If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Northwestern Quality Improvement, Research, & Education in Surgery, Northwestern University, Feinberg School of Medicine, Chicago, IllDepartment of Surgery, Canning Thoracic Institute, Northwestern University, Feinberg School of Medicine, Chicago, Ill
Northwestern Quality Improvement, Research, & Education in Surgery, Northwestern University, Feinberg School of Medicine, Chicago, IllDepartment of Surgery, Canning Thoracic Institute, Northwestern University, Feinberg School of Medicine, Chicago, Ill
Northwestern Quality Improvement, Research, & Education in Surgery, Northwestern University, Feinberg School of Medicine, Chicago, IllSurgery Service, Jesse Brown VA Medical Center, Chicago, Ill
Address for reprints: David D. Odell, MD, MMSc, Department of Surgery, Surgical Outcomes and Quality Improvement Center, Northwestern University, Feinberg School of Medicine, 633 N St Clair St, 20th Floor, Chicago, IL 60611.
Northwestern Quality Improvement, Research, & Education in Surgery, Northwestern University, Feinberg School of Medicine, Chicago, IllDepartment of Surgery, Canning Thoracic Institute, Northwestern University, Feinberg School of Medicine, Chicago, Ill
Regionalization of surgery for non–small cell lung cancer (NSCLC) to high-volume centers (HVCs) improves perioperative outcomes but frequently increases patient travel distance. Travel might decrease rates of adjuvant chemotherapy (AC) use, however, the relationship of distance, volume, and receipt of AC with outcomes is unknown. Our objective was to evaluate the association of distance, volume, and receipt of AC with overall survival among patients with NSCLC.
Methods
Patients with stage I to IIIA (N0-N1) NSCLC were identified between 2004 and 2018 using the National Cancer Database. Distance to surgical facility was categorized into quartiles (<5.1, 5.1 to <11.5, 11.5 to <28.1, and ≥28.1 miles), and HVCs were defined as those that perform ≥40 annual resections. Patient characteristics and likelihood of receiving AC anywhere were determined. Propensity score-matched survival analysis was performed using Cox models and Kaplan–Meier curves.
Results
Of the 131,982 patients included, 35,658 (27.0%) were stage II to IIIA. Of the stage II to IIIA cohort, 49.6% received AC, 13.1% traveled <5.1 miles to low-volume centers (LVCs), and 18.1% traveled ≥28.1 miles to HVCs (P < .001). Among stage II to IIIA patients who traveled ≥28.1 miles to HVCs, 45% received AC versus 51.5% who traveled <5.1 miles to LVCs (incidence rate ratio, 0.88; 95% CI, 0.83-0.94; <5.1 miles to LVC reference). Patients with stage II to IIIA NSCLC who traveled ≥28.1 miles to HVCs and did not receive AC had higher mortality rates than those who traveled <5.1 miles to LVCs and received AC (median overall survival, 52.3 vs 36.7 months; adjusted hazard ratio, 1.41; 95% CI, 1.26-1.57).
Conclusions
Increasing travel distance to surgical treatment is associated with decreased likelihood of receiving AC for patients with stage II to IIIA (N0-N1) NSCLC.
Because regionalization of complex surgery results in increased patient travel to receive care, it is important that vital oncological services remain accessible.
Use of adjuvant chemotherapy for lung cancer is becoming more common. However, regionalization of surgery has increased distance patients travel to receive care. This increased travel distance to high-volume hospitals is associated with decreased likelihood of receiving adjuvant chemotherapy. Nonreceipt of adjuvant chemotherapy is associated with worse survival.
Despite recent improvements in survival, lung cancer remains the foremost cause of cancer death in the United States and the world for men and women.
Non–small cell lung cancer (NSCLC) is the most common form of lung cancer among patients who undergo surgical resection, and hospital surgical volume has been established as a having an inverse relationship with mortality in complex or high-risk cases.
These findings have led to proposals for hospital and surgeon surgical volume minimums, and a trend toward regionalization (centralization) of complex surgery to high-volume centers (HVCs).
Increased travel distance might also lead to decreased access to care after surgical resection, possibly because of loss of care coordination resulting in care fragmentation when more than 1 treatment team is involved.
Association between geographic access to cancer care, insurance, and receipt of chemotherapy: geographic distribution of oncologists and travel distance.
Because of these findings, we hypothesize that: 1) patients with NSCLC who travel long distances for surgical resection are less likely to receive adjuvant chemotherapy (AC), and 2) patients who travel long distances to HVCs for surgery and do not receive indicated AC have worse overall survival (OS) than patients who travel short distances to low-volume centers (LVCs) and successfully receive AC. Therefore, the objective of this study was to evaluate the association of travel distance, hospital surgical volume, and receipt of AC with OS for patients with resected NSCLC.
Methods
Data Source and Study Cohort
The National Cancer Database (NCDB) is a large hospital-based cancer registry.
More than 1500 hospitals contribute to the NCDB, and it is estimated to capture approximately 82% of cancers of the lung and bronchus in the United States.
Data are entered by trained registrars, and the database undergoes regular audits to confirm accuracy and completeness. The NCDB was used to retrospectively identify surgically treated NSCLC patients diagnosed from 2004 to 2017 with outcomes through 2018. Patients were excluded if they had multiple malignancies, received neoadjuvant chemotherapy, underwent surgical resection at a nonreporting facility (so that travel distance to surgery is accurate), had interruptions in facility reporting from 2004 to 2018, had nonintact geographic data, had 1-way travel distance >250 miles, had missing tumor grade or size, or they did not meet the widely accepted definition of resectable (stage I-IIIA; N0-N1; Figure 1). Patient information in the NCDB is deidentified and use in research was determined to be exempt from review by our institution's institutional review board.
Figure 1Inclusion criteria flow diagram. The National Cancer Database was used to identify patients with non–small cell lung cancer diagnosed between 2004 and 2017 with outcomes available through 2018. Population excluded at each stage of analysis is detailed. PUF, Participant User File.
Travel distance to the reporting facility was determined by calculating the distance between the centroid of the patient's zip code to the address of the reporting facility where surgery was performed. Travel distance was evaluated as a continuous variable and also categorized into quartiles (<5.1, 5.1 to <11.5, 11.5 to <28.1, and ≥28.1 to 250 miles) for the stage I to IIIA (N0-N1) cohort. Facility geographic region was categorized into 9 US Census Divisions. The United States Department of Agriculture Economic Research Service publishes a 9-level rural-urban continuum code, and “metro” was defined as rural-urban continuum codes 1 to 3, and “nonmetro” was defined as rural-urban continuum codes 4 to 9. Patients with missing region, rurality, travel distance, and those who traveled >250 miles were excluded.
Facility surgical volume was determined by the mean number of resections performed annually by facilities without missing years of NCDB reporting. Annual surgical volume was evaluated as continuous variable as well as a categorical variable. HVCs versus LVCs were defined using LeapFrog criteria with a cutoff of ≥40 annual pulmonary resections.
Agency for Healthcare Research and Quality Safety in numbers: hospital performance on Leapfrog's surgical volume standard based on results of the 2019 Leapfrog Hospital Survey.
Program type, as determined by Commission on Cancer accreditation, was categorized as academic, comprehensive, integrated, and community. Designations reflect program-level capabilities and organizational characteristics. Care fragmentation was categorized as those who received care at multiple facilities versus a single facility.
Treatment Characteristics
All included patients underwent surgical resection for cure, with a pathological specimen obtained, at the reporting facility, with a known extent of resection. Extent of resection was categorized as wedge, segmentectomy, lobectomy, or pneumonectomy. Surgical margin status was categorized as negative (R0) versus positive (R1, R2, or residual tumor present). Number of intraoperative lymph nodes sampled was included as a continuous variable. Patients were considered to have received AC if treatment was administered at any facility (including facilities other than the facility where they received surgical treatment).
Patient Characteristics
Patient age at diagnosis was included as a continuous variable. Race and ethnicity were categorized as non-Hispanic White, non-Hispanic Black, Hispanic, Asian (including Pacific Islander, East Asian, Southeast Asian, and South Asian), and other or unknown race or ethnicity (American Indian, Alaska Native, and “other” or “unknown” categories, which are NCDB categories). Educational attainment was on the basis of zip code tabulation area estimates, matched with patient year of diagnosis, as quartiles of the population aged ≥25 years without a high school degree. Income level was similarly on the basis of zip code tabulation area estimates matched to patient year of diagnosis and was categorized as quartiles. Insurance status was categorized as uninsured or Medicaid, Medicare, or private, and other. Patients with missing education, income, or insurance data were excluded. The NCDB reports comorbidities with the Charlson–Deyo score, which was grouped as 0, 1, 2, and ≥3. Tumor size was categorized as <1, ≥1, >1-2, >2-3, >3-5, >5-7, or >7 cm. Tumor grade was categorized as well differentiated, moderately differentiated, poorly differentiated, and undifferentiated. Histology was categorized as adenocarcinoma, squamous, large cell, carcinoid, and other (Table E1). American Joint Committee on Cancer stage was categorized as pathological stage I, stage II, and stage IIIA. Nodal stage was categorized to N0, N1, or N2 to N3. Patients with pathological stage 0 or stage IV, N2 to N3 disease, and missing tumor size data were excluded.
Statistical Analysis
The χ2 test was used to determine the patient population differences for those who underwent AC and those who did not. Continuous variables are reported as mean and SD if normally distributed, and median and interquartile range (IQR) if not normally distributed. The t test and Mann–Whitney test were used as appropriate. Changes in rates of AC receipt over time were assessed using Cochran–Armitage tests. P values were 2-sided.
Hypothesis 1: likelihood of receipt of AC
To assess hypothesis 1, Poisson regression was performed to evaluate the association of travel distance and hospital surgical volume with receipt of AC in the stage II to IIIA (N0-N1) patients. Model 1 was used to evaluate likelihood of receipt of AC with annual hospital volume and travel distance as continuous variables with an interaction term, whereas model 2 was used to evaluate hospital volume and travel distance subgroups. Both models were similarly adjusted for covariates. Standard errors were adjusted for clustering within facilities.
Hypothesis 2: survival analysis
To assess hypothesis 2, Cox proportional hazards models with Breslow method for ties were used to evaluate the simultaneous association of travel distance, hospital surgical volume, and receipt of AC with OS of stage II to IIIA (N0-N1) patients. Model 3 was used to evaluate subgroups of volume, travel distance, and AC, whereas model 4 was used to evaluate AC as a dichotomous variable and annual surgical volume and travel distance as continuous variables with an interaction term. Both models were similarly adjusted for covariates with robust SEs adjusted for clustering within facilities.
Next, patients who traveled the lowest quartile of travel distance were compared with those who traveled the highest quartile of travel distance. Propensity scores were generated for the probability of receipt of AC, adjusted for clustering and for covariates. Calipers were set to 0.2 multiplied by the SD, and nearest-neighbor matching with no replacement was performed using logit of the propensity score. All parameters were evaluated and were determined to have standardized mean differences within 10%.
Cox models were used to examine differences in mortality for travel distance, surgical volume, and receipt of AC in the propensity score-matched cohort with robust standard errors adjusted for clustering within facilities (model 5, subgroups; model 6, continuous variables). Sensitivity analyses were conducted that excluded perioperative 90-day mortality (model 7).
Kaplan–Meier survival estimates with log rank tests were used to determine the significance of differences in survival of patients grouped according to those who traveled short distances to undergo surgery at LVCs and received AC versus patients who traveled long distances to undergo surgery at HVCs but did not receive AC. Kaplan–Meier curves were generated for comparisons at stage II to IIIA (N0-N1). A Kaplan–Meier curve was also generated for comparison of stage I patients to evaluate if the association with survival was similar at early stages, which do not typically receive AC. All analyses were done with Stata version 17 (StataCorp).
Results
Overall, 131,982 patients with stage I to IIIA (N0-N1) NSCLC met criteria for inclusion with 34,658 (27.0%) stage II to IIIA (N0-N1; Figure 1). Patients were surgically treated at 758 facilities with continuous reporting for all years of the study. A total of 222 (29.3%) facilities that met HVC criteria (≥40 annual pulmonary resections) treated 74,570 (56.5%) patients. Rates of perioperative 90-day mortality were superior at HVCs compared with LVCs in the overall stage I to IIIA (N0-N1) cohort (4.1% vs 5.2%), and in the stage II to IIIA (N0-N1) cohort (6.2% vs 7.5%; all P < .001).
Of the total cohort of 131,982 patients, 33,540 (18.6%) received AC (Table 1). For stage II to IIIA (N0-N1) patients, 17,195 (49.6%) received AC at any facility with a median time to initiation of 47 (95% CI, 36-61) days from surgical resection. Patient AC refusal rate was 7.5% and was not significantly different according to travel distance. Only 16.5% of stage II to IIIA (N0-N1) patients who traveled long distances (28.1-250 miles) to HVCs received AC at the same facility as their surgical treatment, whereas those who traveled short distances (<5.1 miles) for surgical care had significantly higher rates of same-site AC at HVCs (29.6%) and LVCs (26.2%; P < .001). Rates of AC at any facility increased for stage II to IIIA (N0-N1) patients from 40.1% in 2004 to 50.4% in 2017 with persistent disparities according to travel distance (P < .001 for trend; Table 2, Figure 2, Figure E1). Patients who traveled short distances (<5.1 miles) to LVCs and received AC (L1C subgroup) had a shorter median time to initiation of AC than patients who traveled short distances (<5.1 miles) to HVCs and received AC or patients who traveled long distances (28.1-250 miles) to HVCs and received AC (46 [IQR, 35-61] vs 48 [IQR, 37-63] and 49 [IQR, 39-63] days, respectively; all P < .001). Women were less likely to travel long distances (28.1-250 miles) for surgical treatment (24.7% vs 28.0% for men) and had lower rates of receipt of AC overall (P < .001; Table 2).
Table 1Population characteristics and rate of receipt of adjuvant chemotherapy at any facility for patients with resected stage I to IIIA (N0-N1) NSCLC
Table 2Population characteristics of stage II to IIIA (N0-N1) patients and Poisson regression model 1 to evaluate the association of travel distance, hospital surgical volume, and receipt of AC at any facility
Figure 2Use of adjuvant chemotherapy (AC) for lung cancer is becoming more common. However, regionalization of surgery has increased travel distance, and increased travel distance to high-volume hospitals is associated with decreased likelihood of receiving AC. Nonreceipt of AC is associated with worse survival. IRR, Incidence rate ratio; CI, confidence interval; HR, hazard ratio.
Hypothesis 1: Multivariable Analysis to Evaluate Receipt of AC
The likelihood of receiving AC was evaluated in the stage II to IIIA (N0-N1) cohort with multivariable Poisson regression (model 1). Model 1 showed an inverse relationship of increasing distance and likelihood of receipt of AC (Table 2; Figure 2). Further analysis with adjusted multivariable Poisson regression of stage II to IIIA (N0-N1) distance and surgical volume subgroups (model 2) was performed. Model 2 showed an inverse relationship of increasing distance and decreasing likelihood of receipt of AC for patients treated at LVCs and HVCs (Figure 3).
Figure 3Forest plot of association between increasing travel distance and receipt of adjuvant chemotherapy (AC) for patients with resected stage II to IIIA (N0-N1) non–small cell lung cancer treated at high- and low-volume centers and Poisson regression model 2 on likelihood of receipt of AC. Low volume, <40 annual surgical resections; high volume, ≥40 annual surgical resections. Distance is in miles. IRR, Incidence rate ratio; CI, confidence interval.
Hypothesis 2: Association of Travel Distance, Hospital Volume, and Receipt of AC With Survival
Adjusted Cox proportional hazards models were used to evaluate the differences in OS for the stage II to IIIA (N0-N1) cohort for distance, surgical volume, and receipt of AC subgroups (model 3; Table 3; Table E2). The propensity score-matched cohort was then evaluated (model 5), and patients who traveled long distances (28.1-250 miles) for surgical treatment at HVCs and did not receive AC (H4N subgroup) had increased risk of death (median OS, 36.7 months; adjusted hazard ratio [aHR], 1.41; 95% CI, 1.26-1.57) compared with patients who traveled short distances (<5.1 miles) and were surgically treated at LVCs and received AC (L1C subgroup; median OS, 52.3 months, reference; Table 4; Figure 4). Patients who traveled short distances (<5.1 miles) to HVCs and successfully received AC had superior outcomes (aHR, 0.86; 95% CI, 0.78-0.96) versus L1C patients (Table 4).
Table 3Travel distance, hospital surgical volume, and receipt of AC at any facility stage II to IIIA (N0-N1) subgroups and Cox proportional hazards model 3
Subgroups denotation uses the following pattern: HVC (H) versus LVC (L)/travel distance quartile/received AC (C) versus did not receive AC (N). Please see Table E2 for detailed definitions.
<.001
L1C
2412
0.0
100.0
Reference
L1N
2273
100.0
0.0
1.43 (1.32-1.55)
H1C
1841
0.0
100.0
0.92 (0.84-1.00)
H1N
1812
100.0
0.0
1.34 (1.23-1.47)
L2C
2115
0.0
100.0
1.01 (0.93-1.09)
L2N
1916
100.0
0.0
1.44 (1.33-1.57)
H2C
2346
0.0
100.0
0.95 (0.88-1.04)
H2N
2102
100.0
0.0
1.42 (1.30-1.55)
L3C
1690
0.0
100.0
1.00 (0.92-1.08)
L3N
1668
100.0
0.0
1.45 (1.33-1.58)
H3C
2630
0.0
100.0
0.94 (0.86-1.02)
H3N
2675
100.0
0.0
1.38 (1.27-1.50)
L4C
1250
0.0
100.0
0.94 (0.85-1.04)
L4N
1466
100.0
0.0
1.46 (1.32-1.61)
H4C
2911
0.0
100.0
0.94 (0.86-1.02)
H4N
3551
100.0
0.0
1.39 (1.28-1.51)
Model 3 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled. AC, Adjuvant chemotherapy; HR, hazard ratio; CI, confidence interval.
∗ AC at any facility.
† Data are presented as n with percentages except where otherwise noted.
‡ Subgroups denotation uses the following pattern: HVC (H) versus LVC (L)/travel distance quartile/received AC (C) versus did not receive AC (N). Please see Table E2 for detailed definitions.
Table 4Population characteristics of the stage II to IIIA (N0-N1) propensity score-matched cohort and Cox proportional hazards model 5 to evaluate the association of travel distance, hospital surgical volume, and receipt of AC at any facility with survival
Subgroups denotation uses the following pattern: HVC (H) versus LVC (L)/travel distance quartile/received AC (C) versus did not receive AC (N). Please see Table E2 for detailed definitions.
<.001
L1C
1603
0.0
100.0
Reference
L1N
1611
100.0
0.0
1.40 (1.28-1.53)
L4C
934
0.0
100.0
0.89 (0.79-1.01)
L4N
931
100.0
0.0
1.43 (1.27-1.62)
H1C
1218
0.0
100.0
0.86 (0.78-0.96)
H1N
1223
100.0
0.0
1.26 (1.13-1.40)
H4C
2169
0.0
100.0
0.93 (0.84-1.03)
H4N
2159
100.0
0.0
1.41 (1.26-1.57)
Median days from surgery to AC (IQR)
5924
48 (37-62)
SMD, %
Rurality
Nonmetro
3522
49.7
50.3
−0.8
Metro
8326
50.1
49.9
0.8
Facility location
New England
591
49.7
50.3
0.2
Middle Atlantic
1586
50.3
49.7
−0.4
South Atlantic
2877
50.1
49.9
−0.3
East North Central
2188
50.0
50.0
0.1
East South Central
1172
49.5
50.5
0.7
West North Central
1175
50.1
49.9
−0.2
West South Central
728
50.0
50.0
0.0
Mountain
392
50.8
49.2
−0.6
Pacific
1139
49.7
50.3
0.4
Year of diagnosis
2004
606
49.7
50.3
0.3
2005
737
49.5
50.5
0.5
2006
711
49.8
50.2
0.2
2007
661
49.9
50.1
0.1
2008
727
49.7
50.3
0.4
2009
640
51.4
48.6
−1.3
2010
1056
50.5
49.5
−0.6
2011
987
49.4
50.6
0.7
2012
1008
50.2
49.8
−0.2
2013
951
50.5
49.5
−0.6
2014
937
50.8
49.2
−0.9
2015
941
49.1
50.9
1.1
2016
936
50.4
49.6
−0.5
2017
950
49.2
50.8
1.0
Sex
Female
5163
50.1
49.9
0.4
Male
6685
49.9
50.1
−0.4
Mean age at diagnosis (SD)
11,848
66.7 (9.7)
67.1 (8.1)
4.5
Race and ethnicity
Non-Hispanic White
9516
50.0
50.0
−0.3
Non-Hispanic Black
1080
49.8
50.2
0.2
Hispanic
246
49.2
50.8
1.5
Asian and Pacific Islander
232
49.1
50.9
0.5
American Indian/Alaska Native or other
774
50.3
49.7
−0.3
Income quartile
Lowest
3081
50.0
50.0
0.0
2
3387
49.8
50.2
0.4
3
2852
49.8
50.2
0.6
Highest
2528
50.5
49.5
−1.1
Education quartiles
Lowest
2818
49.7
50.3
0.6
2
3493
50.1
49.9
−0.2
3
3302
49.8
50.2
0.6
Highest
2235
50.6
49.4
−1.2
Insurance status
Medicare or private
10,648
50.0
50.0
0.6
Medicaid or uninsured
1033
50.7
49.3
−0.9
Other
167
48.5
51.5
0.7
Charlson-Deyo score
0
5614
50.4
49.6
−1.6
1
4175
49.6
50.4
1.2
2
1532
50.1
49.9
−0.1
≥3
527
48.6
51.4
1.2
Tumor size, cm
≤1
238
49.6
50.4
0.2
>1 to 2
1638
49.6
50.4
0.6
>2 to 3
2261
50.2
49.8
−0.5
>3 to 5
3266
50.4
49.6
−1.1
>5 to 7
2690
49.8
50.2
0.4
>7
1755
49.6
50.4
0.7
Pathological stage
IIA
4577
49.8
50.2
0.5
IIB
5270
49.8
50.2
0.7
IIIA (N0-N1)
2001
50.9
49.1
−1.7
Nodal status
N0
5770
49.4
50.6
2.3
N1
6078
50.6
49.4
−2.3
Grade
Well differentiated
850
48.2
51.8
1.07
Moderately differentiated
5074
49.8
50.2
0.3
Poorly differentiated
5611
50.4
49.6
−0.86
Undifferentiated
313
49.8
50.2
0.06
Histology
Adenocarcinoma
6326
50.6
49.4
−2.4
Squamous
4472
49.2
50.8
2.6
Large cell
355
51.8
48.2
−1.3
Carcinoid
136
47.1
52.9
1.3
Other
559
49.9
50.1
0.1
Facility program
Community
348
49.7
50.3
0.11
Comprehensive
4885
50.5
49.5
−0.91
Academic
4491
49.4
50.6
1.08
Integrated
2124
50.2
49.8
−0.24
Care fragmentation
Single facility
10,107
50.0
50.0
0.08
Multiple facilities
1741
49.9
50.1
−0.08
Extent of resection
Wedge
561
49.7
50.3
0.13
Segmentectomy
163
46.6
53.4
0.87
Lobectomy
9505
50.0
50.0
0.02
Pneumonectomy
1619
50.5
49.5
−0.4
Surgical margin status
R1, R2, or unspecified residual
1236
51.3
48.7
−0.96
R0
10,612
49.8
50.2
0.96
Median lymph nodes sampled, n (IQR)
11,848
10 (6-16)
10 (6-16)
−0.83
Model 5 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled. AC, Adjuvant chemotherapy; aHR, adjusted hazard ratio; CI, confidence interval; IQR, interquartile range; SMD, standardized mean difference; SD, standard deviation.
∗ AC at any facility
† Data are presented as n with percentages except where otherwise noted.
‡ Subgroups denotation uses the following pattern: HVC (H) versus LVC (L)/travel distance quartile/received AC (C) versus did not receive AC (N). Please see Table E2 for detailed definitions.
Figure 4Kaplan–Meier curves and SEs for the stage II to IIIA (N0-N1) propensity score-matched cohort; comparison of overall survival of patients who traveled long distances to high-volume centers and did not receive adjuvant chemotherapy (H4N) with patients who traveled short distances to low-volume centers and received adjuvant chemotherapy (L1C). CI, Confidence interval; SE, survival estimate.
Additional analyses were performed before and after propensity score matching to evaluate the association of AC with OS when travel distance and surgical volume were included as continuous variables (model 4, Table E3; model 6, Table E4). Also, we performed a conditional survival analysis to address potential confounding with perioperative mortality by excluding those with 90-day postoperative mortality and determined that the inference was unchanged with worse OS for H4N patients (aHR, 1.20 [95% CI, 1.07-1.34] vs L1C patients; model 7; Table E5).
Kaplan–Meier survival estimates showed worse OS for patients in the H4N subgroup compared with the L1C subgroup (Figure 4). Kaplan–Meier survival curves were also generated to evaluate bivariate differences in OS for the H4N subgroup and the L1C subgroup at stage I as a control (median OS, 90.1 vs 97.3 months; P = .09; Figure E2).
Discussion
The trend toward regionalization of complex surgical procedures has resulted in patients traveling further for treatment than ever before.
Whereas surgery is a discrete event for which travel is feasible for many patients, treatment over a longer period with chemotherapy or immunotherapy is more difficult if the patient is remote from the treating center.
The recent broadening of the role for adjuvant treatment in patients with NSCLC is likely to further accentuate this issue for rural patients. In this study, patients with stage II to IIIA (N0-N1) NSCLC who traveled long distances for surgical treatment were less likely to receive AC than patients who traveled short distances. Although perioperative outcomes are improved by travel to HVCs in our study and others, the protective effect of surgery at HVCs is mitigated when care fragmentation occurs because of distance. We found that patients with NSCLC who traveled long distances for surgical treatment at HVCs and did not receive AC had worse survival compared with patients who traveled short distances to LVCs but received AC.
A large body of research spanning nearly 5 decades has identified an association of higher surgical volume for complex procedures with improved patient outcomes.
The definition of a complex surgical procedure in the context of the volume–outcome relationship has not been definitively established, but surgical treatment of the lungs, heart, esophagus, pancreas, hepatobiliary system, and rectum are extensively reported in this literature.
Wasif N, Etzioni D, Habermann EB, Mathur A, Pockaj BA, Gray RJ, et al. Racial and socioeconomic differences in the use of high-volume commission on cancer-accredited hospitals for cancer surgery in the United States.
These data suggest that the relationship is widely generalizable to operations across organ systems that are characterized by increased patient risk, and that risk can be ameliorated by increased surgeon and institutional experience through operative volume.
noted: volume is a “crude indicator of the quality of care.” The volume–outcome relationship might be a surrogate for differences in the quality of care at multiple levels related to surgical management, including preoperative patient selection and prehabilitation, intraoperative expertise, perioperative monitoring and intervention leading to fewer “failures to rescue” and increased rates of guideline-concordant care.
Although patients treated at HVCs have been shown to have improved perioperative mortality, fewer complications, and shorter lengths of stay; some studies have called into question the validity of the volume–outcome relationship, and few studies have validated proposed thresholds or minimums for surgical volume.
Risk-adjusted mortality rates as a quality proxy outperform volume in surgical oncology-a new perspective on hospital centralization using national population-based data.
Using G-computation to estimate the effect of regionalization of surgical services on the absolute reduction in the occurrence of adverse patient outcomes.
Agency for Healthcare Research and Quality Safety in numbers: hospital performance on Leapfrog's surgical volume standard based on results of the 2019 Leapfrog Hospital Survey.
showed that the number of hospitals that perform lung resections in a US state decreased from 49 to 31 between 2005 and 2015, while the proportion of patients treated at HVCs, as well as patient travel distance, increased.
Intuitively, increased distance might be expected to decrease care access. However, the actual relationship between travel distance and patient outcomes has not been definitively established.
However, previous studies have also reported an association of longer travel distance with decreased access to chemotherapy among patients with colorectal cancer, as well as decreased access to surveillance screening after lung resection.
Association between geographic access to cancer care, insurance, and receipt of chemotherapy: geographic distribution of oncologists and travel distance.
reported that patients are more willing to travel long distances for surgery than for recurring therapies. In our study increasing travel distance was associated with a decreased likelihood of receipt of AC for patients with surgically treated NSCLC at HVCs or LVCs. This raised the question of whether there might be an association with worse OS for certain populations of cancer patients who travel long distances, even to HVCs, but then fail to receive indicated AC. Upon investigation, OS was worse for patients who traveled long distances to HVCs for surgical resection but did not receive AC with stage II to IIIA (N0-N1) disease.
However, in patients with early stage NSCLC, AC is not typically indicated and was only administered in approximately 2.1% of 60,378 stage IA cases in our study cohort. We found no association with worse OS for those with stage I NSCLC who traveled long distances to HVCs (Figure E2), further supporting the hypothesis that the survival differences in the stage II to IIIA (N0-N1) cohort were affected by care fragmentation. Although this retrospective study cannot show a causal relationship, we can reasonably infer that reduced access and uptake of indicated postoperative oncological services might adversely influence patient outcomes, including OS. It is also possible that survival disparities related to nonreceipt of AC might worsen in the future as systemic therapies improve and are extended to a larger group of patients with earlier-stage disease.
The benefits of surgery at HVCs for complex procedures have garnered widespread support for regionalization of surgical care, and the resulting trend toward regionalization is likely irreversible.
Agency for Healthcare Research and Quality Safety in numbers: hospital performance on Leapfrog's surgical volume standard based on results of the 2019 Leapfrog Hospital Survey.
However, it is important to recognize that regionalization does not occur in a vacuum and attention must be paid to unintended consequences of this approach. Thus, efforts should focus on maintaining care throughout the treatment continuum and improving communication between patients, local treatment teams, and regional HVCs to mitigate the effect of care fragmentation on treatment.
Maintaining local access to other services is likely dependent on the locoregional health care system's ability to continue effectively coordinating care within their catchment area, and some travel for treatment likely will remain necessary for many patients. Indeed, Rhodin and colleagues
reported that successful multi-institutional coordination to deliver neoadjuvant therapy and surgery for esophageal cancer was not associated with worse OS. Theoretically, no system is better poised to serve this role for patients with cancer than a National Cancer Institute-designated facility that is also a regional HVC.
In fact, the original hospital systems that publicly took the “volume pledge” were established regional systems with extensive resources and vast experience.
These systems are characterized by the ability to conduct rigorous outcomes research and quality improvement to identify barriers to care and implement solutions.
Evaluating breast cancer care coordination at a rural National Cancer Institute Comprehensive Cancer Center using network analysis and geospatial methods.
Cancer Epidemiol Biomarkers Prev.2019; 28: 455-461
However, our study shows that over-reliance on these centers to provide all care might result in important quality gaps for more rural patients, particularly with respect to adjuvant treatment.
This study has several limitations, which should be considered when interpreting the findings. First, although the NCDB captures approximately 70% of total cancers in the United States, and more specifically, 82% of cancers of the lung and bronchus, there are reporting gaps.
Comparison of commission on cancer-approved and -nonapproved hospitals in the United States: implications for studies that use the National Cancer Data Base.
Travel distance, as reported in the NCDB, should not be applied directly to policy. Additionally, we were unable to determine in the NCDB if patients who traveled long distances bypassed a nearby HVC to seek care at another facility, although this would be expected to be few patients.
Conclusions
Patients with stage II to IIIA (N0-N1) NSCLC who travel long distances for surgical treatment are less likely to receive AC than patients who travel short distances for surgical treatment. Furthermore, patients with NSCLC who traveled long distances for surgical treatment at HVCs and did not receive AC had worse OS compared with patients who traveled short distances to LVCs but received AC. Understanding the reasons underlying this lack of receipt of AC will provide actionable opportunities to improve care coordination and treatment outcomes, thereby maximizing the benefit of travel to HVCs for surgical treatment.
Conflict of Interest Statement
The authors reported no conflicts of interest.
The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.
The data used are derived from a deidentified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology used, or the conclusions drawn from these data by the investigators.
Appendix E1
Table E1Histological categories defined in the International Classification of Disease for Oncology, third edition
Table E3Population characteristics of travel distance, hospital surgical volume, and receipt of adjuvant chemotherapy for stage II to IIIA (N0-N1) patients and Cox proportional hazards model 4
Model 4 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
<.001
Yes
17,195
0.0
100.0
0.68 (0.66-0.70)
No
17,463
100.0
0.0
Reference
Median travel distance to surgical treatment (IQR)
Model 4 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Model 4 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Model 4 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Data are presented as n with percentages except where otherwise noted. HR, Hazard ratio; CI, confidence interval; IQR, interquartile range.
∗ Adjuvant chemotherapy at any facility.
† Model 4 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Table E4Population characteristics of travel distance, hospital surgical volume, and receipt of adjuvant chemotherapy for stage II to IIIA (N0-N1) propensity score-matched cohort and Cox proportional hazards model 6
Model 6 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
<.001
Yes
5924
0.0
100.0
0.67 (0.64-0.71)
No
5924
100.0
0.0
Reference
SMD (%)
Median travel distance to surgical treatment (IQR)
Model 6 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Model 6 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Model 6 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
† Model 6 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled.
Subgroups denotation uses the following pattern: HVC (H) versus LVC (L)/travel distance quartile/received AC (C) versus did not receive AC (N). Please see Table E2 for detailed definitions.
<.001
L1C
1583
0.0
100.0
Reference
L1N
1395
100.0
0.0
1.16 (1.06-1.27)
L4C
924
0.0
100.0
0.90 (0.80-1.02)
L4N
792
100.0
0.0
1.16 (1.02-1.31)
H1C
1213
0.0
100.0
0.86 (0.77-0.96)
H1N
1097
100.0
0.0
1.06 (0.95-1.19)
H4C
2149
0.0
100.0
0.93 (0.84-1.03)
H4N
1914
100.0
0.0
1.20 (1.07-1.34)
Model 7 adjusted for age, race, sex, income, education, insurance, comorbidities, rurality, region, year of diagnosis, stage, nodal status, tumor size, histology, grade, care fragmentation, facility type, extent of resection, margin status, and number of lymph nodes sampled. AC, Adjuvant chemotherapy; aHR, adjusted hazard ratio; CI, confidence interval.
∗ AC at any facility.
† Subgroups denotation uses the following pattern: HVC (H) versus LVC (L)/travel distance quartile/received AC (C) versus did not receive AC (N). Please see Table E2 for detailed definitions.
Figure E1Trends of travel distance to surgical treatment and rate of receipt of adjuvant chemotherapy at any facility according to year of diagnosis for resected stage II to IIIA (N0-N1) non–small cell lung cancer (NSCLC). Trend P < .001.
Figure E2Kaplan–Meier curves and survival estimates (SEs) for patients with stage I disease; comparison of overall survival of patients who traveled long distances to high-volume centers and did not receive adjuvant chemotherapy (H4N) with patients who traveled short distances to low-volume centers and received adjuvant chemotherapy (L1C). CI, Confidence interval.
Association between geographic access to cancer care, insurance, and receipt of chemotherapy: geographic distribution of oncologists and travel distance.
Wasif N, Etzioni D, Habermann EB, Mathur A, Pockaj BA, Gray RJ, et al. Racial and socioeconomic differences in the use of high-volume commission on cancer-accredited hospitals for cancer surgery in the United States.
Risk-adjusted mortality rates as a quality proxy outperform volume in surgical oncology-a new perspective on hospital centralization using national population-based data.
Using G-computation to estimate the effect of regionalization of surgical services on the absolute reduction in the occurrence of adverse patient outcomes.
Evaluating breast cancer care coordination at a rural National Cancer Institute Comprehensive Cancer Center using network analysis and geospatial methods.
Cancer Epidemiol Biomarkers Prev.2019; 28: 455-461
Comparison of commission on cancer-approved and -nonapproved hospitals in the United States: implications for studies that use the National Cancer Data Base.
Research reported in this publication was supported by the National Institute on Minority Health and Health Disparities of the National Institutes of Health under award number T37MD014248 (C.D.L.), the National Cancer Institute under award number K07 CA216330 (D.D.O.), the National Heart, Lung, and Blood Institute under award number R01HL145478 (A.B.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Read at the 48th Annual Meeting of the Western Thoracic Surgical Association, Koloa, Hawaii, June 22-25, 2022.
00414-4&id=fx1.jpg){kind=link}
00414-4&id=fx2.jpg){kind=link}
00414-4&id=gr1.jpg){kind=link}
00414-4&id=gr2.jpg){kind=link}
00414-4&id=gr3.jpg){kind=link}
00414-4&id=gr4.jpg){kind=link}
00414-4&id=fx3.jpg){kind=link}
00414-4&id=fx4.jpg){kind=link}