Abstract
Background
Incomplete lung cancer resection connotes poor prognosis, the incidence varies with patient demographic, clinical, and institutional factors. We sought to develop a valid, survival-impactful facility-based surgical quality metric which adjusts for related patient demographic and clinical characteristics.
Methods
Facilities performing resections for patients diagnosed with stage I-IIIA non-small-cell lung cancer in National Cancer Data Base between 2004-2011 were identified. Multivariate logistic regression modeling was used to estimate the expected number of margin-positive cases by adjusting for patient risk-mix and calculate the observed: expected (O/E) ratio for each facility. Facilities were categorized as outperformers (O/E ratio<1, p<.05), non-outliers (p>.05), and underperformers (O/E ratio>1, p<.05); and their characteristics across performance categories were compared by chi-square tests. Multivariate Cox proportional hazard analyses were conducted, adjusting for patient demographic and clinical characteristics.
Results
A total of 96,324 patients underwent surgery at 809 facilities. The overall observed margin-positive rate was 4.4%. Sixty-one facilities (8%) were outperformers, 644(80%) were non-outliers, and 104(13%) were underperformers. One-third (36%) of National Cancer Institute-designated facilities, 13% Academic-Comprehensive-Cancer-Programs, 5% Comprehensive-Community-Cancer-Programs, and 13% ‘other’ facilities achieved outperforming status but no Community-Cancer-Programs did. Interestingly, 9% of National Cancer Institute-designated facilities and 11% of Academic-Comprehensive-Cancer-program facilities were underperformers. Adjusting for patient demographic and clinical characteristics, outperformers had a 5-year all-cause hazard ratio of 0.88 (95%CI: 0.85–0.91, p<.0001) compared to non-outliers; and 0.80 (95%CI: 0.77–0.84, p<.0001) compared to underperformers.
Conclusions
Facility performance in lung cancer surgery can be captured by the risk-adjusted margin- positivity rate, potentially providing a valid quality improvement metric.
Surgical resection is the most important curative treatment modality for early-stage non-small-cell lung cancer (NSCLC). However, resection with positive surgical margins, whether microscopic or macroscopic , is associated with significantly inferior survival [1-3]. The value of adjuvant therapy in this setting has been questioned [1,4,5]. Nevertheless, adjuvant therapy is insufficient to completely correct the excessive death risk associated with incomplete resection [3,5].
Patient demographic and clinical factors, as well as institutional and surgeon characteristics, have been associated with disparate survival in lung cancer patients receiving surgery [6–14]. Studies have suggested that institutions and surgeons with higher volumes of high-risk surgical cases have better short-term outcomes than those with lower volumes [7–9,14]. However, these studies have been criticized for failing to account for differing patient risk-mix [15,16].
Given evidence of significant patient and institutional-level disparities in the incidence of incomplete resection for NSCLC [3], we sought to develop a valid, survival-impactful facility-based surgical quality metric which adjusts for related patient demographic and clinical characteristics.
PATIENTS AND METHODS
Data source
The National Cancer Data Base (NCDB), jointly sponsored by the American College of Surgeons and the American Cancer Society, is a hospital-based cancer registry collecting cancer cases from more than 1500 American College of Surgeons Commission on Cancer (CoC)-accredited facilities. The NCDB captures approximately 70 percent of newly diagnosed cancer cases in the US and has been used to analyze cancer treatment and outcomes [17]. This study was granted exempt review by the Institutional Review Boards of the Morehouse School of Medicine, Atlanta, GA and Baptist Cancer Center, Memphis, TN.
Study population
We selected patients aged 18 to 90 years, diagnosed with a first primary invasive American Joint Committee on Cancer stage I-IIIA NSCLC (International Classification of Diseases for Oncology, 3rd Edition, site code: C34.0-C34.3, C34.8, C34.9), who underwent cancer-directed surgery within six months of diagnosis in the reporting facilities between 2004 and 2011. We excluded patients with missing information on sex, diagnosis date, surgery date, surgical margin status, tumor size, last contact date, and those who received neoadjuvant therapy (Figure 1). We also excluded all facilities that performed fewer than 25 NSCLC resections and those that reported zero margin-positive cases, because of concerns about the quality of data.
Figure 1.
Patient selection schema.
*Non-small cell histology was identified through International Classification of Diseases for Oncology, 3rd version (ICD-O-3) histology codes: 8010-8040, 8050-8076, 8140, 8143, 8211, 8230-8231, 8246, 8250-8260, 8310, 8320, 8323, 8430, 8470-8490, 8550-8573, 8980, 8981.
†Cancer-directed surgery was identified through site-specific surgical codes (21, 22, 30 –80), including sub-lobectomy, lobectomy, bi-lobectomy, and pneumonectomy.
Outcomes and Covariates
The primary outcome of this study is the risk-adjusted margin-positivity (RAMP) ratio for each facility in which NSCLC resection was performed from 2004 to 2011. The RAMP ratio is defined as the ratio of observed surgical margin-positivity cases versus expected margin-positivity cases after adjustment for patient demographic and clinical characteristics (O/E ratio). Observed surgical margin-positive cases were captured at each facility from the final pathology report as microscopic, macroscopic, or margin-positive but unspecified, after resection of the primary tumor. Expected surgical margin-positivity cases were predicted based on patient demographic and clinical characteristics in each facility using multivariate logistic regression models. Using previously-specified methodology [18,19], we determined the significance of the O/E ratio by binominal function and used this to define facility performance status (details in supplemental methodology, supplemental Table 1). Facilities were categorized as outperforming (O/E ratio<1, p<0.05), non-outlier (p>=0.05), and underperforming (O/E ratio>1, p<0.05).
The main variables of interest were facility characteristics, including facility type, facility location, payer-mix and surgical-volume. Facilities were accredited by the CoC as ‘Community Cancer Program’, ‘Comprehensive Community Cancer Program’, ‘Academic Comprehensive Cancer Program’, ‘National Cancer Institute (NCI)-designated Cancer Program’ and ‘Other’, which includes ‘Integrated Network Cancer Programs’, ‘Hospital Associate Cancer Programs’ and ‘Free Standing Cancer Center Programs’. We categorized facility location by US census region, and measured payer-mix by the proportion of patients with no insurance or insured with Medicaid, which we then categorized in quartiles.
We measured surgical volume in two ways: total number of cases of cancer-directed surgery for all cancer sites and the proportion of cancer-directed surgery for lung cancer, which we also categorized in quartiles. We ascertained surgeons who performed cancer-directed surgery by their National Provider Identifier, a unique 10-digit identification number issued to health care providers in the United States by the Centers for Medicare and Medicaid Services, which has been collected by the NCDB since 2010. In 2010 and 2011, 24,933 patients underwent cancer-directed surgery, and 3596 (14.4%) had missing surgeon identifier, and thus their surgeon characteristics were categorized as unknown. For both institution and surgeon, surgical-volume was categorized as high if the institutional or surgeon average was greater than the 75th percentile. Institutions or surgeons were categorized as having a high proportion of underserved patients if their average proportion of Medicaid and uninsured patients was greater than the 75th percentile.
Patient demographic characteristics considered in the analysis included age at diagnosis, gender, race/ethnicity, insurance, census region of residence, diagnosis year, and median income level of neighborhood of residence. Median income level of neighborhood was derived from U.S. 2000 Census data and categorized based on national quartiles by zip-code level. Clinical characteristics included the primary tumor site, histology, grade, size, stage and comorbidity. Comorbidity was measured using the Charlson-Deyo comorbidity score through information on pre-existing medical conditions captured in the NCDB by International Classification of Diseases, 9th revision, Clinical Modification diagnosis codes.
Statistical Analyses
We plotted the distribution of O/E ratio of surgical margin-positivity with binominal p-value and used descriptive analysis to summarize patient, physician and facility characteristics, which we stratified by facility performance status, and compared by the Chi-square test, with a significance level at 0.05. The predictive accuracy of the model was measured by the concordance statistic [20] using bootstrapping with 200 re-samples. The concordance statistic was 0.75, indicating that the model had fair discrimination [21]. Five-year unadjusted overall survival rates were estimated by Kaplan-Meier analysis and compared across categories with the log-rank test. Log-rank p-value under 0.05 was considered statistically significant.
We examined the proportional hazard assumption of facility performance status by Kaplan-Meier log(-log) survival curves and found no violation. We conducted univariate and multivariate Cox proportional hazard analyses, adjusting patient demographic and clinical characteristics in multivariate analyses. Since surgical outcome among patients with stage IIIA or squamous histology is associated with risk for margin-positivity, we conducted sensitivity analyses among patients with stage I-II with all histology types and with adenocarcinoma histology only. All statistical analyses were performed using SAS version 9.4 (Cary, NC).
RESULTS
A total of 96,324 patients underwent surgery for stage I-IIIA NSCLC in 809 facilities. The overall margin-positivity rate was 4.4% (Table 1), but differed by stage (Figure 2). Supplemental figure 1 shows the distribution of O/E ratio and binominal p-value. We identified 61 facilities as ‘outperformers’ (mean O/E ratio=0.40, 95% confidence interval [CI]: 0.36–0.45), 644 as ‘non-outliers’ (mean O/E ratio=1.04, 95%CI: 1.00–1.07), and 104 as ‘underperformers’ (mean O/E ratio=2.43, 95%CI: 2.27–2.59).
Table 1.
Patient characteristics by outlier status
| Categories | Total | Underperforming | Non- Outlier | Outperforming | p-value |
|---|---|---|---|---|---|
| Number of patients(%) | 96324 | 11622(12.1) | 65886(68.4) | 18816(19.5) | |
| Number of facilities(%) | 809 | 104(12.9) | 644(79.6) | 61(7.5) | |
| Margin positivity(%) | 4259(4.4) | 1108(9.5) | 2770(4.2) | 381(2) | < 0.0001 |
| Stage Group | |||||
| Stage IA(T1aN0) | 42204(43.8) | 4832(41.6) | 29072(44.1) | 8300(44.1) | < 0.0001 |
| Stage IB(T2aN0) & Stage IIA(T2bN0) | 29433(30.6) | 3645(31.4) | 20104(30.5) | 5684(30.2) | |
| Stage IIA(T1ab-T2aN1) & Stage IIB(T3N0; T2bN1) | 16167(16.8) | 2057(17.7) | 11010(16.7) | 3100(16.5) | |
| Stage IIIA(T1-T3N2; T3N1) | 8520(8.8) | 1088(9.4) | 5700(8.7) | 1732(9.2) | |
| Age, years | |||||
| 18–49 | 4900(5.1) | 621(5.3) | 3320(5) | 959(5.1) | 0.051 |
| 50–64 | 30410(31.6) | 3753(32.3) | 20673(31.4) | 5984(31.8) | |
| 65–74 | 36904(38.3) | 4454(38.3) | 25218(38.3) | 7232(38.4) | |
| 75–90 | 24110(25) | 2794(24) | 16675(25.3) | 4641(24.7) | |
| Sex | |||||
| Male | 47080(48.9) | 5827(50.1) | 32198(48.9) | 9055(48.1) | 0.003 |
| Female | 49244(51.1) | 5795(49.9) | 33688(51.1) | 9761(51.9) | |
| Race/ethnicity | |||||
| Non-Hispanic, White | 75786(78.7) | 9073(78.1) | 52021(79) | 14692(78.1) | < 0.0001 |
| Hispanic | 2041(2.1) | 158(1.4) | 1477(2.2) | 406(2.2) | |
| Black | 7820(8.1) | 966(8.3) | 5183(7.9) | 1671(8.9) | |
| Other | 2412(2.5) | 227(2) | 1594(2.4) | 591(3.1) | |
| Missing | 8265(8.6) | 1198(10.3) | 5611(8.5) | 1456(7.7) | |
| Diagnosis year | |||||
| 2004 | 11068(11.5) | 1297(11.2) | 7707(11.7) | 2064(11) | 0.011 |
| 2005 | 11941(12.4) | 1419(12.2) | 8113(12.3) | 2409(12.8) | |
| 2006 | 12184(12.6) | 1477(12.7) | 8250(12.5) | 2457(13.1) | |
| 2007 | 12110(12.6) | 1456(12.5) | 8343(12.7) | 2311(12.3) | |
| 2008 | 12119(12.6) | 1491(12.8) | 8334(12.6) | 2294(12.2) | |
| 2009 | 11969(12.4) | 1495(12.9) | 8043(12.2) | 2431(12.9) | |
| 2010 | 12148(12.6) | 1442(12.4) | 8313(12.6) | 2393(12.7) | |
| 2011 | 12785(13.3) | 1545(13.3) | 8783(13.3) | 2457(13.1) | |
| Insurance | |||||
| Uninsured | 1795(1.9) | 231(2) | 1180(1.8) | 384(2) | < 0.0001 |
| Medicaid | 3852(4) | 526(4.5) | 2649(4) | 677(3.6) | |
| Younger Medicare | 5385(5.6) | 761(6.5) | 3583(5.4) | 1041(5.5) | |
| Older Medicare | 51606(53.6) | 6167(53.1) | 35263(53.5) | 10176(54.1) | |
| Government | 180(0.2) | 19(0.2) | 133(0.2) | 28(0.1) | |
| Private | 32277(33.5) | 3795(32.7) | 22286(33.8) | 6196(32.9) | |
| Missing | 1229(1.3) | 123(1.1) | 792(1.2) | 314(1.7) | |
| Median income-quartile 2000 | |||||
| <$30,000 | 12964(13.5) | 1871(16.1) | 8397(12.7) | 2696(14.3) | < 0.0001 |
| $30,000–$34,999 | 18041(18.7) | 2667(22.9) | 11702(17.8) | 3672(19.5) | |
| $35,000–$45,999 | 26601(27.6) | 3319(28.6) | 18446(28) | 4836(25.7) | |
| $46,000+ | 33548(34.8) | 3174(27.3) | 23911(36.3) | 6463(34.3) | |
| Missing | 5170(5.4) | 591(5.1) | 3430(5.2) | 1149(6.1) | |
| Comorbidity | |||||
| 0 | 45055(46.8) | 4889(42.1) | 30693(46.6) | 9473(50.3) | < 0.0001 |
| 1 | 34882(36.2) | 4501(38.7) | 23785(36.1) | 6596(35.1) | |
| 2+ | 16387(17) | 2232(19.2) | 11408(17.3) | 2747(14.6) | |
| Histology | |||||
| Not Otherwise Specified | 281(0.3) | 34(0.3) | 193(0.3) | 54(0.3) | < 0.0001 |
| Large Cell | 4437(4.6) | 604(5.2) | 3015(4.6) | 818(4.3) | |
| Squamous | 28819(29.9) | 3674(31.6) | 19812(30.1) | 5333(28.3) | |
| Other | 4981(5.2) | 567(4.9) | 3359(5.1) | 1055(5.6) | |
| Adeno | 57806(60) | 6743(58) | 39507(60) | 11556(61.4) | |
| Tumor grade | |||||
| well/moderately differentiated | 56437(58.6) | 6664(57.3) | 38691(58.7) | 11082(58.9) | < 0.0001 |
| poorly/undifferentiated | 35678(37) | 4527(39) | 24348(37) | 6803(36.2) | |
| Unknown | 4209(4.4) | 431(3.7) | 2847(4.3) | 931(4.9) | |
| Tumor size | |||||
| ≤3cm | 60312(62.6) | 7088(61) | 41308(62.7) | 11916(63.3) | 0.003 |
| >3cm-≤5cm | 23502(24.4) | 2977(25.6) | 16061(24.4) | 4464(23.7) | |
| >5cm | 12185(12.7) | 1519(13.1) | 8299(12.6) | 2367(12.6) | |
| Unknown | 325(0.3) | 38(0.3) | 218(0.3) | 69(0.4) | |
| Location | |||||
| Rural | 18518(19.2) | 2787(24) | 11566(17.6) | 4165(22.1) | < 0.0001 |
| Urban | 71973(74.7) | 8210(70.6) | 50478(76.6) | 13285(70.6) | |
| Unknown | 5833(6.1) | 625(5.4) | 3842(5.8) | 1366(7.3) | |
| Census region | |||||
| Northeast | 18875(19.6) | 2360(20.3) | 12544(19) | 3971(21.1) | < 0.0001 |
| Midwest | 26469(27.5) | 2953(25.4) | 19108(29) | 4408(23.4) | |
| South | 39535(41) | 5186(44.6) | 25542(38.8) | 8807(46.8) | |
| West | 11319(11.8) | 1114(9.6) | 8626(13.1) | 1579(8.4) | |
| Missing | 126(0.1) | 9(0.1) | 66(0.1) | 51(0.3) | |
| Primary site | |||||
| C340-Main bronchus | 560(0.6) | 75(0.6) | 372(0.6) | 113(0.6) | < 0.0001 |
| C341-upper lobe | 57737(59.9) | 6930(59.6) | 39393(59.8) | 11414(60.7) | |
| C342-Middle lobe | 4717(4.9) | 552(4.7) | 3236(4.9) | 929(4.9) | |
| C343-Lower lobe | 30414(31.6) | 3709(31.9) | 20815(31.6) | 5890(31.3) | |
| C348-Overlapping lesion | 1514(1.6) | 206(1.8) | 1047(1.6) | 261(1.4) | |
| C349-Lung, not specified | 1382(1.4) | 150(1.3) | 1023(1.6) | 209(1.1) | |
| Pathologic T-category | |||||
| T1 | 48689(50.5) | 5603(48.2) | 33575(51) | 9511(50.5) | < 0.0001 |
| T2 | 41162(42.7) | 5160(44.4) | 27995(42.5) | 8007(42.6) | |
| T3 | 6473(6.7) | 859(7.4) | 4316(6.6) | 1298(6.9) | |
| Pathologic N-category | |||||
| N0 | 75908(78.8) | 9053(77.9) | 52018(79) | 14837(78.9) | < 0.0001 |
| N1 | 13320(13.8) | 1664(14.3) | 9143(13.9) | 2513(13.4) | |
| N2 | 7096(7.4) | 905(7.8) | 4725(7.2) | 1466(7.8) | |
| Extent of resection | |||||
| Sublobar | 12425(12.9) | 1454(12.5) | 8405(12.8) | 2566(13.6) | < 0.0001 |
| Lobe/bilobectomy | 78865(81.9) | 9537(82.1) | 54167(82.2) | 15161(80.6) | |
| Pneumonectomy | 5034(5.2) | 631(5.4) | 3314(5) | 1089(5.8) | |
Figure 2.
Margin-positive rate by stage.
Patient Characteristics
Twenty percent of patients were treated at outperforming facilities, 68% at non-outliers, and 12.% at underperforming facilities (Table 1). The observed surgical margin positivity rate was highest at underperforming facilities: outperforming, 2%; non-outliers, 4.2%; underperforming, 9.5% (p<.0001). Overall, the median age of the study cohort was 68 (interquartile range 14) years. Most patients (48.6%) were treated at facilities designated as Comprehensive Community Cancer Programs. Majority were non-Hispanic whites (78.7%), Medicare insured (59.2%), diagnosed with adenocarcinoma (60%), had well-differentiated tumors (58.6%), tumor size ≤3cm (62.6%), lobectomy (81.9%), and resided in urban areas (74.7%).
Compared to patients treated at underperforming facilities, patients treated at outperforming facilities were more likely to be female (51.9% vs. 49.9%), reside in high-income neighborhoods (34.3% vs.27.3%), have no comorbidity (50.3% vs. 42.1%), and have adenocarcinoma (61.4% vs. 58%), well/moderately-differentiated tumors (58.9% vs. 57.3%), smaller tumors, and earlier stage disease (Table 1). Patients at underperforming institutions were more likely to have squamous cell lung cancer (31.6% vs.28.3%).
Facility Characteristics
Approximately 8% of institutions were outperformers while 13% were underperformers. Institutional performance status was strongly associated with facility characteristics (Table 2). Twelve of 33 (36.4%) NCI-designated programs and 22 of 167 (13.2%) Academic Comprehensive Cancer programs, 22 of 475 (4.6%) Comprehensive Community Cancer Programs, and 5 of 40 (12.5%) ‘other’ facilities achieved outperforming status, but no Community Cancer Program did. NCI- or Academic-Comprehensive Cancer Program-designation did not guarantee performance: 9.1% of NCI-designated, and 10.8% of Academic-Comprehensive Cancer Programs were underperformers.
Table 2.
Facility characteristics by performance status
| Categories | Total | Underperforming | Non- Outlier | Outperforming | p-value |
|---|---|---|---|---|---|
| Number of facilities(%) | 809 | 104(12.9) | 644(79.6) | 61(7.5) | |
| N(column%) | N(row %) | N(row %) | N(row %) | ||
| Census region | |||||
| Northeast | 162(20) | 22(13.6) | 129(79.6) | 11(6.8) | 0.28 |
| Midwest | 243(30) | 30(12.3) | 198(81.5) | 15(6.2) | |
| South | 281(34.7) | 40(14.2) | 212(75.4) | 29(10.3) | |
| West | 123(15.2) | 12(9.8) | 105(85.4) | 6(4.9) | |
| Facility type | |||||
| Community Cancer Program | 94(11.6) | 12(12.8) | 82(87.2) | 0(0) | < 0.0001 |
| Comprehensive Community Cancer Program | 475(58.7) | 69(14.5) | 384(80.8) | 22(4.6) | |
| Academic Comprehensive Cancer Program | 167(20.6) | 18(10.8) | 127(76) | 22(13.2) | |
| NCI Program/Network | 33(4.1) | 3(9.1) | 18(54.5) | 12(36.4) | |
| Other | 40(4.9) | 2(5) | 33(82.5) | 5(12.5) | |
| Proportion of Medicaid/Uninsured patients | |||||
| Q1(low) | 152(18.8) | 17(11.2) | 128(84.2) | 7(4.6) | 0.11 |
| Q2 | 173(21.4) | 15(8.7) | 147(85) | 11(6.4) | |
| Q3 | 192(23.7) | 30(15.6) | 148(77.1) | 14(7.3) | |
| Q4(high) | 292(36.1) | 42(14.4) | 221(75.7) | 29(9.9) | |
| Proportion of surgery which were lung cancer resection | |||||
| Q1(low) | 50(6.2) | 5(10) | 44(88) | 1(2) | 0.09 |
| Q2 | 227(28.1) | 30(13.2) | 188(82.8) | 9(4) | |
| Q3 | 248(30.7) | 29(11.7) | 195(78.6) | 24(9.7) | |
| Q4(high) | 284(35.1) | 40(14.1) | 217(76.4) | 27(9.5) | |
| Total cancer surgical volume | |||||
| Q1(low) | 24(3) | 3(12.5) | 21(87.5) | 0(0) | < 0.0001 |
| Q2 | 155(19.2) | 20(12.9) | 135(87.1) | 0(0) | |
| Q3 | 264(32.6) | 34(12.9) | 224(84.8) | 6(2.3) | |
| Q4(high) | 366(45.2) | 47(12.8) | 264(72.1) | 55(15) | |
Institutional performance status was also associated with surgical volume. The average number of patients treated at outperforming facilities was almost three times that of non-outliers or underperforming facilities (outperforming, 308; non-outliers, 102; underperforming, 112, t-test p<.001). However, greater surgical volume did not guarantee institutional performance. Among 630 facilities which had a total cancer surgical volume larger than the median, 61 (9.7%) were outperformers while 81 (12.9%) were underperformers. Facilities with a total cancer surgical volume less than the median were not more likely to be underperformers. However, none of them were outperformers. Similarly, specialization in lung cancer surgery did not assure outperformance. Among 532 facilities with a proportion of lung cancer surgery larger than the median, 51 (9.6%) were outperformers and 69(13%) were underperformers. Of those facilities with a proportion of lung cancer surgery less than the median, 10 (3.6%) were outperformers and 35 (12.6%) were underperformers. Outperforming facilities were as likely as underperforming institutions to have a high proportion of uninsured/Medicaid-insured patients (47.5% vs. 40.4%, p=0.59).
Physician Characteristics
A total of 2011 unique surgeons performed cancer-directed surgery for patients diagnosed between 2010 and 2011 (Table 3). The average annual surgical volume and 75th percentile surgical volume were 12.8 (standard deviation, 14.9) and 16.5, respectively. The case-volume among surgeons in outperforming facilities was almost twice that of surgeons in underperforming or non-outlier facilities: outperforming, 20.9; non-outlier, 11.6; underperforming, 12.8 (t-test p<.001). Surgeons in outperforming facilities were more likely to be at the highest case-volume quartiles (52.2%), than those at underperforming (37.2%) and non-outlier (34.4%) facilities.
Table 3.
Physician characteristics by facility performance status (patients treated between 2010 and 2011)
| Categories | Total | Underperforming | Non- Outlier | Outperforming | p-value |
|---|---|---|---|---|---|
| Number of Physicians(%) | 2011 | 250(12.4) | 1537(76.4) | 224(11.1) | |
| N(column%) | N(column%) | N(column%) | N(column%) | ||
| Census region | |||||
| Northeast | 357(17.8) | 42(16.8) | 269(17.5) | 46(20.5) | 0.052 |
| Midwest | 571(28.4) | 66(26.4) | 449(29.2) | 56(25) | |
| South | 811(40.3) | 117(46.8) | 595(38.7) | 99(44.2) | |
| West | 272(13.5) | 25(10) | 224(14.6) | 23(10.3) | |
| Facility type | |||||
| Community Cancer Program | 141(7) | 17(6.8) | 124(8.1) | 0(0) | < 0.0001 |
| Comprehensive Community Cancer Program | 1111(55.2) | 165(66) | 866(56.3) | 80(35.7) | |
| Academic Comprehensive Cancer Program | 454(22.6) | 42(16.8) | 349(22.7) | 63(28.1) | |
| NCI Program/Network | 138(6.9) | 12(4.8) | 63(4.1) | 63(28.1) | |
| Other | 167(8.3) | 14(5.6) | 135(8.8) | 18(8) | |
| Proportion of Medicaid/uninsured patients | |||||
| Low(<75th percentile) | 1440(71.6) | 177(70.8) | 1107(72) | 156(69.6) | 0.73 |
| High(≥75th percentile) | 571(28.4) | 73(29.2) | 430(28) | 68(30.4) | |
| Surgical volume | |||||
| Low(<75th percentile) | 1273(63.3) | 157(62.8) | 1009(65.6) | 107(47.8) | < 0.0001 |
| High(≥75th percentile) | 738(36.7) | 93(37.2) | 528(34.4) | 117(52.2) | |
Survival impact
Patients treated at outperforming facilities had better unadjusted 5-year overall survival than those treated at non-outlying and underperforming facilities: 60.1% vs. 56.2% vs. 53.9%, respectively, log-rank p<.0001 (Figure 3). Before adjusting for patient and clinical characteristics, patients treated at outperforming facilities had 12% lower 5-year all-cause mortality risk than patients treated at non-outlier facilities (Hazard Ratio 0.88, 95%CI 0.85–0.91, p<.0001) and 20% lower risk than patients treated at underperforming facilities (Hazard Ratio 0.8, 95%CI 0.77–0.84, p<.0001) (Table 4, supplemental Tables 2a,b). After adjusting for patient demographic and clinical characteristics, patients treated at outperforming facilities had better survival than those treated at non-outlier and underperforming facilities (Hazard Ratio: 0.88 vs. non-outliers; Hazard Ratio: 0.85 vs. underperforming, p<.0001). The results from the sensitivity analyses were similar to those from primary analyses (Supplemental Tables 3a–4b).
Figure 3.
Kaplan-Meier survival curves by facility performance category.
Table 4.
Association between facility performance status and 5 year overall survival
| Performing status | Hazard Ratio(95% CI) | p value |
|---|---|---|
| Unadjusted | ||
| Outperforming | 0.88(0.85–0.91) | <.0001 |
| Non-outlier | 1 | |
| Underperforming | 1.09(1.05–1.13) | <.0001 |
| Outperforming vs. Underperforming(reference) | 0.80(0.77–0.84) | <.0001 |
| Adjusted for demographics* | ||
| Outperforming | 0.88(0.85–0.90) | <.0001 |
| Non-outlier | 1 | |
| Underperforming | 1.07(1.04–1.11) | 0.0001 |
| Outperforming vs. Underperforming(reference) | 0.82(0.79–0.86) | <.0001 |
| Adjusted for demographics & clinical characteristics† | ||
| Outperforming | 0.88(0.86–0.91) | <.0001 |
| Non-outlier | 1 | |
| Underperforming | 1.04(1.00–1.08) | 0.031 |
| Outperforming vs. Underperforming(reference) | 0.85(0.82–0.89) | <.0001 |
includes age, gender, race/ethnicity, insurance, median income level, urban/rural
includes age, gender, race/ethnicity, insurance, median income level, urban/rural, comorbidity, T stage, N stage, tumor size, histology, tumor grade, and tumor location.
COMMENT
In this contemporary nation-wide study, patients treated at facilities with a significantly lower-than-expected rate of margin-positive NSCLC resections had significantly lower long-term mortality risk than patients treated at facilities with equal-to or greater-than-expected incomplete resection rates. The 20% of patients who had surgery at CoC-accredited facilities categorized by their RAMP ratio as ‘outperformers’ had a 12% lower adjusted 5-year all-cause mortality risk than the 68% of patients whose surgery was performed at the facilities classified as ‘non-outliers’, and 15% lower risk than the 12% whose surgery was performed in facilities categorized as ‘underperformers’.
Despite the strong association between CoC facility type and institutional performance status, CoC-designation was not a guarantee against institutional underperformance. No Community Cancer Program outperformed, 5% of Comprehensive Community Cancer Programs,13% of Academic Comprehensive Cancer Programs and 36% of NCI-designated Programs were outperformers. However, 11% of Academic Comprehensive Cancer Programs and 9% of NCI-designated Programs underperformed.
RAMP-based institutional category was strongly associated with institutional surgical volume. All outperforming facilities had above-median total cancer surgical volume, suggesting that practice makes for better outcomes. However, 75% of underperforming facilities also had above-median total cancer surgical volume. A similar observation generally applied to surgeon-level case-volume. This suggests a need to further understand the specific institutional and provider practices that distinguish outperforming from underperforming facilities and surgeons. Categorization by RAMP rate can help identify such institutions for further comparative evaluation. Clearly, high-volume institutions and surgeons to can adopt practices that lead to bad patient outcomes.
Several previous reports have suggested links between institutional surgical volume, operative mortality risk and long-term survival probability of lung cancer patients [7–9,14]. Short-term postoperative risk has also been linked with surgeon case-volume, training and clinical experience [6,8,11–13]. The utility of volume-based structural measures for corrective intervention is limited, and the validity of the volume-outcome relationship has been questioned [15,16]. Volume-based outcome disparities have mostly been related to the probability of surviving the operation itself, indicated by major perioperative complication rates, and failure to rescue patients after such complications [22,23]. The oncologic quality of resection influences the probability of surviving the cancer, its impact is delayed and, consequently, harder to detect [24]. A major challenge for quality improvement is to find reversible institutional and human factors associated with survival disparities, which serve as targets for intervention in underperforming environments.
Process measures of quality are potentially more readily susceptible to corrective intervention, and use for quality improvement [25,26]. Such measures must be clinically relevant, rigorously validated, readily available and standardized before they can serve as meaningful quality metrics. Margin-positive resection is a healthcare outcome disparity, associated with patient, provider and institutional factors [3]. Margin-status has been a mandatory component of the College of American Pathologists’ standard reporting template since at least 2004 [27].
Our retrospective study design precludes direct examination of human and organizational practices causing disparate incomplete resection rates and their survival influence. We cannot directly test important hypotheses such as provider proficiency, the thoroughness of preoperative evaluation, and the quality of intraoperative pathology support. Furthermore, since 14% of cases diagnosed in 2010 and 2011 did not have surgeon information, we might have underestimated surgeon case-volume. Finally, because non-CoC-accredited facilities generally have structural characteristics predictive of worse outcomes than NCDB institutions, [28] we have probably underestimated the value of applying this metric for quality improvement. Individual facilities, irrespective of CoC-accreditation status, can use it for internal quality improvement. Researchers can use it to stratify facilities for further inquiry into the attitudes, beliefs, and practices driving lung cancer process and outcome disparities.
Supplementary Material
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Hancock JG, Rosen JE, Antonicelli A, et al. Impact of adjuvant treatment for microscopic residual disease after non-small cell lung cancer surgery. Ann Thorac Surgery. 2015;99:406–413. doi: 10.1016/j.athoracsur.2014.09.033. [DOI] [PubMed] [Google Scholar]
- 2.Wang EH, Corso CD, Rutter CE, et al. Postoperative Radiation Therapy Is Associated With Improved Overall Survival in Incompletely Resected Stage II and III Non-Small-Cell Lung Cancer. J Clin Oncol. 2015;33:2727–2734. doi: 10.1200/JCO.2015.61.1517. [DOI] [PubMed] [Google Scholar]
- 3.Osarogiagbon RU, Lin CC, Smeltzer MP, Jemal A. Prevalence, Prognostic Implications, and Survival Modulators of Incompletely Resected Non-Small Cell Lung Cancer in the U.S. National Cancer Data Base. J Thorac Oncol. 2016;11:e5–e16. doi: 10.1016/j.jtho.2015.08.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Wind J, Smit EJ, Senan S, Eerenberg JP. Residual disease at the bronchial stump after curative resection for lung cancer. Eur J Cardiothorac Surg. 2007;32:29–34. doi: 10.1016/j.ejcts.2007.04.003. [DOI] [PubMed] [Google Scholar]
- 5.Smeltzer MP, Lin CC, Yu X, Jemal A, Osarogiagbon RU. Survival Impact of Post-Operative Therapy Modalities after Incomplete and Complete Surgical Resection for Non-Small Cell Lung Cancer in the US. J Thorac Oncol. 2015;10(9) suppl 2:S213–4. [Google Scholar]
- 6.Silvestri GA, Handy J, Lackland D, et al. Specialists achieve better outcomes than generalists for lung cancer surgery. CHEST. 1998;114:675–80. doi: 10.1378/chest.114.3.675. [DOI] [PubMed] [Google Scholar]
- 7.Bach PB, Cramer LD, Schrag D, Downey RJ, Gelfand SE, Begg CB. The influence of hospital volume on survival after resection for lung cancer. N Engl J Med. 2001;345:181–8. doi: 10.1056/NEJM200107193450306. [DOI] [PubMed] [Google Scholar]
- 8.Birkmeyer JD, Siewers AE, Finlayson EV, et al. Hospital volume and surgical mortality in the United States. N Engl J Med. 2002;346:1128–37. doi: 10.1056/NEJMsa012337. [DOI] [PubMed] [Google Scholar]
- 9.Birkmeyer JD, Stukel TA, Siewers AE, et al. Surgeon volume and operative mortality in the United States. N Engl J Med. 2003;349:2117–27. doi: 10.1056/NEJMsa035205. [DOI] [PubMed] [Google Scholar]
- 10.Meguid RA, Brooke BS, Chang DC, et al. Are surgical outcomes for lung cancer resections improved at teaching hospitals? Ann Thorac Surg. 2008;85:1015–24. doi: 10.1016/j.athoracsur.2007.09.046. [DOI] [PubMed] [Google Scholar]
- 11.Farjah F, Glum DR, Varghese TK, et al. Surgeon specialty and long-term survival after pulmonary resection for lung cancer. Ann Thorac Surg. 2009;87:995–1004. doi: 10.1016/j.athoracsur.2008.12.030. [DOI] [PubMed] [Google Scholar]
- 12.Schipper PH, Diggs BS, Ungerleider RM, et al. The influence of surgeon specialty on outcomes in general thoracic surgery: a national sample 1996 to 2005. Ann Thorac Surg. 2009;88:1566–72. doi: 10.1016/j.athoracsur.2009.08.055. [DOI] [PubMed] [Google Scholar]
- 13.Ferraris VA, Saha SP, Davenport DL, et al. Thoracic surgery in the real world: does surgical specialty affect outcomes in patients having general thoracic operations? Ann Thorac Surg. 2012;93:1041–7. doi: 10.1016/j.athoracsur.2011.12.061. [DOI] [PubMed] [Google Scholar]
- 14.Lüchtenborg M, Riaz SP, Coupland VH, et al. High procedure volume is strongly associated with improved survival after lung cancer surgery. J Clin Oncol. 2013;31:3141–6. doi: 10.1200/JCO.2013.49.0219. [DOI] [PubMed] [Google Scholar]
- 15.Kozower BD, Stukenborg GJ. The relationship between hospital lung cancer resection volume and patient mortality risk. Ann Surg. 2011;254:1032–7. doi: 10.1097/SLA.0b013e31821d4bdd. [DOI] [PubMed] [Google Scholar]
- 16.French B, Farjah F, Flum DR, Heagerty PJ. A general framework for estimating volume-outcome associations from longitudinal data. Stat Med. 2012;31:366–82. doi: 10.1002/sim.4410. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lerro CC, Robbins AS, Phillips JL, et al. Comparison of cases captured in the national cancer data base with those in population-based central cancer registries. Ann Surg Oncol. 2013;20:1759–65. doi: 10.1245/s10434-013-2901-1. [DOI] [PubMed] [Google Scholar]
- 18.Russell MC, You YN, Hu CY, et al. A novel risk-adjusted nomogram for rectal cancer surgery outcomes. JAMA Surg. 2013 Aug;148(8):769–77. doi: 10.1001/jamasurg.2013.2136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Massarweh NN, Hu CY, You YN, et al. Risk-adjusted pathologic margin positivity rate as a quality indicator in rectal cancer surgery. J Clin Oncol. 2014 Sep 20;32(27):2967–74. doi: 10.1200/JCO.2014.55.5334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Agresti A, editor. Categorical Data Analysis. 3. Wiley; 2013. Building, checking, and applying logistic regression models; p. 224. [Google Scholar]
- 21.Kleinbaum DG, Klein M, editors. Logistic Regression. A Self-Learning Text. Springer; 2010. Assessing discriminatory performance of a binary logistic model; p. 357. [Google Scholar]
- 22.Gonzalez AA, Dimick JB, Bierkmeyer JD, Ghaferi AA. Understanding the Volume-Outcome Effect in Cardiovascular Surgery. The Role of Failure to Rescue. JAMA Surg. 2014;149:119–123. doi: 10.1001/jamasurg.2013.3649. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Grenda TR, Revels SL, Yin J, Birkmeyer JD, Wong SL. Lung Cancer Resection at Hospitals with High vs Low Mortality Rates. JAMA Surg. 2015;150:1034–40. doi: 10.1001/jamasurg.2015.2199. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Bilimoria KY, Bentrem DJ, Feinglass JM, Stewart AK, Winchester DP, Talamonti MS, Ko CY. Directing surgical quality improvement initiatives: comparison of perioperative mortality and long-term survival for cancer surgery. J Clin Oncol. 2008;26:4626–33. doi: 10.1200/JCO.2007.15.6356. [DOI] [PubMed] [Google Scholar]
- 25.Donabedian A. The quality of care. How can it be assessed? JAMA. 1988;260:1743–8. doi: 10.1001/jama.260.12.1743. [DOI] [PubMed] [Google Scholar]
- 26.Farjah F, Detterbeck FC. What is quality, and can we define it in lung cancer? The case for quality improvement. Transl Lung Cancer Res. 2015;4:365–72. doi: 10.3978/j.issn.2218-6751.2015.07.12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gal AA, Marchevsky AM, Travis WD. Updated Protocol for the Examination of Specimens From Patients With Carcinoma of the Lung. Arch Pathol Lab Med. 2003;127:1304–13. doi: 10.5858/2003-127-1304-UPFTEO. [DOI] [PubMed] [Google Scholar]
- 28.Bilimoria KY, Bentrem DJ, Stewart AK, Winchester DP, Ko CY. Comparison of Commission on Cancer-Approved and –Nonapproved Hospitals in the United States: Implications for Studies That Use the National Cancer Data Base. J Clin Oncol. 2009;27:4177–4181. doi: 10.1200/JCO.2008.21.7018. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.