oncology outreach traveled 16% (95% CI: 11–21%) less minutes to chemotherapy and 11% (95% CI: 9–13%) less minutes to radiotherapy than those who did not receive oncology outreach. These estimates corresponded to expected one-way savings of 15.9 (95% CI: 15.5–16.4) minutes for chemotherapy and 11.9 (95% CI: 11.7–12.2) minutes for radiotherapy (Table 3; Appendix Table A5). Primary associations were robust to changes in travel time restrictions (Appendix Figure A1).
Table 3.
Adjusted main effects, expected marginal means, and one-way savings estimates for oncology outreach, stratified by procedure cohorta
| Chemotherapy | Radiotherapy | Surgery | ||||
|---|---|---|---|---|---|---|
| Est (95% CI) | P | Est (95% CI) | P | Est (95% CI) | P | |
| Main effects b | ||||||
| Outreach: Yes vs. No | 0.84 (0.79, 0.89) | <0.01 | 0.89 (0.87, 0.91) | <0.01 | 1.00 (0.98, 1.03) | 0.93 |
| Expected marginal means c | ||||||
| Outreach: No | 100.6 (90.5, 111.8) | NA | 106.2 (100.2, 112.6) | NA | 91.4 (88.7, 94.2) | NA |
| Outreach: Yes | 84.6 (75.1, 95.4) | NA | 94.3 (88.7, 100.3) | NA | 91.5 (88.1, 95.0) | NA |
| One-way savings c,d,e | ||||||
| Difference (A-B) | 15.9 (15.5, 16.4) | <0.01 | 11.9 (11.7, 12.2) | <0.01 | -0.04 (−0.18, 0.10) | 0.47 |
Adjusted for all encounter-, patient-, and physician-level variables outlined in the Methods section.
Exponentiated point estimates and confidence intervals are presented.
Point estimates and confidence intervals are presented on the outcome variable’s original scale.
One-way savings estimates may not appear to sum due to rounding.
Confidence intervals and P-values were calculated via a parametric bootstrap with 10,000 iterations. In each iteration, we generated Normal distributions for each expected marginal mean group (using the estimated coefficient and standard error), conducted Welch’s two-sample t-test to compare distributions, and extracted the confidence interval and P-value. After completion of all iterations, we averaged across the simulated confidence intervals (lower and upper bounds) and P-values to derive the bootstrapped estimates.
In supplementary analysis, the inverse association between patient receipt of oncology outreach and travel time to chemotherapy was positively modified by breast cancer patients (Est [95% CI]: 0.83 [0.76–0.90]) and those living in large rural areas (0.81 [0.71–0.93]). Additionally, the inverse association between patient receipt of oncology outreach and travel time to radiotherapy was positively modified by high-travel physicians (0.89 [0.83–0.95]), breast cancer patients (0.89 [0.86–0.91]), and patients living in isolated rural areas (0.87 [0.83–0.91]) (Table 4).
Table 4.
Tests for effect modification of the primary associations, stratified by procedure cohorta
| Chemotherapy | Radiotherapy | Surgery | ||||
|---|---|---|---|---|---|---|
| Est (95% CI)b | P | Est (95% CI)b | P | Est (95% CI)b | P | |
| Outreach × Physician Travel Frequency c | ||||||
| Outreach : Low (ref) | 0.93 (0.75, 1.15) | 0.49 | 0.99 (0.92, 1.05) | 0.64 | 0.99 (0.95, 1.04) | 0.77 |
| Outreach : Med | 0.84 (0.63, 1.11) | 0.21 | 0.90 (0.82, 0.99) | 0.03 | 0.98 (0.92, 1.06) | 0.65 |
| Outreach : High | 0.90 (0.72, 1.13) | 0.36 | 0.89 (0.83, 0.95) | 0.001d | 1.01 (0.94, 1.09) | 0.81 |
| Outreach × Patient Cancer Type | ||||||
| Outreach : Breast (ref) | 0.83 (0.76, 0.90) | <0.001d | 0.89 (0.86, 0.91) | <0.001d | 1.01 (0.98, 1.04) | 0.65 |
| Outreach : Colorectal | 1.08 (0.95, 1.23) | 0.22 | 1.05 (0.98, 1.11) | 0.19 | 0.95 (0.90, 1.00) | 0.07 |
| Outreach : Lung | 1.00 (0.90, 1.12) | 0.99 | 1.02 (0.97, 1.06) | 0.48 | 1.00 (0.93, 1.09) | 0.92 |
| Outreach × Patient Rurality | ||||||
| Outreach : Isolated (ref) | 0.95 (0.84, 1.07) | 0.40 | 0.87 (0.83, 0.91) | <0.001d | 0.98 (0.93, 1.04) | 0.58 |
| Outreach : Small Rural | 0.99 (0.84, 1.16) | 0.91 | 1.08 (1.02, 1.14) | 0.007 | 0.95 (0.88, 1.02) | 0.13 |
| Outreach : Large Rural | 0.81 (0.71, 0.93) | 0.003d | 1.01 (0.95, 1.06) | 0.86 | 1.04 (0.97, 1.11) | 0.28 |
Separate models were fit for each set of interactions, adjusting for all encounter-, patient-, and physician-level variables outlined in the Methods section.
Exponentiated point estimates and confidence intervals are presented.
Procedure cohorts were restricted to patients who encountered traveling oncologists, as non-traveling oncologists do not conduct outreach.
Statistically significant after Bonferroni correction (9 tests per cohort: P=0.05/9).
DISCUSSION
In this study, we introduced a claims-based approach for identifying traveling oncologists, characterized the attributes of this unique workforce, and evaluated the association between their outreach and rural patient travel time to cancer care. We found that patient receipt of oncology outreach was associated with one-way savings of 15.9 and 11.9 minutes to chemotherapy and radiotherapy, respectively. To our knowledge, this is the first study to leverage nationwide claims to profile the traveling oncology workforce and quantify their influence on rural access to cancer care.
The identified characteristics of the traveling oncology workforce are consistent with prior research. We observed that higher travel physicians were more likely to be men. Similarly, a previous study found that men comprised a greater proportion of visiting specialists compared to non-visiting specialists,25 and another reported that men were more likely to deliver care at a greater number of practice sites than women.35 Additionally, we observed that higher travel physicians had greater patient volumes. The decisions to travel and provide care to more patients may relate to clinical availability or reflect a physician financial incentive, which research has previously documented.25,36 Further, we found that higher travel physicians were comprised of a greater mix of medical and radiation oncologists, whereas the opposite was true for surgeons. Prior work showed that radiation oncologists maintain the largest geographic dispersion of practices, followed by medical and surgical oncologists,37 which may partially explain why medical and radiation oncologists comprised an increasing density of higher travel physicians; in turn, this shift in specialty mix at higher travel frequencies likely explains why the average number of HSAs per physician also increased with travel frequency. Lastly, we showed that the proportion of oncologists who delivered most of their care in rural HSAs rose with travel frequency. This indicates that oncologists engaged in greater outreach spent more time providing care to rural communities, which aligns with our expectations.
In contrast, our primary results somewhat diverge from previous research that reported the one-way median travel time to the nearest medical oncology clinic fell from 51.6 to 19.2 minutes for rural Iowans when considering outreach clinics.20 This effect is noticeably larger than our expected 15.9-minute savings, and differences in effect sizes are likely due to a few sources of variation. Our study considers the influence of oncology outreach across the U.S., whereas the former focusses on oncology outreach in Iowa. Because Iowa is a pioneer of oncology outreach,19 the geographic presence of Iowan outreach is likely more extensive than in other states, which may partially account for effect size differences. Additionally, our study compared travel times to chemotherapy among patients who received oncology outreach versus those who did not, while the former compared travel times to the closest medical oncology clinic across two scenarios: when considering outreach clinics versus not. Despite the validity of this approach, their comparison assumes patients visit their closest clinic and does not recognize patient preferences, the availability/supply of physicians, or services provided at these facilities; therefore, their comparison quantifies improvements in potential access rather than realized access, which may contribute to higher savings estimates.
Another important finding was the expected 11.9-minute savings to radiotherapy for outreach-receiving patients, as research is yet to explore the influence of oncology outreach on rural access to radiotherapy. Beyond this first-order association, of particular interest are the interaction effects identified in supplementary analysis. Substantial research has highlighted that rural breast cancer patients have significantly reduced uptake of lumpectomy and adjuvant radiotherapy compared to urban patients, mostly due to high travel burden to radiation oncologists.34,38–46 It is therefore promising that the inverse association between oncology outreach and patient travel time to radiotherapy was pronounced among breast cancer patients, those living in isolated rural areas, and those receiving care from high-travel radiation oncologists. These insights suggest that high-frequency radiation oncology outreach is helpful in extending access to radiotherapy for rural breast cancer patients.
While we detected significant inverse associations between oncology outreach and patient travel times to chemotherapy and radiotherapy, we recognize that our results may be understated. It is likely that patients who did not receive oncology outreach resided closer to their preferred clinic, resulting in shorter travel times to care relative to those who received oncology outreach. Despite our efforts to control for geographic variation during modeling, the presence of unaccounted geographic heterogeneity between exposure groups may have diluted the observed effect sizes. Additionally, our analysis did not quantify the total savings attributable to oncology outreach due to high variation in the number and type of subsequent treatment visits among patients in our sample. However, ≈80% of patients who received their first chemotherapy and radiotherapy at a physician outreach visit received all subsequent treatment from the same physician at the same location; therefore, the total savings attributable to oncology outreach is likely substantial.
Our study has several limitations. First, the approach for classifying traveling oncologists is based on Medicare claims. As a result, oncologists who encountered large numbers of Medicare beneficiaries had increased likelihood of being classified as high-travel physicians. Second, our definition of oncology outreach relies on HSA boundaries and does not consider physicians who resided on the border of contiguous HSAs or those who moved mid-year. Third, patient travel time estimates assume that individuals commuted from their residential location to treatment by car, which does not consider those who used public transportation, received cancer care while away from home, or moved mid-year. Fourth, this study is prone to residual confounding due to the absence of certain patient and physician variables that may distort the exposure-outcome relationship. Fifth, our results may not extrapolate to younger patients or those with less common cancers, as the sample was confined to Medicare patients with breast, colorectal, and lung cancer. Sixth, the influence of oncology outreach on other measures of access, quality, and cost was not examined as it is beyond the scope of this study. Finally, the cross-sectional assessment of the exposure and outcome precludes any causal inference.
CONCLUSION
In conclusion, we used nationwide Medicare claims to identify traveling oncologists, characterize their attributes, and understand how their outreach influences patient travel burden to cancer care. We found that receipt of oncology outreach was associated with significant reductions in travel time to chemotherapy and radiotherapy for rural patients. Future research should consider exploring these associations in all-payer claims data, conducting a prospective study to assess causality, or modeling additional outcomes. Overall, our study introduces a novel claims-based approach for tracking the traveling oncology workforce and supports oncology outreach as an effective means for extending rural access to cancer care.
Supplementary Material
CONTEXT SUMMARY.
Key objective:
To characterize the nationwide traveling oncology workforce and quantify the travel time savings for rural Medicare patients who received oncology outreach.
Knowledge generated:
Approximately a quarter of identified surgical, medical, and radiation oncologists conducted annual oncology outreach. Patients who received treatment at a physician outreach visit had an expected one-way savings of 15.9 minutes for chemotherapy and 11.8 minutes for radiotherapy compared to those who did not receive treatment at a physician outreach visit.
Relevance:
Traveling oncologists play a critical role in rural cancer care delivery, and our results support oncology outreach as an effective strategy for extending rural access to cancer care.
Funding:
National Institutes of Health (R37CA263936)
REFERENCES
- 1.Rural-Urban Disparities in Cancer. Accessed October 1, 2022. https://gis.cancer.gov/mapstory/rural-urban/index.html
- 2.Zahnd WE, James AS, Jenkins WD, et al. Rural-Urban Differences in Cancer Incidence and Trends in the United States. Cancer Epidemiol Biomark Prev Publ Am Assoc Cancer Res Cosponsored Am Soc Prev Oncol. 2018;27(11):1265–1274. doi: 10.1158/1055-9965.EPI-17-0430 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Hussaini S, Blackford AL, Gupta A. Rural-urban disparities in cancer mortality in the United States from 1999 to 2019. J Clin Oncol. 2022;40(16_suppl):6553–6553. doi: 10.1200/JCO.2022.40.16_suppl.6553 [DOI] [Google Scholar]
- 4.Yabroff KR, Han X, Zhao J, Nogueira L, Jemal A. Rural Cancer Disparities in the United States: A Multilevel Framework to Improve Access to Care and Patient Outcomes. JCO Oncol Pract. 2020;16(7):409–413. doi: 10.1200/OP.20.00352 [DOI] [PubMed] [Google Scholar]
- 5.American Society of Clinical Oncology. The state of cancer care in America, 2014: a report by the American Society of Clinical Oncology. J Oncol Pract. 2014;10(2):119–142. doi: 10.1200/JOP.2014.001386 [DOI] [PubMed] [Google Scholar]
- 6.2014 Update of the Rural-Urban Chartbook | NORC.org. norc. Accessed January 18, 2023. https://www.norc.org/Research/Projects/Pages/2014-update-of-the-rural-urban-chartbook.aspx
- 7.Cyr ME, Etchin AG, Guthrie BJ, Benneyan JC. Access to specialty healthcare in urban versus rural US populations: a systematic literature review. BMC Health Serv Res. 2019;19(1):974. doi: 10.1186/s12913-019-4815-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Fordyce MA, Chen FM, Doescher MP, Hart LG. 2005 physician supply and distribution in rural areas of the United States. WWAMI Rural Health Research Center, University of Washington. Accessed January 9, 2023. https://familymedicine.uw.edu/rhrc/publications/2005-physician-supply-and-distribution-in-rural-areas-of-the-united-states/ [Google Scholar]
- 9.Ward MM, Ullrich F, Matthews K, et al. Where Do Patients With Cancer in Iowa Receive Radiation Therapy? J Oncol Pract. 2014;10(1):20–25. doi: 10.1200/JOP.2013.001191 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.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(2):426–433. doi: 10.1111/jrh.12693 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Chan L, Hart LG, Goodman DC. Geographic Access to Health Care for Rural Medicare Beneficiaries. J Rural Health. 2006;22(2):140–146. doi: 10.1111/j.1748-0361.2006.00022.x [DOI] [PubMed] [Google Scholar]
- 12.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: 10.1002/cncr.23229 [DOI] [PubMed] [Google Scholar]
- 13.Ambroggi M, Biasini C, Del Giovane C, Fornari F, Cavanna L. Distance as a Barrier to Cancer Diagnosis and Treatment: Review of the Literature. The Oncologist. 2015;20(12):1378–1385. doi: 10.1634/theoncologist.2015-0110 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lin CC, Bruinooge SS, Kirkwood MK, et al. Association Between Geographic Access to Cancer Care, Insurance, and Receipt of Chemotherapy: Geographic Distribution of Oncologists and Travel Distance. J Clin Oncol. 2015;33(28):3177–3185. doi: 10.1200/JCO.2015.61.1558 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Levit LA, Byatt L, Lyss AP, et al. Closing the Rural Cancer Care Gap: Three Institutional Approaches. JCO Oncol Pract. 2020;16(7):422–430. doi: 10.1200/OP.20.00174 [DOI] [PubMed] [Google Scholar]
- 16.Gruen RL, Weeramanthri TS, Knight SE, Bailie RS. Specialist outreach clinics in primary care and rural hospital settings. Cochrane Database Syst Rev. 2004;(1):CD003798. doi: 10.1002/14651858.CD003798.pub2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Tracy R, Saltzman KL, Wakefield DS. Considerations in Establishing Visiting Consultant Clinics in Rural Hospital Communities. J Healthc Manag. 1996;41(2):255. [PubMed] [Google Scholar]
- 18.Charlton M, Schlichting J, Chioreso C, Ward M, Vikas P. Challenges of Rural Cancer Care in the United States. Oncol Williston Park N. 2015;29(9):633–640. [PubMed] [Google Scholar]
- 19.Gruca TS, Nam I, Tracy R. Trends in medical oncology outreach clinics in rural areas. J Oncol Pract. 2014;10(5):e313–320. doi: 10.1200/JOP.2013.001350 [DOI] [PubMed] [Google Scholar]
- 20.Tracy R, Nam I, Gruca TS. The influence of visiting consultant clinics on measures of access to cancer care: evidence from the state of Iowa. Health Serv Res. 2013;48(5):1719–1729. doi: 10.1111/1475-6773.12050 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.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: 10.1002/cncr.30727 [DOI] [PubMed] [Google Scholar]
- 22.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: 10.1097/MLR.0000000000000875 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Research Methods. Dartmouth Atlas of Health Care. Accessed October 19, 2023. https://www.dartmouthatlas.org/research-methods/
- 24.Smith TJ, Desch CE, Grasso MA, et al. The Rural Cancer Outreach Program: clinical and financial analysis of palliative and curative care for an underserved population. Cancer Treat Rev. 1996;22 Suppl A:97–101. doi: 10.1016/s0305-7372(96)90069-9 [DOI] [PubMed] [Google Scholar]
- 25.Drew J, Cashman SB, Savageau JA, Stenger J. The Visiting Specialist Model of Rural Health Care Delivery: A Survey in Massachusetts. J Rural Health. 2006;22(4):294–299. doi: 10.1111/j.1748-0361.2006.00049.x [DOI] [PubMed] [Google Scholar]
- 26.Kellerman R, Ast T, Dorsch J, Frisch L. Itinerant surgical and medical specialist care in Kansas: report of a survey of rural hospital administrators. J Rural Health Off J Am Rural Health Assoc Natl Rural Health Care Assoc. 2001;17(2):127–130. doi: 10.1111/j.1748-0361.2001.tb00269.x [DOI] [PubMed] [Google Scholar]
- 27.Comorbidity SAS Macro (2014 version). Accessed November 1, 2022. https://healthcaredelivery.cancer.gov/seermedicare/considerations/macro-2014.html
- 28.Sundararajan V, Henderson T, Perry C, Muggivan A, Quan H, Ghali WA. New ICD-10 version of the Charlson comorbidity index predicted in-hospital mortality. J Clin Epidemiol. 2004;57(12):1288–1294. doi: 10.1016/j.jclinepi.2004.03.012 [DOI] [PubMed] [Google Scholar]
- 29.Rural Urban Commuting Area Codes Data. Accessed November 17, 2022. https://depts.washington.edu/uwruca/ruca-uses.php
- 30.Bureau UC. 2015–2019 ACS 5-year Estimates. Census.gov. Accessed November 16, 2022. https://www.census.gov/programs-surveys/acs/technical-documentation/table-and-geography-changes/2019/5-year.html
- 31.NPI Files. Accessed November 1, 2022. https://download.cms.gov/nppes/NPI_Files.html
- 32.Hu Y, Wang C, Li R, Wang F. Estimating a Large Travel Time Matrix Between Zip Codes in the United States: A Differential Sampling Approach. J Transp Geogr. 2020;86:102770. doi: 10.1016/j.jtrangeo.2020.102770 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Yu L, Liu YC, Cornelius SL, et al. Telehealth Use Following COVID-19 Within Patient-Sharing Physician Networks at a Rural Comprehensive Cancer Center: Cross-sectional Analysis. JMIR Cancer. 2023;9:e42334. doi: 10.2196/42334 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Longacre CF, Neprash HT, Shippee ND, Tuttle TM, Virnig BA. Evaluating Travel Distance to Radiation Facilities Among Rural and Urban Breast Cancer Patients in the Medicare Population. J Rural Health Off J Am Rural Health Assoc Natl Rural Health Care Assoc. 2020;36(3):334–346. doi: 10.1111/jrh.12413 [DOI] [PubMed] [Google Scholar]
- 35.Xierali IM. Physician Multisite Practicing: Impact on Access to Care. J Am Board Fam Med. 2018;31(2):260–269. doi: 10.3122/jabfm.2018.02.170287 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Wakefield DS, Tracy R, Einhellig J. Trends and implications of visiting medical consultant outpatient clinics in rural hospital communities. Hosp Health Serv Adm. 1997;42(1):49–66. [PubMed] [Google Scholar]
- 37.Frosch ZA, Hicks-Courant K, Bekelman JE, Ko EM, Kanter GP. Multi-site practice and physician travel burden by oncology specialty. J Clin Oncol. 2021;39(15_suppl):e13513-e13513. doi: 10.1200/JCO.2021.39.15_suppl.e13513 [DOI] [Google Scholar]
- 38.Schroen AT, Brenin DR, Kelly MD, Knaus WA, Slingluff CL. Impact of Patient Distance to Radiation Therapy on Mastectomy Use in Early-Stage Breast Cancer Patients. J Clin Oncol. 2005;23(28):7074–7080. doi: 10.1200/JCO.2005.06.032 [DOI] [PubMed] [Google Scholar]
- 39.Celaya MO, Rees JR, Gibson JJ, Riddle BL, Greenberg ER. Travel Distance and Season of Diagnosis Affect Treatment Choices for Women with Early-stage Breast Cancer in A Predominantly Rural Population (United States). Cancer Causes Control. 2006;17(6):851–856. doi: 10.1007/s10552-006-0025-7 [DOI] [PubMed] [Google Scholar]
- 40.Farrow DC, Hunt WC, Samet JM. Geographic Variation in the Treatment of Localized Breast Cancer. N Engl J Med. 1992;326(17):1097–1101. doi: 10.1056/NEJM199204233261701 [DOI] [PubMed] [Google Scholar]
- 41.Voti L, Richardson LC, Reis IM, Fleming LE, MacKinnon J, Coebergh JWW. Treatment of local breast carcinoma in Florida. Cancer. 2006;106(1):201–207. doi: 10.1002/cncr.21557 [DOI] [PubMed] [Google Scholar]
- 42.Nattinger AB, Kneusel RT, Hoffmann RG, Gilligan MA. Relationship of distance from a radiotherapy facility and initial breast cancer treatment. J Natl Cancer Inst. 2001;93(17):1344–1346. doi: 10.1093/jnci/93.17.1344 [DOI] [PubMed] [Google Scholar]
- 43.Athas WF, Adams-Cameron M, Hunt WC, Amir-Fazli A, Key CR. Travel Distance to Radiation Therapy and Receipt of Radiotherapy Following Breast-Conserving Surgery. JNCI J Natl Cancer Inst. 2000;92(3):269–271. doi: 10.1093/jnci/92.3.269 [DOI] [PubMed] [Google Scholar]
- 44.Baldwin LM, Patel S, Andrilla CHA, Rosenblatt RA, Doescher MP. Receipt of recommended radiation therapy among rural and urban cancer patients. Cancer. 2012;118(20):5100–5109. doi: 10.1002/cncr.27488 [DOI] [PubMed] [Google Scholar]
- 45.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 Biomark Prev Publ Am Assoc Cancer Res Cosponsored Am Soc Prev Oncol. 2013;22(10):1657–1667. doi: 10.1158/1055-9965.EPI-13-0404 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Wheeler SB, Kuo TM, Durham D, Frizzelle B, Reeder-Hayes K, Meyer AM. Effects of distance to care and rural or urban residence on receipt of radiation therapy among North Carolina Medicare enrollees with breast cancer. N C Med J. 2014;75(4):239–246. doi: 10.18043/ncm.75.4.239 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.