Abstract
Purpose:
To assess the association between travel distance to an academic health system and overall survival for patients with human papillomavirus (HPV)-associated cancers.
Methods:
Using hospital-based cancer registry data from 2005–2019, we calculated unidirectional travel distance from each patient's geocoded address to our academic health center through network analysis. We categorized distance as short (<25 miles), intermediate (25–74.9 miles), or long (≥75 miles). The primary outcome was time from the date of initial diagnosis to the date of death or last contact. We used multivariable Cox proportional hazards regression to evaluate the association between travel distance and overall survival. We also estimated the adjusted observed 5-year survival rate.
Results:
Patients with HPV-associated cancers traveling distances that were intermediate (hazard ratio [HR], 1.23; 95% CI, 1.06–1.43) and long (HR, 1.15; 95% CI, 1.01–1.32) had a higher hazard of death than the short-distance group. The adjusted 5-year observed survival rates for HPV-associated cancers were lowest in the intermediate-distance group (60.4%) compared with the long-(62.6%) and short-distance (66.2%) groups.
Conclusions:
Our findings indicate that travel distance to an academic health center was associated with overall survival for patients with HPV-associated cancers, reflecting the importance of considering travel burden in improving patient outcomes.
Keywords: 5-year survival, directed acyclic graph, HPV-associated cancers, overall survival, travel distance
Introduction
Travel burden, as measured by travel distance or time, is an important aspect of cancer care. Studies have shown that travel burden can delay diagnosis,1-3 impact treatment,4-6 and affect quality of life for cancer patients.7,8 More importantly, it can influence patient outcomes, including overall survival.9-11
Approximately 46,000 incident cases of human papillomavirus (HPV)-associated cancers, such as cancers of the cervix, vagina, vulva, anus, penis, and oropharynx, occurred annually in the United States.12 In 2019, the incidence of HPV-associated cancers in Oklahoma (15.2 per 100,000) exceeded the national average (12.5 per 100,000) and was the fifth highest in the nation.13 Furthermore, from 2015 to 2019, Oklahoma had the highest mortality rate (3.5 per 100,000 women) and the second-highest incidence rate (9.2 per 100,000 women) for cervical cancer nationally.13
Specialized, complex, and multidisciplinary management are needed for certain HPV-associated cancers. Such services are usually centralized and require patients to travel for care. Data on the impact of travel distance on survival outcomes for HPV-associated cancers in the United States are limited, and results from the few available studies are conflicting. An analysis of a clinic database reported that cervical cancer patients living ≥100 miles from a comprehensive cancer center in Alabama had worse overall survival than patients living <100 miles from the center (hazard ratio [HR], 1.68; 95% CI, 1.11–2.54).11 Other studies reported no association between travel distance and survival for cervical cancer cases.14-16 Conversely, in an investigation using the National Cancer Database sourced from Commission on Cancer–accredited facilities, traveling a long distance for head and neck cancer treatment was associated with improved survival, especially for patients receiving nonsurgical care.17 For other HPV-associated cancers, our literature review yielded no studies assessing the relation between travel burden and survival outcomes.
The present study was undertaken because of the lack of literature on the relation between travel distance and survival for HPV-associated cancers, as well as the importance of this issue to patients, providers, and policymakers. The objective of this study was to evaluate the association between travel distance to an academic health center containing a National Cancer Institute (NCI)-designated cancer center and estimated overall observed survival for patients with HPV-associated cancers from 2005 to 2019. We also estimated the adjusted 5-year survival rate for these patients.
Material and Methods
Data Source and Study Population
We obtained data from the University of Oklahoma cancer registry, which is part of an academic health system known as OU Health. OU Health is primarily located in the center of Oklahoma City and includes the state's only NCI-designated cancer center. The registry is accredited by the American College of Surgeons' Commission on Cancer. Data were collected in accordance with the standards set by the North American Association of Central Cancer Registries.
Cases of HPV-associated cancers were identified according to the International Classification of Diseases for Oncology, Third Edition (ICD-O-3).18 Primary site codes were used to define cases of cervical (C53.0–C53.9), vaginal (C52.9), vulvar (C51.0–C51.9), anal (C20.9, C21.0–C21.8), penile (C60.0–C60.9), and oropharyngeal (C01.9, C02.4, C02.8, C05.1–C05.2, C09.0–C09.1, C09.8–C09.9, C10.0, C10.1–10.4, C10.8–C10.9, C14.0, C14.2, C14.8) cancers.
We restricted our analysis to patients with HPV-associated cancers diagnosed between January 1, 2005, and June 30, 2019 (n = 3,515). We excluded patients who were younger than 18 years (n = 3), only diagnosed through a death certificate (n = 4), incarcerated (n = 95), residing outside of Oklahoma (n = 50), missing the date of diagnosis or last contact information (n = 28), or classified as in situ (n = 115). Further, we excluded patients whose addresses could not be geocoded, as this was required to calculate the exposure of interest (n = 35). Duplicate records of cancer patients were identified by patient medical record number, reviewed manually, and removed (n = 243). Figure 1 provides a flow diagram of the study population.
Figure 1.
Inclusion and Exclusion Flowchart
Exposure and Outcome
The exposure, travel distance, was defined as the unidirectional travel distance from a patient's residence to OU Health. We calculated the shortest travel distance using ArcGIS with the Network Analyst extension (version 10.8; ESRI). For travel distance calculation, we used previously geocoded residential addresses in the cancer registry at the time of cancer diagnosis (n = 2823). If an address was not geocoded, we cleaned and geocoded the address using the address locator in ArcGIS (n = 119). If an address was missing, we used LexisNexis Accurint (LexisNexis Risk Solutions Inc) and Melissa Personator19 to find the address based on the patient's name and date of birth.
As a categorical variable, we classified travel distance as short (<25 miles), intermediate (25–74.9 miles), or long (≥75 miles) based on the geographical layout of our study population. These cut points fit the local context and roughly reflect the urban, suburban, and rural regions of Oklahoma. The first 25 miles include metropolitan Oklahoma City (central Oklahoma) and surrounding urban and suburban areas. The remaining distance groups represent more rural areas of the state, except for metropolitan areas of Tulsa (northeastern Oklahoma) and Lawton (southwestern Oklahoma).
The primary outcome of interest was overall observed survival. Overall survival time was calculated from the date of initial diagnosis of the first HPV-associated cancer to the date of death, last contact, or end of the study period. The secondary outcome of interest was adjusted 5-year observed survival rate for HPV-associated cancers.
Covariates
Patient-level (demographic and clinical factors) and census tract-level (socioeconomic status) characteristics were examined. At the patient level, demographic variables included sex, age, race, ethnicity, and insurance status. Patients with insurance included those with private insurance, Medicare, Medicaid, TRICARE or Veterans Affairs, and Indian Health Service.
Clinical factors included cancer stage at diagnosis (localized, regional, and distant based on the Surveillance, Epidemiology, and End Results Program [SEER] Summary Stage 2000), cancer site, and treatment receipt. If a patient underwent either surgery, chemotherapy, radiation, hormone therapy, or immunotherapy, we categorized treatment receipt as yes; else, it was categorized as either no or unknown.
Census tract-level socioeconomic status indicators, including the proportion of households with total income less than the federal poverty line (0–5%, 5–<10%, 10–<20%, 20–100%) and the proportion of individuals with less than high school or equivalent educational attainment (0–5%, 5–<10%, 10–<20%, 20–100%), were obtained from the 5-year average of the 2014 American Community Survey.20
Analysis
Demographic and clinical characteristics of patients were summarized and compared by distance categories using the χ2 test. To better understand the distribution of patients geographically, we developed a heat map. The heat map was displayed using donut method geomasking to maintain patient confidentiality while trying to preserve the geospatial pattern between the residential locations and survival.21
To test the association between travel distance and overall survival, we performed multivariable Cox proportional hazards regression and estimated unadjusted and adjusted (for race, ethnicity, and insurance status) HRs with 95% CIs. We specified race, cancer stage, and cancer site as potential effect modifiers a priori and evaluated them using a directed acyclic graph for interaction.22 To assess confounding, the relation between variables in the causal diagram were identified and summarized through a literature review and drawn using a directed acyclic graph (DAG).23 The search criteria, relation justifications, model code, and DAG are provided in Supplementary File 1. Based on the minimal sufficient adjustment set obtained through our DAG, we adjusted for race, ethnicity, and insurance status in the multivariable Cox proportional hazards model. There was no evidence of collinearity among covariates according to chi-square association tests.
For survival analyses, we assumed survival times to be independent of each other. The proportional hazards assumption, which states that hazards are proportional and nonoverlapping at all points in time, was verified using survival plots and formal quantitative tests (P values of time-dependent variables, Grambsch and Therneu “zph” test, and the Supremum test for nonproportional hazards). The assessment of assumptions is provided in Supplementary File 2.
We estimated the unadjusted and adjusted 5-year survival rates by distance categories (short, <25 miles; intermediate, 25–74.9 miles; and long, ≥75 miles). We also graphed a directly adjusted survival curve, which plots the Cox proportional hazard survival estimates against time, to display the differences in overall survival by distance categories.
We conducted a series of sensitivity analyses to evaluate different exposure classifications. First, we used a different cut point for travel distance in line with previous work: short (<12.5 miles), intermediate (12.5–49.9 miles), and long (≥50 miles).2,17,24 Second, we evaluated distance as a continuous variable instead of classifying it categorically into distance groups. Third, we fitted a locally weighted scatterplot smoothing (LOESS) curve to martingale residuals based on the continuous distance variable to inform the selection of exposure cut points. We used the Contal and O'Quiley statistical test25 in the %findcut SAS Macro26 to determine cut points. Based on the agreement of the LOESS curve and the statistical test, travel distance was categorized as <21 and ≥21 miles.
Less than 1% of the data in our records were missing, except for the cancer stage variable. Before 2018, over half of the cancer summary stage variable in our data set was classified as unstaged, unknown, or unspecified. These categories for the summary stage were excluded from our primary analysis while still incorporating the patient's data for other variables in the model. However, to assess and compare if discrepancies in staging skewed the findings, we undertook a subanalysis of the summary stage variable, including the unstaged, unknown, or unspecified variable as a separate category from January 1, 2018, to June 30, 2019 (Supplementary File 3).
All analyses were conducted using ArcGIS and SAS (version 9.4; SAS Institute). We used an α level of 0.05 to determine a difference in patient characteristics by distance categories and the presence of effect modification.
Results
Patient Characteristics
Overall, we identified and analyzed 2,942 incident cases of HPV-associated cancers in this study. Demographic, socioeconomic, and clinical characteristics of patients are summarized in Table 1. Most patients traveled a short distance to care (45.9%) compared to long (32.0%) or intermediate (22.1%) distances. The majority of the patient population was non-Hispanic or Latino (92.9%), White (85.2%), female (65.8%), and aged 50–64 years (40.3%). Cervical cancer was the most common HPV-associated cancer (38.7%), followed by oropharyngeal (26.6%), anal (18.3%), vulvar (12.6%), vaginal (2.3%), and penile (1.5%) cancers. We observed differences in travel distance by race (P < .001), ethnicity (P = .04), income (P < .001), and education (P < .001). Other patient characteristics, such as age, sex, cancer site, cancer stage, treatment receipt, and insurance status, were relatively balanced between the travel distance groups.
Table 1.
Characteristics of Patients with HPV-Associated Cancers by Travel Distance Categories, 2005-2019
Characteristic | Travel distance category, n (%) | ||||
---|---|---|---|---|---|
Overall (n = 2,942) | <25 Miles (n = 1,351) | 25-74.9 Miles (n = 650) | ≥75 Miles (n = 941) | P value* | |
Demographic | |||||
Age groups (y) | .280 | ||||
<50 | 995 (33.8) | 463 (34.3) | 207 (31.9) | 325 (34.5) | |
50-64 | 1185 (40.3) | 522 (38.6) | 264 (40.6) | 399 (42.4) | |
65-74 | 471 (16.0) | 225 (16.7) | 113 (17.4) | 133 (14.1) | |
≥75 | 291 (9.9) | 141 (10.4) | 66 (10.2) | 84 (8.9) | |
Sex | .799 | ||||
Male | 1004 (34.1) | 469 (34.7) | 221 (34.0) | 314 (33.4) | |
Female | 1936 (65.8) | 881 (65.3) | 429 (66.0) | 626 (66.6) | |
Race | <.001 | ||||
White | 2507 (85.2) | 1118 (82.8) | 582 (89.5) | 807 (85.8) | |
Black or African American | 179 (6.1) | 137 (10.1) | 11 (1.7) | 31 (3.3) | |
American Indian or Alaska Native | 143 (4.9) | 27 (2.0) | 45 (6.9) | 71 (7.6) | |
Other/unknown** | 113 (3.8) | 69 (5.1) | 12 (1.9) | 32 (3.4) | |
Ethnicity | .039 | ||||
Non-Hispanic or Latino | 2733 (92.9) | 1238 (91.6) | 619 (95.2) | 876 (93.1) | |
Hispanic or Latino | 105 (3.6) | 61 (4.5) | 14 (2.2) | 30 (3.2) | |
Unknown | 104 (3.5) | 52 (3.9) | 17 (2.6) | 35 (3.7) | |
Insurance status | .145 | ||||
Uninsured | 316 (10.7) | 149 (11.0) | 60 (9.2) | 107 (11.4) | |
Insured*** | 2591 (88.1) | 1189 (88.0) | 585 (90.0) | 817 (86.8) | |
Unknown | 35 (1.2) | 13 (1.0) | 5 (0.8) | 17 (1.8) | |
Socioeconomic | |||||
Percent of population living below the federal poverty level**** | <.001 | ||||
0-<5 | 233 (7.9) | 209 (15.5) | 5 (0.8) | 19 (2.0) | |
5-<10 | 456 (15.5) | 278 (20.6) | 86 (13.2) | 92 (9.8) | |
10-<20 | 1209 (41.1) | 355 (26.3) | 380 (58.5) | 474 (50.4) | |
20-100% | 1044 (35.5) | 509 (37.7) | 179 (27.5) | 356 (37.8) | |
Education less than high school or equal (%)**** | <.001 | ||||
0-<5 | 377 (12.8) | 323 (23.9) | 18 (2.8) | 36 (3.8) | |
5-<10 | 605 (20.6) | 323 (23.9) | 114 (17.5) | 168 (17.9) | |
10-<20 | 1398 (47.5) | 433 (32.1) | 427 (65.7) | 538 (57.2) | |
20-100 | 562 (19.1) | 272 (20.1) | 91 (14.0) | 199 (21.2) | |
Cancer site | .343 | ||||
Cervix | 1138 (38.7) | 521 (38.6) | 244 (37.5) | 373 (39.6) | |
Vagina | 68 (2.3) | 32 (2.4) | 13 (2.0) | 23 (2.4) | |
Vulva | 372 (12.6) | 160 (11.8) | 84 (12.9) | 128 (13.6) | |
Penis | 43 (1.5) | 13 (1.0) | 10 (1.5) | 20 (2.1) | |
Anus | 539 (18.3) | 255 (18.9) | 130 (20.0) | 154 (16.4) | |
Oropharynx | 782 (26.6) | 370 (27.4) | 169 (26.0) | 243 (25.8) | |
Clinical | |||||
Stage | .450 | ||||
Localized | 367 (12.5) | 165 (35.8) | 78 (39.4) | 124 (37.2) | |
Regional | 488 (16.6) | 239 (51.8) | 88 (44.4) | 161 (48.4) | |
Distant | 137 (4.7) | 57 (12.4) | 32 (16.2) | 48 (14.4) | |
Treatment | .338 | ||||
No | 104 (3.5) | 48 (3.6) | 21 (3.2) | 35 (3.7) | |
Yes | 2366 (80.4) | 1066 (78.9) | 531 (81.7) | 769 (81.7) | |
Unknown | 472 (16.0) | 237 (17.5) | 98 (15.1) | 137 (14.6) |
HPV, human papillomavirus.
The parametric P value is calculated by χ2 test.
Other/unknown race includes Asian, Native Hawaiian or other Pacific Islanders, and other races. These racial groups were combined due to some groups having fewer than 5 patients.
Patients with insurance included private insurance, Medicare, Medicaid, TRICARE or Veterans Affairs, and Indian Health Service.
Census tract-level variables are from the 5-year average of the 2014 American Community Survey.20
Survival Analysis
We did not observe any effect modification (P > .05 for all interaction terms). In the multivariable model, after adjusting for race, ethnicity, and insurance status, distances that were intermediate (HR, 1.23; 95% CI, 1.06–1.43) and long (HR, 1.15; 95% CI, 1.01–1.32) were associated with a higher hazard of death than short distances (Table 2). Similarly, the adjusted 5-year survival rates were highest in the short-distance group (66.2%) compared with the long-(62.6%) and intermediate-distance (60.4%) groups (Table 2; Figure 3).
Table 2.
Multivariable Analysis of Overall Survival and 5-Year Survival Rates by Distance Categories in Patients with HPV-Associated Cancers, 2005–2019
Overall survival | 5-Year survival rate | |||
---|---|---|---|---|
Distance | Unadjusted HR (95% CI) | Adjusted* HR (95% CI) | Unadjusted % (95% CI) | Adjusted* % (95% CI) |
Short (<25 miles) | Reference | Reference | 66.0 (63.5-68.7) | 66.2 (63.6-68.9) |
Intermediate (25-74.9 miles) | 1.23 (1.06-1.42) | 1.23 (1.06-1.43) | 60.1 (56.5-63.9) | 60.4 (57.1-64.4) |
Long (≥75 miles) | 1.14 (1.00-1.31) | 1.15 (1.01-1.32) | 62.3 (59.2-65.6) | 62.6 (59.5-65.9) |
HPV, human papillomavirus; HR, hazard ratio.
Adjusted for race, ethnicity, insurance status.
Figure 3.
Directly Adjusted Survival Curves with 95% CIs for Patients with HPV-Associated Cancers by Distance Groups in Miles, 2005–2019 (n = 2,942)
Figure 2.
Geographical Distribution of Patients with HPV-Associated Cancers
Sensitivity and Subanalysis
In our analysis of alternate classifications of travel distance, 26.1% of patients traveled short distances (<12.5 miles), 32.3% traveled intermediate distances (12.5–49.9 miles), and 41.6% traveled long distances (≥50 miles) to care (Supplementary File 3). Similar to our primary analysis, there was no effect modification by race, cancer stage, or cancer site. However, unlike our primary analysis, there was no association between travel distance and overall survival in the adjusted multivariable model (P = .28). In addition, the hazard of death did not differ among those in the intermediate-distance group (HR, 0.93; 95% CI, 0.79–1.09) and the long-distance group (HR, 1.04; 95% CI, 0.89–1.20) compared with the short-distance group (Supplementary File 3, Table S5).
In our secondary sensitivity analysis of dichotomized travel distance based on the LOESS curves, 42.6% of patients traveled short distances (<21.0 miles) and 57.4% traveled intermediate and long distances (≥21.0 miles). Like our first sensitivity analysis, there was no evidence of effect modification (P > .05 for all interaction terms). However, there was an association between travel distance and overall survival in the adjusted multivariable model (P = .002). After adjusting for race, ethnicity, and insurance status, the hazard of death among those in the intermediate- and long-distance groups was 1.21 times greater (95% CI, 1.07–1.37) than those in the short-distance group (Supplementary File 3, Table S6).
In our subanalysis of the summary stage variable including unstaged, unknown, or unspecified values from January 1, 2018, to June 30, 2019, 48.3% of patients traveled short distances, 16.6% traveled intermediate distances, and 35.1% traveled long distances. We did not observe any effect modification, and travel distance was not associated with overall survival (P = .40) (Supplementary File 3). Furthermore, when compared to those in the short-distance group, the hazard of death did not differ among those in the intermediate-distance group (HR, 1.60; 95% CI, 0.81–3.18) or the long-distance group (HR, 1.26; 95% CI, 0.68–2.31) (Supplementary File 3, Table S8).
Discussion
In this study of patients with HPV-associated cancers from 2005 to 2019, travel distance to an academic health center was associated with overall survival. Specifically, patients who traveled intermediate (25–74.9 miles) and long (≥75 miles) distances for care had poorer overall survival. Similarly, the adjusted 5-year survival rate for patients with HPV-associated cancers traveling intermediate and long distances was comparatively lower than that of patients traveling short distances. Furthermore, in our sensitivity analyses, travel distance was still associated with overall survival when distance was dichotomized based on LOESS curves (short, <21 miles; intermediate and long, ≥21 miles). However, there was no association when travel distance was classified as short (<12.5 miles), intermediate (12.5–49.9 miles), and long (>50.0 miles), indicating that exposure classification may influence the relation between travel distance and overall survival for HPV-associated cancers.
Our finding that intermediate and longer travel distances to care led to poorer survival may stem from multiple factors. Patients could face issues with transportation, including lack of social support, inability to drive or drive long distances, lack of a reliable vehicle or public transportation option, and time and costs, among other factors.27 In the present study, we did not ascertain the mode of transportation or explore issues with transportation. However, future studies examining the influence of travel burden could benefit from investigating these factors. Another reason for poorer survival among patients with travel burden could be differences in treatment and treatment compliance.4,7,28 In a study of patients with cervical cancer, distance to the treatment center was an important factor for completing radiation therapy in the recommended time frame.28 Several studies have also reported advanced cancer stage at diagnosis among patients traveling long distances for care,1,2,7 likely leading to poorer outcomes. However, in our analysis, the proportion of patients with advanced-stage disease traveling intermediate or long distances was slightly, but not significantly, elevated.
Beyond improvements in early detection, treatment, and care, outreach efforts, telemedicine, patient programs, and cancer navigation services could improve overall survival and minimize the gap in survival observed by travel distance. For instance, patient navigation systems have increased adherence to treatment care29,30 and improved timeliness of care treatment,31 and could reduce disparities in underserved populations.32 OU Health offers patient navigation services, including specialized services for American Indians.33
Our findings are subject to several limitations. First, as cancer registries do not routinely assess HPV status, information on HPV DNA status in cancer tissue was unavailable. Therefore, HPV-associated cancers were defined as cancers at specific anatomic sites in which HPV DNA is frequently found. Second, selection bias could occur if patients living and traveling farther from our academic center are less likely to be captured in our registry. It could also arise if patients from rural areas captured in our registry were referred to our academic health center, which includes an NCI-designated cancer center, based on cancer severity or complications. In this case, selection bias may make it appear that patients living in rural areas or traveling long distances have advanced cancer and a potentially higher hazard of death. Adjustment for disease severity through standard methods may be inappropriate as travel distance could affect severity through its impact on treatment and treatment completion, among other factors.34 Third, addresses on a rural route or listed as a PO box are less likely to be geocoded35 and thus were excluded. We tried to locate the addresses based on patient information and were unable to geocode less than 1% of the addresses. However, excluding these patients may introduce selection bias if the association between travel distance and overall survival differs from those included in our study. Fourth, several factors could influence the accuracy and quality of geocoded data, especially among addresses in rural counties of Oklahoma,36 which may lead to measurement error in travel distance calculation, particularly for patients traveling long distances. In addition, 732 records were geocoded at the zip code and city level, which may lead to differential measurement error as addresses in rural counties were more likely to be geocoded at the zip code and city level. We did not calculate travel times in this study, as they would be subject to the same errors along with traffic and congestion, road and weather conditions, and time of the day and season. Fifth, as we used patient addresses at diagnosis only and not during or after treatment, there could be potential misclassification. For example, patients could have changed residence during treatment, or those traveling long distances could have stayed closer for care. Since these data were not uniformly captured in our registry, we could not assess the influence of misclassification on our estimates. Sixth, over half of the data for cancer stage were unknown, unspecified, or unstaged before 2018, warranting a subanalysis with data from January 1, 2018, to June 30, 2019, which only had about 4.2% unknown, unspecified, or unstaged data. Seventh, cause-specific mortality data in our registry were incomplete. Therefore, we could not undertake competing risk analysis to account for causes of death other than HPV-associated cancers. Future studies should evaluate competing risk events and account for the assumption that the censoring mechanism is informative.37 Lastly, as our study was conducted using registry data from an academic health center and limited to patients from Oklahoma, our findings need to be interpreted in this context, which may limit generalizability to other settings and patients with HPV-associated cancers. Furthermore, our cancer registry is not population-based. From 2005 to 2018, 1,097 cases of cervical cancer were included in our registry, representing 43.3% of the incident cervical cancer cases in the state of Oklahoma. In the same period, our registry only included 13.7% of all patients with penile cancer statewide.
Despite these limitations, the present study is the first to assess the association between travel distance and overall survival for HPV-associated cancers. We found that travel distance influences overall survival for HPV-associated cancers at certain cut points. Future quantitative and qualitative research could further improve our understanding of the travel burden and identify barriers and areas for improvement through surveys, interviews, and focus groups. Future studies may account for competing risk events, include risk factors for HPV-associated cancers in analyses, adjust for selection bias, and assess unmeasured confounding, when appropriate. Future work may also benefit from assessing and understanding health care system characteristics across the cancer control continuum. Strengthening navigation programs, coordinating follow-up and long-term care, and increasing access may improve survival outcomes for patients with HPV-associated cancers.
Acknowledgments
The authors are grateful for the efforts of Kathy J. Kyler (Staff Editor, Office of the Vice President for Research, University of Oklahoma Health Sciences Center) in preparing this manuscript for publication.
Footnotes
Supplementary material for this article is available for download at the following URL: https://www.ncra-usa.org/Portals/68/Journal%20of%20Registry%20Management/Gopalani_HPV_Supplementary%20Files%20Final.pdf
SVG, LF, and HDD are supported by the Hudson Fellows in Public Health program through the University of Oklahoma Health Sciences Center. SC, JEC, and JDP were partially supported by the Oklahoma Shared Clinical and Translational Resources (U54GM104938) with an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences. SC and JEC were partially supported by the National Cancer Institute Cancer Center Support (Grant P30CA225520) awarded to the University of Oklahoma Stephenson Cancer Center for the use of the Biostatistics and Research Design Shared Resources. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
The study was approved by the University of Oklahoma Health Sciences Center Institutional Review Board (IRB #12619) and was conducted in compliance with its requirements.
References
- 1.Farquhar DR, Masood MM, Lenze NR, et al. Travel time to provider is associated with advanced stage at diagnosis among low income head and neck squamous cell carcinoma patients in North Carolina. Oral Oncol. 2019;89:115-120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Massarweh NN, Chiang YJ, Xing Y, et al. Association between travel distance and metastatic disease at diagnosis among patients with colon cancer. J Clin Oncol. 2014;32(9):942-948. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Scoggins JF, Fedorenko CR, Donahue SM, Buchwald D, Blough DK, Ramsey SD. Is distance to provider a barrier to care for Medicaid patients with breast, colorectal, or lung cancer? J Rural Health. 2012;28(1):54-62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Spees LP, Wheeler SB, Varia M, et al. Evaluating the urban-rural paradox: the complicated relationship between distance and the receipt of guideline-concordant care among cervical cancer patients. Gynecol Oncol. 2019;152(1):112-118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Goyal S, Chandwani S, Haffty BG, Demissie K. Effect of travel distance and time to radiotherapy on likelihood of receiving mastectomy. Ann Surg Oncol. 2015;22(4):1095-1101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Boscoe FP, Johnson CJ, Henry KA, et al. Geographic proximity to treatment for early stage breast cancer and likelihood of mastectomy. Breast. 2011;20(4):324-328. [DOI] [PubMed] [Google Scholar]
- 7.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]
- 8.Thomas AA, Gallagher P, O'Céilleachair A, Pearce A, Sharp L, Molcho M. Distance from treating hospital and colorectal cancer survivors' quality of life: a gendered analysis. Support Care Cancer. 2015;23(3):741-751. [DOI] [PubMed] [Google Scholar]
- 9.Daruvala A, Lucas FL, Sammon J, Darus C, Bradford L. Impact of geography and travel distance on outcomes in epithelial ovarian cancer: a national cancer database analysis. Int J Gynecol Cancer. 2021;31(2):209-214. [DOI] [PubMed] [Google Scholar]
- 10.Loree JM, Javaheri KR, Lefresne SV, et al. Impact of travel distance and urban-rural status on the multidisciplinary management of rectal cancer. J Rural Health. 2017;33(4):393-401. [DOI] [PubMed] [Google Scholar]
- 11.Barrington DA, Dilley SE, Landers EE, et al. Distance from a comprehensive cancer center: a proxy for poor cervical cancer outcomes? Gynecol Oncol. 2016;143(3):617-621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Centers for Disease Control and Prevention. Cancers associated with human papillomavirus, United States, 2014–2018. USCS Data Brief. 2021. [Google Scholar]
- 13.US Cancer Statistics Working Group. US Cancer Statistics Data Visualizations Tool, based on 2021 submission data (1999-2019): US Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute. Accessed October 26, 2022. https://www.cdc.gov/cancer/dataviz
- 14.Hung P, Zahnd WE, Brandt HM, Adams SA, Wang S, Eberth JM. Cervical cancer treatment initiation and survival: the role of residential proximity to cancer care. Gynecol Oncol. 2021;160(1):219-226. [DOI] [PubMed] [Google Scholar]
- 15.Rauh LA, Saks EJ, Nakad-Rodriguez D, Showalter TN, Duska LR. Cervical cancer care in rural Virginia: the impact of distance from an academic medical center on outcomes & the role of non-specialized radiation centers. Gynecol Oncol. 2018;150(2):338-342. [DOI] [PubMed] [Google Scholar]
- 16.Gunderson CC, Nugent EK, McMeekin DS, Moore KN. Distance traveled for treatment of cervical cancer: who travels the farthest, and does it impact outcome? Int J Gynecol Cancer. 2013;23(6):1099-1103. [DOI] [PubMed] [Google Scholar]
- 17.Graboyes EM, Ellis MA, Li H, et al. Racial and ethnic disparities in travel for head and neck cancer treatment and the impact of travel distance on survival. Cancer. 2018;124(15):3181-3191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.World Health Organization. International Classification of Diseases for Oncology (ICD-O). 3rd ed. World Health Organization; 2013. [Google Scholar]
- 19.Melissa Lookups Personator. Melissa Lookups website. Accessed August 10, 2020. https://lookups.melissa.com/home/personator [Google Scholar]
- 20.American Community Survey 5-Year data (2009-2019). US Census Bureau website. Published December 8, 2022. https://www.census.gov/data/developers/data-sets/acs-5year.2014.html
- 21.Hampton KH, Fitch MK, Allshouse WB, et al. Mapping health data: improved privacy protection with donut method geomasking. Am J Epidemiol. 2010;172(9):1062-1069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Nilsson A, Bonander C, Strömberg U, Björk J. A directed acyclic graph for interactions. Int J Epidemiol. 2021;50(2):613-619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. 23.Textor J, van der Zander B, Gilthorpe MS, Liskiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package ‘dagitty’. Int J Epidemiol. 2016;45(6):1887-1894. [DOI] [PubMed] [Google Scholar]
- 24.Vetterlein MW, Löppenberg B, Karabon P, et al. Impact of travel distance to the treatment facility on overall mortality in US patients with prostate cancer. Cancer. 2017;123(17):3241-3252. [DOI] [PubMed] [Google Scholar]
- 25.Contal C, O'Quigley J. An application of changepoint methods in studying the effect of age on survival in breast cancer. Comp Stat Data Anal. 1999;30(3):253-270. [Google Scholar]
- 26.Williams BA, Mandrekar JN, Mandrekar SJ, Cha SS, Furth AF. Finding Optimal Cutpoints for Continuous Covariates with Binary and Time-To-Event Outcomes. Mayo Clinic; 2006. [Google Scholar]
- 27.Charlton M, Schlichting J, Chioreso C, Ward M, Vikas P. Challenges of Rural Cancer Care in the United States. Oncology (Williston Park). 2015;29(9):633-640. [PubMed] [Google Scholar]
- 28.Cohen J, Harper A, Nichols EM, Rao GG, Mohindra P, Roque DM. Barriers to timely completion of radiation therapy in patients with cervical cancer in an urban tertiary care center. Cureus. 2017;9(9):e1681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wells KJ, Battaglia TA, Dudley DJ, et al. Patient navigation: state of the art or is it science? Cancer. 2008;113(8):1999-2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ell K, Vourlekis B, Xie B, et al. Cancer treatment adherence among low-income women with breast or gynecologic cancer: a randomized controlled trial of patient navigation. Cancer. 2009;115(19):4606-4615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Freund KM, Battaglia TA, Calhoun E, et al. Impact of patient navigation on timely cancer care: the Patient Navigation Research Program. J Natl Cancer Inst. 2014;106(6):dju115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Natale-Pereira A, Enard KR, Nevarez L, Jones LA. The role of patient navigators in eliminating health disparities. Cancer. 2011;117(15 suppl):3543-3552. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Cancer Navigation Services. Stephenson Cancer Center website. Accessed July 8, 2021. https://www.ouhealth.com/stephenson-cancer-center/cancer-patients-families/cancer-navigation-services/ [Google Scholar]
- 34.Caniglia EC, Zash R, Swanson SA, et al. Methodological challenges when studying distance to care as an exposure in health research. Am J Epidemiol. 2019;188(9):1674-1681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Ha S, Hu H, Mao L, Roussos-Ross D, Roth J, Xu X. Potential selection bias associated with using geocoded birth records for epidemiologic research. Ann Epidemiol. 2016;26(3):204-211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Dilekli N, Janitz A, Campbell J. improved geocoding of cancer registry addresses in urban and rural Oklahoma. J Registry Manag. 2020;47(1):13-20. [PubMed] [Google Scholar]
- 37.Satagopan JM, Ben-Porat L, Berwick M, Robson M, Kutler D, Auerbach AD. A note on competing risks in survival data analysis. Br J Cancer. 2004;91(7):1229-1235. [DOI] [PMC free article] [PubMed] [Google Scholar]