Association Between Surgical Quality Metric Adherence and Overall Survival Among US Veterans With Early-Stage Non–Small Cell Lung Cancer

This cohort study develops and validates a surgical quality score by examining the association between surgical quality metric adherence and the overall survival and recurrence-free survival among US veterans diagnosed with early-stage non–small cell lung cancer.

Key Points

Question

What is the association between surgical quality metric adherence and overall survival and recurrence-free survival among US veterans with early-stage non–small cell lung cancer?

Findings

In this cohort study of 9628 veterans, adherence to intraoperative quality metrics was found to be associated with improved overall survival and recurrence-free survival. These findings were validated in a cohort of 107 674 nonveteran patients using data from the National Cancer Database.

Meaning

The findings of this study suggest that efforts to improve adherence to surgical quality metrics may improve patient outcomes following curative-intent resection of early-stage lung cancer.

Abstract

Importance

Surgical resection remains the preferred treatment for functionally fit patients diagnosed with early-stage non–small cell lung cancer (NSCLC). Process-based intraoperative quality metrics (QMs) are important for optimizing long-term outcomes following curative-intent resection.

Objective

To develop a practical surgical quality score for patients diagnosed with clinical stage I NSCLC who received definitive surgical treatment.

Design, Setting, and Participants

This retrospective cohort study used a uniquely compiled data set of US veterans diagnosed with clinical stage I NSCLC who received definitive surgical treatment from October 2006 through September 2016. The data were analyzed from April 1 to September 1, 2022. Based on contemporary treatment guidelines, 5 surgical QMs were defined: timely surgery, minimally invasive approach, anatomic resection, adequate lymph node sampling, and negative surgical margin. The study developed a surgical quality score reflecting the association between these QMs and overall survival (OS), which was further validated in a cohort of patients using data from the National Cancer Database (NCDB). The study also examined the association between the surgical quality score and recurrence-free survival (RFS).

Exposures

Surgical treatment of early-stage NSCLC.

Main Outcomes and Measures

Overall survival and RFS.

Results

The study included 9628 veterans who underwent surgical treatment between 2006 and 2016. The cohort consisted of 1446 patients who had a mean (SD) age of 67.6 (7.9) years and included 9278 males (96.4%) and 350 females (3.6%). Among the cohort, 5627 individuals (58.4%) identified as being smokers at the time of surgical treatment. The QMs were met as follows: timely surgery (6633 [68.9%]), minimally invasive approach (3986 [41.4%]), lobectomy (6843 [71.1%]) or segmentectomy (532 [5.5%]), adequate lymph node sampling (3278 [34.0%]), and negative surgical margin (9312 [96.7%]). The median (IQR) follow-up time was 6.2 (2.5-11.4) years. An integer-based score (termed the Veterans Affairs Lung Cancer Operative quality [VALCAN-O] score) from 0 (no QMs met) to 13 (all QMs met) was constructed, with higher scores reflecting progressively better risk-adjusted OS. The median (IQR) OS differed substantially between the score categories (score of 0-5 points, 2.6 [1.0-5.7] years of OS; 6-8 points, 4.3 [1.7-8.6] years; 9-11 points, 6.3 [2.6-11.4] years; and 12-13 points, 7.0 [3.0-12.5] years; P < .001). In addition, risk-adjusted RFS improved in a stepwise manner between the score categories (6-8 vs 0-5 points, multivariable-adjusted hazard ratio [aHR], 0.62; 95% CI, 0.48-0.79; P < .001; 12-13 vs 0-5 points, aHR, 0.39; 95% CI, 0.31-0.49; P < .001). In the validation cohort, which included 107 674 nonveteran patients, the score remained associated with OS.

Conclusions and Relevance

The findings of this study suggest that adherence to intraoperative QMs may be associated with improved OS and RFS. Efforts to improve adherence to surgical QMs may improve patient outcomes following curative-intent resection of early-stage lung cancer.

Introduction

Lung cancer remains the leading cause of cancer-related mortality in the United States.¹ For functionally fit patients diagnosed with early-stage disease, the preferred treatment is surgical resection.² Inherent to the modern treatment of lung cancer is the delivery of high-quality, evidence-based care.³ Specifically in the setting of surgical resection, most contemporary lung cancer treatment guidelines recommend several modifiable, process-based surgical quality metrics (QMs) that should be met for all patients diagnosed with early-stage non–small cell lung cancer (NSCLC).^4,5,6 These metrics include timely surgery,^7,8,9 receipt of anatomic resection,^10,11 a minimally invasive approach,^12,13 negative surgical margin,¹⁴ and adequate lymph node sampling.^15,16,17 Prior work from our group has demonstrated that adherence to these surgical QMs is associated with improved outcomes among patients diagnosed with NSCLC who received definitive surgical treatment.¹⁸ Despite this, guideline-concordant surgical care is infrequently achieved in early-stage NSCLC.¹⁹

Several barriers to the standardization of surgical quality have been described. For example, steep learning curves related to video-assisted thoracic surgery have been associated with poor adoption of minimally invasive lung cancer resection^20,21; similarly, various clinician- and patient-specific factors have been associated with inadequate nodal sampling.²² Collectively, these barriers highlight the challenge of disseminating and implementing evidence-based, guideline-concordant practice standards within thoracic surgery.²³

The Veterans Health Administration (VHA) is the largest integrated health care system in the US.²⁴ Although the VHA serves a unique patient population,^25,26 patterns of lung cancer care (including rates of adherence to intraoperative QMs) and outcomes are similar between veterans and the general US population.¹⁰ Standardization of lung cancer treatment quality within the VHA may therefore have a disproportionate impact on early-stage NSCLC outcomes among veterans and provide a road map for implementation in the general US population.

In this study, we sought to develop a pragmatic surgical quality score (termed the Veterans Affairs Lung Cancer Operative quality [VALCAN-O] score) based on the association between previously proposed process-based, intraoperative, modifiable QMs and overall survival (OS). We also examined the association between this score and recurrence-free survival (RFS). We then validated this score using a cohort of patients from the National Cancer Database (NCDB).

Methods

Study Population

In this retrospective cohort study, a unique cohort of veterans diagnosed with early-stage NSCLC who underwent definitive surgical treatment through the VHA from October 2006 through September 2016 was compiled. The cohort was assembled using the VHA Informatics and Computing Infrastructure, which contains several clinical and administrative data sets within the Corporate Data Warehouse, such as Oncology Raw and Veterans Affairs Surgical Quality Improvement Program.²⁷ Veterans diagnosed with NSCLC were identified using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3) codes. Surgical treatments were confirmed using either the International Classification of Diseases, Ninth Revision (ICD-9) or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) procedure or Current Procedural Terminology codes. Inclusion criteria were all adults diagnosed with clinical stage I (≤5 cm, node-negative) NSCLC who underwent surgery (American Joint Committee on Cancer, seventh edition).²⁸ We excluded patients who received neoadjuvant therapy, those who presented with recurrent disease from a previously diagnosed cancer, patients with unavailable date of diagnosis, and those who were missing any of the prespecified QMs. Race was reported as Black, White, or other (ie, any code other than Black or White) as coded in the Corporate Data Warehouse and defined according to the American College of Surgery Facility Oncology Registry Data Standards. Race was included to assess potential factors associated with surgical quality. A dedicated team of researchers performed manual medical record review—augmented with natural language processing techniques—across 24 months to ensure minimal missingness of data, especially in relation to the surgical QMs. The study was approved by the St Louis VHA’s Research and Development Committee and institutional review board, which also waived the requirement for informed consent given the deidentified analysis.

Quality Metrics and Covariates

Based on contemporary treatment guidelines and previously validated literature, including some of our prior work, we defined 5 process-based QMs that should be met in all patients diagnosed with early-stage NSCLC: timely surgery (defined as surgery within 12 weeks of radiographic suspicion, as previously described by our group),^7,8,9 anatomic resection (via lobectomy or segmentectomy),^10,11,29 minimally invasive approach,^12,13 negative margin,¹⁴ and adequate lymph node sampling.^15,16 Timely surgery has been previously associated with worse cancer-specific outcomes.⁷ Adequate nodal sampling was defined as sampling at least 10 lymph nodes³⁰; however, several sampling levels were assessed in the model (0, 1-4, 5-9, and ≥10 nodes) to avoid dichotomization of a continuous variable. Of note, while several sampling guidelines currently exist for early-stage NSCLC, we chose a count-based standard, as these were prevailing recommendations during the study period.³¹ Additional patient-, treatment-, and tumor-related covariates were also extracted, as described previously by our group.^7,10,25,26

Outcomes

Our primary outcome was OS, defined as the time between date of surgery and death from any cause. Overall survival was assessed using the VHA Vital Status Files³² and was censored at the end of the study follow-up period (May 1, 2020). We also examined the association between the developed score and RFS. Recurrence-free survival was defined as the time between the date of surgery and cancer recurrence and was characterized using a combination of clinical documentation and billing codes suggestive of recurrence, as described previously in the VHA.^25,33 For this study, RFS was censored at the date of the last follow-up or death (which was treated as a competing event in multivariable models).

Score Development

To develop the surgical quality score, a multivariable Cox proportional hazards regression model was constructed to examine the association between OS and surgical QMs. The model also controlled for pertinent patient-, treatment-, and tumor-related variables (including patient age, sex, race, smoking status, body mass index, comorbidity burden, number of prescriptions, hospital volume, tumor location, tumor size, histology, and pathologic stage) in addition to the 5 QMs. Beta coefficients for the QMs were fixed from this model to develop the integer-based VALCAN-O score (0-13 points), with higher scores representing “higher-quality” (ie, guideline-concordant) operations.³⁴ To provide clinical context, patients were further subdivided into risk categories based on the estimated probability of survival at 5 years (ie, score of 12-13: approximately 60% 5-year survival; score of 9-11: approximately 50%-60%; score of 6-8: approximately 40%-50%; and score of 0-5: approximately 40%). Because certain QMs (ie, sublobar resection) may be associated with pulmonary function, we also assessed the VALCAN-O score in subgroup analyses of patients with normal and impaired pulmonary function (forced expiratory volume percentage predicted, ≥80%, 50%-79%, or <50%) with available data.

Statistical Analysis

The data were analyzed from April 1, 2022, to September 1, 2022. Cohort descriptive statistics were presented as mean (SD) for continuous variables and absolute numbers (proportions) for categorical variables. Patients diagnosed with pathologic stage IV disease were excluded from OS and RFS analyses (including score development). Kaplan-Meier curves were used to display OS and RFS. Given the highly curated nature of this data set, missing data were minimal and were displayed using unknown categories. Multivariable, proportional, cause-specific hazard-competing risk regression was used to examine the association between VALCAN-O score and RFS, with recurrence as the event and death as the competing event.³⁵ The 2-sided P < .05 was considered statistically significant. All analyses were performed in SAS, version 9.3 (SAS Institute Inc).

To further apply our findings to current practice trends, we extended our cohort of veterans with early-stage NSCLC using the same methods previously described in this manuscript to include data from 2017 to 2019 (validation cohort 1). We used this extended cohort to describe longitudinal trends in the VALCAN-O score within the VHA from 2006 to 2019 (prior to the COVID-19 pandemic).

In addition, to validate our findings in a nonveteran population, we obtained a similar cohort of patients with early-stage NSCLC from the NCDB (2010-2016; minimally invasive approach data were unavailable prior to 2010). We applied the same inclusion and exclusion criteria to this cohort (validation cohort 2). Since the NCDB defines dates of diagnosis based on clinical criteria,⁷ we defined timely surgery as occurring within 8 weeks of the clinical diagnosis based on prior studies from the NCDB.^8,9

Results

Study Cohort and Quality Metrics

The study included 9628 veterans undergoing surgical treatment for early-stage NSCLC from October 2006 through September 2016. Demographic and tumor-related factors are shown in Table 1. The mean (SD) age of the patients was 67.6 (7.9) years. The cohort included 9278 males (96.4%) and 350 females (3.6%), and all of them identified as being of the following races and ethnicities: Black (1446 [15.0%]), White (7961 [82.7%]), other (130 [1.4%]), or unknown (91 [1.0%]). Among the patients, 5627 (58.4%) were currently smoking at the time of surgery. Most of the tumors were adenocarcinomas (5136 [53.3%]) and exhibited higher-grade features (tumor grades II-IV, 7868 [81.7%]).

Table 1. Characteristics of the Veterans Health Administration Study Population.

Characteristic	Study cohort, No. (%) (N = 9628)
Demographic variables
Age, mean (SD), y	67.6 (7.9)
Sex
Male	9278 (96.4)
Female	350 (3.6)
Racea
Black	1446 (15.0)
White	7961 (82.7)
Other	130 (1.4)
Unknown	91 (0.9)
Smoking status (at time of surgical treatment)
Current	5627 (58.4)
Former	3868 (40.2)
Never	133 (1.4)
BMI
<18.5	797 (8.3)
18.5-24.9	3086 (32.0)
25-29.9	3281 (34.1)
30-34.9	1757 (18.2)
≥35	707 (7.3)
Charlson Comorbidity Index score, median (IQR)	7 (5-8)
No. of unique prescriptions, median (IQR)	13 (8-18)
Distance from hospital, miles
<10	2105 (21.9)
10-50	3878 (40.3)
≥50	3645 (37.9)
Area Deprivation Index
Q1 (least deprived)	2334 (24.2)
Q2	2427 (25.2)
Q3	2494 (25.9)
Q4 (most deprived)	2339 (24.3)
Unknown	34 (0.4)
Tumor-related variables
Histology
Adenocarcinoma	5136 (53.3)
Squamous cell carcinoma	3258 (33.8)
Other	1234 (12.8)
Grade
I	1186 (12.3)
II	4781 (49.7)
III	2959 (30.7)
IV	128 (2.3)
Unknown	574 (6.0)
Location
Right upper lobe	3492 (36.3)
Right middle lobe	535 (5.6)
Right lower lobe	1518 (15.8)
Left upper lobe	2639 (27.4)
Left lower lobe	1329 (13.8)
Unknown	115 (1.2)
Tumor size, mm
≤10	852 (8.8)
11-20	3786 (39.3)
21-30	2620 (27.2)
31-40	1451 (15.1)
>40	711 (7.4)
Unknown	208 (2.2)
Pathologic stage
I	8385 (87.1)
II	794 (8.2)
III	408 (4.2)
IV	41 (0.4)
Surgical quality metrics
Timely surgery
≤12 wk	6633 (68.9)
>12 wk	2995 (31.1)
Incision
Thoracotomy	5642 (58.6)
Minimally invasive	3986 (41.4)
Resection
Lobectomy	6843 (71.1)
Wedge	2101 (21.8)
Segmentectomy	532 (5.5)
Pneumonectomy	152 (1.6)
Nodal sampling adequacy, No. of LNs
0	1000 (10.4)
1-4	2362 (24.5)
5-9	2988 (31.0)
≥10	3278 (34.0)
Surgical marginb
R0	9312 (96.7)
R1+	316 (3.3)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); LNs, lymph nodes.

SI conversion factor: To convert miles to kilometers, multiply by 1.6.

^a As coded in the Veterans Health Administration Corporate Data Warehouse. Other includes any code other than Black or White in the American College of Surgery Facility Oncology Registry Data Standards manual.

^b For the surgical margin, R0 indicates negative margin and R1+ indicates positive margin.

In terms of QMs, most patients received timely surgery (6633 [68.9%]) and underwent lobectomies (6843 [71.1%]) or wedge resections (2101 [21.8%]), while a smaller number of people underwent segmentectomy (532 [5.5%]). The most common surgical approach was thoracotomy (5642 [58.6%]) followed by minimally invasive surgery (3986 [41.4%]). Adequate nodal sampling (defined as ≥10 lymph nodes) was achieved in 3278 patients (34.0%). Finally, most of the operations attained negative surgical margins (9312 [96.7%]).

Score Development

To develop the score, we examined the association between the specified surgical QMs and OS. The median (IQR) follow-up period was 6.2 (2.5-11.4) years. An integer-based score (VALCAN-O score) from 0 (no QMs met) to 13 (all QMs met) was constructed, with higher scores reflecting progressively improved risk-adjusted OS (Table 2; eTable 1 in the Supplement). The score distribution is shown in eFigure 1 in the Supplement.

Table 2. Development of the VALCAN-O Score.

Variablea	aHR (95% CI)	β	P value	VALCAN-O points
Delayed surgery
>12 wk	1 [Reference]			0
≤12 wk	0.89 (0.84 to 0.94)	−0.12	<.001	1
Surgical approach
Open	1 [Reference]			0
Minimally invasive	0.92 (0.87 to 0.97)	−0.09	.003	1
Extent of resection
Wedge	1 [Reference]			0
Lobectomy	0.83 (0.77 to 0.89)	−0.19	<.001	2
Segmentectomy	0.82 (0.72 to 0.94)	−0.19	.004	2
Pneumonectomyb	1.08 (0.88 to 1.34)	0.08	.47	0
Nodal sampling adequacy, LNs
0	1 [Reference]			0
1-4	0.88 (0.89 to 0.96)	−0.13	.008	1
5-9	0.82 (0.74 to 0.90)	−0.20	<.001	2
≥10	0.78 (0.70 to 0.86)	−0.25	<.001	3
Surgical marginc
R1+	1 [Reference]			0
R0	0.58 (0.51 to 0.67)	−0.54	<.001	6

Abbreviations: aHR, adjusted hazard ratio; LNs, lymph nodes; VALCAN-O, Veterans Affairs Lung Cancer Operative quality.

^a Model controlling for displayed covariates in addition to age, sex, race, body mass index, smoking status, Charlson Comorbidity Index score, number of unique prescriptions, hospital volume, location of tumor, histology, pathologic stage, and tumor size. The full model is available in eTable 1 in the Supplement.

^b Pneumonectomy was given a score of 0 to prevent negative scores in the model.

^c For the surgical margin, R0 indicates negative margin and R1+ indicates positive margin.

To further aid with data visualization and to provide a clinical perspective, the patients were divided into risk categories based on the VALCAN-O score. As shown in Figure 1, the median (IQR) OS differed substantially between score categories (0-5 points, 2.6 [1.0-5.7] years; 6-8 points, 4.3 [1.7-8.6] years; 9-11 points, 6.3 [2.6-11.4] years; and 12-13 points, 7.0 [3.0-12.5] years; P < .001). Additionally, risk-adjusted RFS improved in a stepwise manner between the score categories (6-8 vs 0-5 points, multivariable-adjusted hazard ratio [aHR], 0.62; 95% CI, 0.48-0.79; P < .001; 12-13 vs 0-5 points, aHR, 0.39; 95% CI, 0.31-0.49; P < .001) (Figure 1 and eTable 2 in the Supplement). To further test the reliability of this score in various subgroups of patients, we performed several subanalyses based on the degree of pulmonary impairment, surgical year, and final pathologic stage. The VALCAN-O score remained associated with improved OS and RFS in these analyses (eFigure 2 in the Supplement).

Figure 1. Kaplan-Meier Curves Showing the Association Between Veterans Affairs Lung Cancer Operative Quality (VALCAN-O) Score and Overall Survival and Recurrence-Free Survival in a Veterans Health Administration Cohort.

Validation Cohort 1: Temporal Geographic Trends in VHA

We next assessed trends in surgical quality in the VHA from 2006 to 2019. We displayed this as the proportion of operations that met the highest-quality VALCAN-O score category (VALCAN-O score ≥12). As shown in Figure 2, VALCAN-O scores improved substantially over the study period; however, there was substantial regional variation. For example, the proportion of patients receiving highest-quality operations (VALCAN-O score ≥12) in Veterans Integrated Services Network (VISN) 19 increased from 32.5% (2006-2009) to 67.2% (2017-2019). Conversely, the proportion of patients receiving highest-quality operations in VISN 15 remained relatively stagnant throughout the study period (26.7% in 2006-2009 vs 29.2% in 2017-2019).

Figure 2. Geospatial Trends in Veterans Affairs Lung Cancer Operative Quality (VALCAN-O) Score According to Veterans Health Administration Administrative Region, 2006-2019.

The images represent the proportion of operations in each Veterans Integrated Services Network (VISN) region obtaining a VALCAN-O score of 12 points or higher (ie, highest-quality operation), with darker blue representing a higher proportion. Alaska (VISN 20), Hawaii (VISN 21), Puerto Rico (VISN 8), and other US territories were omitted for ease of viewing.

Validation Cohort 2: NCDB Cohort

We then sought to validate our findings in a cohort of patients from the NCDB who had been diagnosed with early-stage NSCLC and were treated at non-VHA hospitals. This cohort included 107 674 patients. Additional demographic information is given in eTable 3 in the Supplement. The QMs were met as follows: timely surgery (85 142 patients [79.1%]), minimally invasive approach (42 199 [39.2%]), lobectomy (78 195 [72.6%]), adequate nodal sampling (37 885 [35.2%]), and negative surgical margin (104 778 [97.3%]). Similar to the VHA cohort, the median (IQR) OS differed substantially between score categories (0-5 points, 3.8 [1.4-8.4] years; 6-8 points, 5.9 [2.6-9.2] years; 9-11 points, 7.6 [3.4 to not reached] years; and 12-13 points, 8.7 [4.4 to not reached] years; P < .001) (Figure 3).

Figure 3. Kaplan-Meier Curves Showing the Association Between Veterans Affairs Lung Cancer Operative Quality (VALCAN-O) Score and Overall Survival in the National Cancer Database Cohort.

Discussion

Using a uniquely compiled data set from the VHA, we developed a practical, easy-to-calculate surgical quality score—the VALCAN-O score—for patients diagnosed with resectable early-stage NSCLC. The score reflects the risk-adjusted association between 5 previously validated, modifiable surgical QMs (ie, timely surgery, anatomic resection, surgical approach, adequate nodal sampling, and negative surgical margin) and long-term OS. We also found that the score was associated with RFS. While the score was developed in a cohort of nearly 10 000 veterans, which represents an admittedly unique patient population, we validated the score in a cohort of more than 100 000 US adults from the NCDB. Our data suggest that efforts to standardize and optimize surgical quality may have disproportionate repercussions for patients diagnosed with early-stage NSCLC who are receiving curative-intent surgical treatment.

Most contemporary treatment guidelines converge in recommending several process-based surgical QMs that should be routinely met in patients diagnosed with early-stage NSCLC. These include timely surgery, anatomic resection, minimally invasive surgical approach, adequate nodal sampling, and negative surgical margin, when possible.^4,5,6 Although these QMs are widely considered guideline-concordant standards of care, it is striking to note the relatively poor adherence to these measures within both the VHA and civilian hospitals (NCDB). For example, we found that only 34.0% (35.2% NCDB) of the patients in either cohort received adequate nodal sampling; similarly, only 41.4% (39.2% NCDB) of the patients received minimally invasive resections and 21.8% (21.3% NCDB) of the patients received nonanatomic wedge resections. Although we found that adherence to these QMs did improve over time within the VHA, we still identified wide variability even in a modern cohort of patients through 2019. Our findings clearly indicate that further efforts are needed to improve the dissemination and implementation of guideline-concordant practices in thoracic surgery to improve lung cancer outcomes in both VHA and non-VHA settings.²³

Efforts to improve surgical quality, such as through the optimization of the VALCAN-O components, may be particularly fruitful within a closed practice environment like the VHA. The VHA has recently invested substantial resources into dissemination and implementation efforts, including through the creation of institutes like the Center for Evaluation and Implementation Resources.³⁶ The VHA has already demonstrated significant success in standardizing lung cancer care beyond what is observed in the general population. For example, the VHA has rapidly implemented robust lung cancer screening programs for veterans³⁷; conversely, screening rates among the general US adult population are much lower.³⁸ Similarly, efforts to standardize surgical quality for early-stage lung cancer may therefore be highly beneficial and readily adoptable within the VHA.

It is also important to emphasize that the VHA provides veterans with a quality of care that is comparable to or better than that received at civilian hospitals. We found that rates of adherence to QMs were similar in the VHA and the NCDB. This is an important observation, as the VHA is often perceived as delivering “low-quality” or inferior care to veterans. However, our findings demonstrate that veterans receive guideline-concordant care at rates similar to those for the general population. More important, the VHA treats a patient population with a much greater comorbidity burden. Therefore, while unadjusted survival appears to be worse among veterans, prior work by our group has demonstrated that after appropriate risk-adjustment, short- and long-term outcomes are similar (if not superior) in veterans compared with the general population.¹⁰ It is important to emphasize this finding given ongoing policy endeavors that may fundamentally alter the paradigm of care being delivered through the VHA.³⁹ Our study again suggests that future policies should focus on improving this unique system that is already performing well.⁴⁰ Nonetheless, both the VHA and NCDB data sets suggest that QM adherence is variable, presenting an opportunity for significant quality improvement efforts in both practice settings.

Last, it is important to reflect on the varying definitions of surgical quality.⁴¹ Traditionally, the surgical literature has focused on minimizing short-term postoperative complications and mortality as a metric of high-quality care. For example, the Society of Thoracic Surgeons Star Rating System grades hospitals based on their risk-adjusted rates of 30-day morbidity and mortality.⁴² Although important, those events happen infrequently after early-stage lung cancer operations.⁴³ Furthermore, they miss a likely more important goal of curative-intent lung cancer resection: achieving long-term survival. Additionally, the diametrically opposed association between certain QMs (eg, sublobar resection, adequate nodal sampling) and short- vs long-term outcomes is noteworthy. For example, while sublobar resections are associated with a significantly lower risk of postoperative morbidity and mortality, this comes at the cost of worse long-term cancer-specific outcomes, including higher rates of cancer recurrence.^11,29 Therefore, thoracic surgeons must continue to adopt a less myopic view of surgical quality that focuses on long-term outcomes, such as that which the VALCAN-O score offers.

Strengths and Limitations

This study is strengthened by the highly unique VHA data set, which was assembled over a period of more than 24 months by a dedicated team of VHA-based researchers who queried electronic health records, creating this relatively homogeneous cohort of patients receiving definitive surgical treatment. In addition, the data set includes outcomes data like recurrence, which is infrequently available through other national, multi-institutional databases. Conversely, this study also had some limitations. First, the VHA treats a unique patient population that consists disproportionately of males with a heavy comorbidity burden. Despite this, previous work by our group has shown that patterns of care and outcomes are similar between veteran and nonveteran populations.¹⁰ In addition, we validated this score in a large cohort of patients with early-stage NSCLC from the general population. Therefore, efforts to standardize and improve surgical quality may improve early-stage lung cancer outcomes in both the VHA and civilian systems. Second, it is important to acknowledge that these QMs cannot be met in all situations. For example, a certain subset of operations will require a thoracotomy for anatomic reasons¹³ or may necessitate a sublobar resection due to diminished lung function. However, our data indicated that even sicker patients with compromised pulmonary function benefit from better adherence to these surgical QMs. Third, cause-specific death, a potentially more appropriate end point, was unavailable. Fourth, we assessed lymph node sampling adequacy using a count-based approach. This was the prevailing recommendation during the study period and also the only available data within the NCDB. However, future studies will need to assess how adherence to other guidelines (eg, station-based standards) may affect outcomes.

Conclusions

The findings of this cohort study suggest that the VALCAN-O score is a practical, easy-to-calculate metric of surgical quality in early-stage NSCLC. Although adherence to quality measures has improved over time, wide variability in guideline-concordant care still exists in the US. Efforts to standardize and optimize long-term cancer-specific outcomes are critical for patients diagnosed with resectable lung cancer who are receiving curative-intent treatment. Future policy efforts should focus on more widespread implementation of these metrics.

Supplement.

eTable 1. Multivariable Cox Proportional Hazards Model for Overall Survival

eFigure 1. VALCAN-O Score Distribution

eTable 2. Multivariable Competing Risk Model for Recurrence-free Survival

eFigure 2. Relationship Between VALCAN-O and Overall Survival and Recurrence-Free Survival in Subgroups Based on Pulmonary Function (A, B), Surgical Year (C, D), and Final Pathologic Stage (E, F)

eTable 3. Validation Cohort (NCDB)