You are here

Article Text

Article menu

Download PDFPDF

Original research
Failure to administer recommended chemotherapy: acceptable variation or cancer care quality blind spot?
  1. Ryan J Ellis1,2,
  2. Cary Jo R Schlick1,
  3. Joe Feinglass3,
  4. Mary F Mulcahy4,5,
  5. Al B Benson4,5,
  6. Sheetal M Kircher4,5,
  7. Tony D Yang1,4,
  8. David D Odell1,4,
  9. Karl Bilimoria1,2,4,
  10. Ryan P Merkow1,2,4
  1. 1 Surgical Outcomes and Quality Improvement Center, Department of Surgery, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA
  2. 2 Division of Research and Optimal Patient Care, American College of Surgeons, Chicago, Illinois, USA
  3. 3 Division of General Internal Medicine and Pediatrics, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  4. 4 Robert H Lurie Comprehensive Cancer Center, Chicago, Illinois, USA
  5. 5 Department of Medicine, Northwestern University Division of Hematology/Oncology, Chicago, Illinois, USA
  1. Correspondence to Dr Ryan P Merkow; ryan.merkow{at}nm.org

Abstract

Background Chemotherapy quality measures consider hospitals compliant when chemotherapy is recommended, even if it is not received. This may mask shortcomings in cancer care delivery. Objectives of this study were to (1) identify patient factors associated with failure to receive recommended chemotherapy without a documented contraindication and (2) assess hospital variation in failure to administer recommended chemotherapy.

Methods Patients from 2005 to 2015 with breast, colon and lung cancers who failed to receive recommended chemotherapy were identified using the National Cancer Database. Hospital-level rates of failure to administer recommended chemotherapy were calculated, and patient and hospital factors associated with failure to receive recommended chemotherapy were identified by multivariable logistic regression.

Results A total of 183 148 patients at 1281 hospitals were analysed. Overall, 3.5% of patients with breast, 6.6% with colon and 10.7% with lung cancers failed to receive recommended chemotherapy. Patients were less likely to receive recommended chemotherapy in all cancers if uninsured or on Medicaid (p<0.05), as were non-Hispanic black patients with both breast and colon cancer (p<0.001). Significant hospital variation was observed, with hospital-level rates of failure to administer recommended chemotherapy as high as 21.8% in breast, 40.2% in colon and 40.0% in lung cancers.

Conclusions and relevance Though overall rates are low, failure to receive recommended chemotherapy is associated with sociodemographic factors. Hospital variation in failure to administer recommended chemotherapy is masked by current quality measure definitions and may define a significant and unmeasured difference in hospital quality.

  • quality measurement
  • surgery
  • health services research

https://doi.org/10.1136/bmjqs-2019-009742

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    While death rates for many cancers have decreased significantly, there are still an estimated 600 000 cancer-related deaths each year in the USA.1 Breast, colon and lung cancers are among the three most common malignancies diagnosed in the USA and have well-defined treatment pathways with known benefits to chemotherapy in select subsets of patients.2–4 The National Quality Forum (NQF) has endorsed several metrics related to the appropriate administration of chemotherapy. Oversight and accreditation organisations, including the Centers for Medicaid and Medicare Services and American College of Surgeons Commission on Cancer (CoC), use these measures to evaluate hospital cancer care quality.5 Thresholds for compliance are defined by the CoC, and cut-offs are defined for subset of measures for formal evaluation of hospital quality. For example, hospitals are considered compliant with the CoC Quality Improvement measure for non-small cell lung cancer if more than 85% of patients receive systemic chemotherapy within 4 months before or 6 months after surgery when pathologically appropriate.6

    The CoC quality metrics regarding appropriate use of adjuvant therapy in patients with breast, colon and lung cancers are structurally similar. Patients treated at CoC hospitals who meet strict inclusion criteria are expected to receive chemotherapy unless alternative documentation is provided.6–8 Patients deemed too frail or systemically ill to tolerate chemotherapy are not counted against the hospital in quality calculations.9 10 Similar exemptions exist for patients who die prior to administration of planned chemotherapy and for those patients that received a chemotherapy recommendation but did not undergo treatment. The goal of these exceptions is to ensure that hospitals are not penalised for situations in which chemotherapy is at least recommended, even if a patient subsequently refuses administration or is lost to follow-up.

    However, given significant evidence regarding disparities in cancer care,11–14 there is a concern that current definitions may mask a true gap in quality. One example would include a vulnerable patient with low health literacy ‘slipping through the cracks’ and never receiving chemotherapy despite a recommendation by a physician, which is considered compliant with current quality measure definitions. Hospitals then artificially appear to be high performers on what is most important: the timely receipt of recommended chemotherapy to patients with cancer. The objectives of this study were (1) to identify patient factors associated with failure to receive recommended chemotherapy and (2) to assess hospital variation in failure to administer recommended chemotherapy in the absence of documented contraindications.

    Methods

    Data source and study population

    The 2005–2015 National Cancer Database (NCDB) was the primary dataset for this retrospective cohort study. The NCDB is sponsored by the CoC of the American College of Surgeons and the American Cancer Society. Approximately 30% of USA hospitals participate in the NCDB, which captures approximately 70% of newly diagnosed cancers nationally each year.15 16 Data are abstracted by trained registrars and periodically audited.17 The NCDB is the primary dataset used for quality evaluation by the CoC. The NCDB collects clinical and pathological data based on the American Joint Committee on Cancer (AJCC) guidelines.18 These staging data, paired with disease-specific inclusion and exclusion criteria, are used to identify patients to be included in hospital-level quality measure calculations.

    The three most common cancer sites in the USA with chemotherapy-based quality measures were selected for study: breast, colon and lung cancer. The quality metrics were reconstructed using the CoC definitions for the three quality measures that are publicly available. Briefly, eligible cases receiving treatment at the reporting facility for each of the three cancer sites were identified using Facility Oncology Registry Data Standards definitions.6–8 Measure inclusion criteria included diagnosis, histology codes and patient record completeness as laid out by the CoC. Measures are designed to capture patient populations with robust evidence underlying the potential benefit of chemotherapeutics (eg, elderly women with breast cancer are excluded from the quality measure). The breast cancer measure includes receipt of adjuvant chemotherapy in patients with T1N0M0 and stage IB–III hormone receptor negative cancer; the colon cancer measure includes receipt of adjuvant chemotherapy in patients with stage III colon cancer; the lung cancer measure includes either adjuvant or neoadjuvant therapy in patients with pN1 or pN2 NSCLC without evidence of metastatic disease. Clinical exclusion criteria were applied per CoC guidelines. Eligible cases were considered the denominator for each quality measure. Cases were included beginning in the calendar year of the initial release of the CoC quality measure (2006 for breast cancer, 2005 for colon cancer, and 2014 for lung cancer).

    Outcome variable: failure to administer recommended chemotherapy

    Adherence to appropriate use of chemotherapy was calculated in two ways. First, cases were considered compliant with the measure using the previously defined, NQF endorsed definitions. For this definition, cases are considered ‘compliant’ if chemotherapy is not administered due to patient comorbidities, death, and if chemotherapy is recommended but not administered due to patient refusal or lost to follow-up. All other eligible cases are considered failures towards the measure. An alternative definition was also calculated excluding cases where chemotherapy was recommended but not administered from the quality measure numerator. The difference in hospital-level compliance using the two definitions above yielded the hospital-level percentage of patients failing to receive recommended chemotherapy.

    Covariates: patient sociodemographic and clinical characteristics

    Patient-level variables of interest included year of treatment, patient age, sex, race/ethnicity, Charlson-Deyo Score (a composite measure of preexisting medical comorbidities) and socioeconomic variables.19 Disease-specific variables of interest included T and N staging. Pathological staging was performed using AJCC staging criteria when appropriate, with stages included for study defined by the NQF defined quality measure.20 Hospital-level variables included hospital type (academic vs non-academic), hospital census region and patient travel distance to the hospital.

    Hospital-level analysis

    Hospital-level rates of failure to administer recommended chemotherapy were calculated based on the difference between the existing CoC compliance definition and the hospital compliance rate when patients failing to receive recommended chemotherapy were excluded from the numerator. All hospitals with at least 10 cases meeting the inclusion criteria were included in the hospital-level analysis.

    Statistical analysis

    Basic descriptive statistics were performed separately for each cancer site. Significant associations between failure to receive recommended chemotherapy and patient-level variables for each individual cancer site were assessed by multivariable logistic regression models with robust SEs adjusted for patient clustering within hospitals. Hospital-level odds of failing to receive recommended chemotherapy was assessed using both adjusted and unadjusted hierarchical multivariable logistic regression models. Adjusted models included all relevant patient-level variables as fixed effects. Hospitals were considered outliers if the hospital-level ORs and CIs were significantly different than 1.0 in hierarchical models. Tests of significance were two sided with p values considered significant at the 0.05 level. Statistical analyses were performed using Stata V.15.1 (StataCorp).

    Results

    Cohort development and descriptive statistics

    Overall, 183 148 patients met inclusion criteria for administration of chemotherapy (82 598 patients with breast cancer, 92 077 patients with colon cancer and 8473 patients with lung cancer; online supplementary appendix figure 1). The smaller number of lung cancer patients was due predominantly to the recency of the measure and the relative infrequency of surgical management in patients with N2 disease. Consistent with CoC definitions, all patients in the breast cancer cohort were female under the age of 70. Patient characteristics of the final cohort for each cancer site are shown in table 1.

    Supplemental material

    View this table:
    Table 1

    Patient cohort characteristics

    Factors associated with failure to receive recommended chemotherapy

    The overall rate of failure to receive recommended chemotherapy was 3.5% in patients with breast cancer. Patients were more likely to fail to receive recommended chemotherapy if over the age of 55 (4.5% vs 2.5% if <55, adjusted OR (aOR) 1.90, 95% CI 1.76 to 2.04), non-Hispanic black (4.1% vs 3.4% in non-Hispanic white patients, aOR 1.26, 95% CI 1.14 to 1.40), from a low-income area (4.3% vs 3.1% in highest income areas, aOR 1.36, 95% CI 1.15 to 1.61), uninsured or on Medicaid (4.0% vs 3.4% if insured, aOR 1.30, 95% CI 1.16 to 1.45) or treated at a non-academic hospital (3.8% vs 3.0% if non-academic, aOR 1.26, 95% CI 1.13 to 1.41). Regional associations were also noted, with patients managed in East North Central, Mountain and Pacific regions more likely to fail to receive recommended chemotherapy (all p<0.05; table 2). There was a stepwise increase in failure to receive recommended chemotherapy with decreasing N-stage, with N0 patients (4.2%, aOR 2.60, 95% CI 2.23 to 3.04 vs N2/3) and N1 patients (2.5%, aOR 1.42, 95% CI 1.20 to 1.68 vs N2/3) significantly more likely to fail to receive chemotherapy than patients with N2 or N3 disease (1.9%).

    View this table:
    Table 2

    Factors associated with failure to receive recommended chemotherapy by disease site

    In patients with colon cancer, the overall rate of failure to receive physician recommended chemotherapy was 6.6%. Patients with colon cancer were more likely to fail to receive recommended chemotherapy if over the age of 70 (10.5% vs 3.4% if <55, aOR 3.56, 95% CI 3.23 to 3.92), non-Hispanic black (8.1% vs 6.3% in non-Hispanic white patients, aOR 1.38, 95% CI 1.26 to 1.51), from a low-income area (7.8% vs 5.5% in highest income areas, aOR 1.30, 95% CI 1.16 to 1.46), uninsured or on Medicaid (7.7% vs 6.4% in insured, aOR 1.62, 95% CI 1.46 to 1.79) or with a high Charlson comorbidity score (9.2% vs 6.1% with a low Charlson comorbidity score, aOR 1.26, (95%CI 1.13 to 1.41). Regional associations were also noted, with several regions demonstrating higher rates of failure (all p<0.05; table 2). Patients with N1 disease were significantly more likely to fail to receive recommended chemotherapy than patients with N2 disease (7.1% vs 5.6%, aOR 1.27, 95% CI 1.19 to 1.34).

    Among patients with lung cancer, 10.7% failed to receive recommended chemotherapy. Patients with lung cancer were more likely to fail to receive recommended chemotherapy if over the age of 70 (14.1% vs 7.4% if <55, aOR 2.26, 95% CI 1.75 to 2.93), uninsured or on Medicaid (11.6% vs 10.6% in insured patients, aOR 1.40, 95% CI 1.08 to 1.82) or with N1 disease (12.5% vs 8.4% for N2, aOR 1.56, 95% CI 1.33 to 1.83).

    Hospital variation in failure to administer recommended chemotherapy by cancer type

    Hospital-level analyses included 1226 hospitals for the breast cancer measure, 1275 hospitals for the colon cancer measure and 302 hospitals for the lung cancer measure. The median hospital-level rates of failure to administer recommended chemotherapy were 3.2% for breast cancer (IQR 0%–5.5%), 6.2% for colon cancer (IQR 3.6%–9.1%) and 9.1% for lung cancer (IQR 4.9%–15.4%; table 3). Rates of hospital-level failure to administer physician recommended chemotherapy were as high as 21.8% in breast cancer, 40.2% in colon cancer and 40.0% in lung cancer. A 0% rate of failure to administer recommended chemotherapy was observed in 309 hospitals (25.2%) in breast cancer, 111 (8.7%) in colon cancer and 62 (20.5%) in lung cancer. Overall variation in hospital-level compliance is illustrated in figure 1, both including and excluding those patients that failed to receive recommended chemotherapy.

    Figure 1
    Figure 1

    Hospital-level compliance rates with and without failure to administer recommended chemotherapy considered in metric. Hospital-level compliance rates with appropriate chemotherapy for breast cancer (n=1226; A), colon cancer (n=1275; B) and lung cancer (n=302; C). The line is hospital-level compliance when including failure to administer recommended chemotherapy. Triangles represent hospital-level compliance when failure to administer recommended chemotherapy is not considered compliant. The distance from each triangle to the line is the rate of failure to administer recommended chemotherapy at a given hospital. Note: these analyses exclude 566 patients at 100 hospitals for breast cancer, 317 patients at 58 hospitals for colon cancer and 2708 patients at 685 hospitals for lung cancer.

    View this table:
    Table 3

    Hospital-Level rates of failure to administer recommended chemotherapy by disease site

    Hierarchical logistic regression modelling was then performed for each cancer type to identify hospital-level odds of failure to receive recommended chemotherapy. In completely unadjusted models, there were 95 high outliers (poor performers, 7.8%) in breast cancer, 184 (14.4%) in colon cancer and 18 (6.0%) in lung cancer. In models adjusting for patient characteristics and sociodemographics, there were 86 high outliers in breast cancer (7.0%), 169 in colon cancer (13.3%) and 17 in lung cancer (5.6%; figure 2). Across all cancer types, adjustment for sociodemographics recategorised 44 hospitals that would have been identified as high outliers in unadjusted models. Notably, only 29 hospitals were high outliers in more than one cancer type. Similar patterns were observed in examining low outliers (high performers).

    Figure 2
    Figure 2

    Unadjusted and adjusted hospital-level odds of failure to administer recommended chemotherapy. Hospital-level ORs for failure to receive recommended chemotherapy for breast (A; n=1277), colon (B, n=1279) and lung cancer (C, n=973) adjusting for both patient sociodemographic and disease characteristics. Adjusted analyses identified 77 high outliers (poor performers; 6.0%) and 21 low outliers (high performers, 1.6%) in breast cancer; 171 high outliers (13.4%) and 108 low outliers (8.4%) in colon cancer; and 13 high outliers (1.3%) and 3 low outliers (0.3%) in lung cancer. Each black dot represents a hospital-level OR of failure to receive recommended chemotherapy, and whiskers representd standard measurement error. Horizontal line marks a hospital OR of 1.0. Note: these analyses exclude 566 patients at 100 hospitals for breast cancer, 317 patients at 58 hospitals for colon cancer and 2708 patients at 685 hospitals for lung cancer.

    Discussion

    In this study, failure to receive recommended chemotherapy was shown to be associated with demographic, socioeconomic and disease-related variables across multiple cancer types. Hospital variation was notable, with many hospitals achieving very low rates of failure to administer recommended chemotherapy and others identified as low performers. Though overall rates are low, several hospitals had extremely high failure rates that are masked by the current quality measure definitions. Hospital variation in failure to administer recommended chemotherapy may define a significant and unmeasured difference in hospital quality.

    Both racial minorities and socioeconomically disadvantaged populations have been shown to have worse outcomes across many studies and cancer types.12–14 21–24 The variation in cancer care delivery demonstrated in this study is simultaneously in line with previously identified disparities and substantially more nuanced, as hospitals are actually given credit on quality measures for those patients who fail to receive recommended chemotherapy.6–8 Moreover, this study may even underestimate this disparity, as those patients that are early lost to follow-up are excluded from the quality measure. While the goal of this policy is to not punish a hospital for a patient that is non-compliant or is lost to follow-up, the striking variation in hospital-level rates of failure to administer recommended chemotherapy implies that there are significant differences in the effectiveness of local efforts in mitigating these demographic and socioeconomic factors at some facilities. Some of this variation is likely due to unmeasured local factors such as health literacy, language barriers and comorbid substance abuse or psychiatric illness. There are also local efforts that may reduce lost to follow-up or reduce financial and social impediments to receiving care, such as social workers, financial counsellors and patient navigators. Future research should explore additional explanatory variables and how it may pertain to patients failing to receive recommended care.

    Beyond the apparent disparity identified, this analysis brings up two notable questions about measurement of hospital quality: risk adjustment for process measures and socioeconomic risk adjustment. Process measures have traditionally not undergone risk adjustment, as they are by definition measures that that should be taken for all patients who meet inclusion criteria (eg, chemotherapy for disseminated cancer). Rather, patients are removed from process measure calculations based on exclusion criteria (eg, patient comorbidities). However, the conclusions from the NQF regarding the role of socioeconomic status in quality measurement have made this policy more fluid.25 Briefly, the NQF concluded that socioeconomic factors could be used for risk adjustment in both outcome and some process measures when there is a logical connexion between the socioeconomic factor and the quality measure (eg, insurance status and expensive chemotherapeutics) and empirical evidence exists that the connexion may be affecting outcomes. Presentation of such data may take the form of clinically adjusted models stratified by sociodemographic groups. Application of these principles to measures such as those studied in this manuscript has far-reaching implications as care moves towards a more bundled, value-based model.26

    There are a few potential solutions to the identified shortcoming in existing quality measures. One method would be to include socioeconomic variables as risk adjusters to ensure that hospitals taking care of more disadvantaged populations are not inadvertently penalised. While our adjusted analyses indicate that some hospitals may have a different assessment based on adjusted models, this may be inappropriate in this setting per NQF guidelines, as this study also suggests that very low rates of failure to administer recommended chemotherapy can be achieved. Alternatively, caveats for failure to administer recommended chemotherapy could be removed entirely from process measure calculations. This approach could disproportionately punish critical access and safety net hospitals, which may treat high proportions of patients that may refuse chemotherapy or be lost to follow-up.27 28 A third potential solution would be a simple cap on the proportion of patients at a facility that are included in the numerator of the quality measure when coded as failing to receive recommended chemotherapy. This straightforward solution would not change the quality measure denominator, minimise hospitals being penalised for patient non-compliance and incentivise hospitals with high rates of failure to administer recommended therapy to improve in that area. As hospitals improved, the benchmark ‘cap’ could be adjusted as necessary to continue to motivate improvement.

    Another notable result of this study was the high rate of failure to receive recommended chemotherapy with lower nodal burden of disease observed across all cancer types. Because chemotherapy was recommended by the physician, this is unlikely to represent guideline non-adherence or a physician-level knowledge issue. Rather, we postulate that physicians may be less likely to emphatically recommend chemotherapy in patients with less perceived benefit from chemotherapy. Further study of this association is warranted.

    This study has limitations. First, the quality measures for the three cancers studied have been in place for variable lengths of time which coincides with increasing awareness of surgical quality. This limitation was mitigated by analysing each cancer separately and adjusting for year of diagnosis in all models. Second, we do not know the nature of the conversations taking place at the individual level between physicians and patients, and any steps taken to attempt to improve overall administration of chemotherapy must be cognizant of patient autonomy. Third, the nature of the CoC quality measures excludes many patients with an indication for chemotherapy (eg, patients with distant metastatic disease) and thus may not accurately capture more global cancer care inequalities. However, we believe this is the appropriate population frame for this study, as we are focused on exploring existing quality measures, rather than cancer care inequalities more generally. Finally, the differential staging and inclusion criteria for the three disease may make interpretation of aggregate analysis challenging. However, the similar patient-level trends across all cancer types indicate that overall hospital-level analysis of failure to administer recommended chemotherapy is valid.

    Conclusion

    Socioeconomic and disease-related factors are associated with failure to administer recommended chemotherapy, and significant hospital variation exists in the proportion of patients that fail to receive recommended chemotherapy. These findings reveal a significant blind spot in the current quality measure definitions. While socioeconomic and cultural factors can be challenging to overcome while maintaining patient autonomy, the presence of many facilities with near-perfect administration of recommended chemotherapy indicates optimisation could occur. Modification of existing quality measures to limit a hospital’s ability to get credit for cases where chemotherapy was recommended but not administered or development of a separate cancer measure assessing chemotherapy failures should be considered to address these shortcomings.

    References

    1. 1.↵
      2. Siegel RL ,
      3. Miller KD ,
      4. Jemal A , et al
      . Cancer statistics, 2018. CA Cancer J Clin 2018;68:7–30.doi:10.3322/caac.21442
      OpenUrlCrossRefPubMed
    2. 2.↵
      2. Kumar Tyagi N ,
      3. Dhesy-Thind S
      . Clinical practice guidelines in breast cancer. Curr Oncol 2018;25:151–S60.doi:10.3747/co.25.3729
      OpenUrl
    3. 3.↵
      2. Messersmith WA
      . Systemic management of colorectal cancer. J Natl Compr Canc Netw 2017;15:699–702.doi:10.6004/jnccn.2017.0077
      OpenUrlAbstract/FREE Full Text
    4. 4.↵
      2. Bironzo P ,
      3. Di Maio M
      . A review of guidelines for lung cancer. J Thorac Dis 2018;10(Suppl 13):S1556–S1563.doi:10.21037/jtd.2018.03.54
      OpenUrl
    5. 5.↵
      1. CoC Quality of Care Measures
      . Standard 4.4 and 4.5 expected performance rates, 2018 surveys, 2018. Available: https://www.facs.org/quality-programs/cancer/ncdb/qualitymeasures [Accessed 21 Dec 2018].
    6. 6.↵
      1. Commission on Cancer Programs Practice Profile Reports
      . Lung measure specifications, 2018. Available: https://www.facs.org/~/media/files/quality%20programs/cancer/ncdb/measure%20specs%20nscl.ashx. [Accessed 21 Dec 2018].
    7. 7.↵
      1. Commission on Cancer Programs Practice Profile Reports
      . Breast measure specifications, 2018. Available: https://www.facs.org/~/media/files/quality%20programs/cancer/ncdb/measure%20specs%20breast.ashx [Accessed 21 Dec 2018].
    8. 8.↵
      1. Commission on Cancer Programs Practice Profile Reports
      . Colon measure specifications, 2018. Available: https://www.facs.org/~/media/files/quality%20programs/cancer/ncdb/measure%20specs%20colon.ashx [Accessed 21 Dec 2018].
    9. 9.↵
      2. van den Broek CBM ,
      3. Puylaert CCEM ,
      4. Breugom AJ , et al
      . Administration of adjuvant chemotherapy in older patients with stage III colon cancer: an observational study. Colorectal Dis 2017;19:O358–O364.doi:10.1111/codi.13876
      OpenUrl
    10. 10.↵
      2. Gridelli C ,
      3. Langer C ,
      4. Maione P , et al
      . Lung cancer in the elderly. JCO 2007;25:1898–907.doi:10.1200/JCO.2006.10.3085
      OpenUrlAbstract/FREE Full Text
    11. 11.↵
      2. Grant SR ,
      3. Walker GV ,
      4. Koshy M , et al
      . Impact of insurance status on radiation treatment modality selection among potential candidates for prostate, breast, or gynecologic brachytherapy. Int J Radiat Oncol Biol Phys 2015;93:968–75.doi:10.1016/j.ijrobp.2015.08.036
      OpenUrl
    12. 12.↵
      2. Parikh AA ,
      3. Robinson J ,
      4. Zaydfudim VM , et al
      . The effect of health insurance status on the treatment and outcomes of patients with colorectal cancer. J Surg Oncol 2014;110:227–32.doi:10.1002/jso.23627
      OpenUrlCrossRefPubMed
    13. 13.↵
      2. Freedman RA ,
      3. Virgo KS ,
      4. He Y , et al
      . The association of race/ethnicity, insurance status, and socioeconomic factors with breast cancer care. Cancer 2011;117:180–9.doi:10.1002/cncr.25542
      OpenUrlCrossRefPubMedWeb of Science
    14. 14.↵
      2. Silber JH ,
      3. Rosenbaum PR ,
      4. Clark AS , et al
      . Characteristics associated with differences in survival among black and white women with breast cancer. JAMA 2013;310:389–97.doi:10.1001/jama.2013.8272
      OpenUrlCrossRefPubMedWeb of Science
    15. 15.↵
      2. Bilimoria KY ,
      3. Stewart AK ,
      4. Winchester DP , et al
      . The National cancer data base: a powerful initiative to improve cancer care in the United States. Ann Surg Oncol 2008;15:683–90.doi:10.1245/s10434-007-9747-3
      OpenUrlCrossRefPubMedWeb of Science
    16. 16.↵
      2. Merkow RP ,
      3. Rademaker AW ,
      4. Bilimoria KY
      . Practical guide to surgical data sets: National cancer database (NCDB). JAMA Surg 2018;153:850–1.
      OpenUrl
    17. 17.↵
      About the Commission on cancer, 2018. Available: https://www.facs.org/quality-programs/cancer/coc/about [Accessed 21 Dec 2018].
    18. 18.↵
      Participant user files, 2018. Available: https://www.facs.org/quality-programs/cancer/ncdb/puf [Accessed 21 Dec 2018].
    19. 19.↵
      2. Deyo R ,
      3. Cherkin DC ,
      4. Ciol MA
      . Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613–9.doi:10.1016/0895-4356(92)90133-8
      OpenUrlCrossRefPubMedWeb of Science
    20. 20.↵
      Cancer staging manual, 2018. Available: https://cancerstaging.org/references-tools/deskreferences/Pages/default.aspx [Accessed 21 Dec 2018].
    21. 21.↵
      2. Parikh-Patel A ,
      3. Morris CR ,
      4. Kizer KW
      . Disparities in quality of cancer care: the role of health insurance and population demographics. Medicine 2017;96:e9125.doi:10.1097/MD.0000000000009125
    22. 22.↵
      2. Al-Husseini MJ ,
      3. Saad AM ,
      4. Jazieh KA , et al
      . Outcome disparities in colorectal cancer: a SEER-based comparative analysis of racial subgroups. Int J Colorectal Dis 2019;34:285-292.doi:10.1007/s00384-018-3195-3
    23. 23.↵
      2. Hildebrand JS ,
      3. Wallace K ,
      4. Graybill WS , et al
      . Racial disparities in treatment and survival from ovarian cancer. Cancer Epidemiol 2019;58:77–82.doi:10.1016/j.canep.2018.11.010
      OpenUrl
    24. 24.↵
      2. Swords DS ,
      3. Mulvihill SJ ,
      4. Brooke BS , et al
      . Disparities in utilization of treatment for clinical stage I-II pancreatic adenocarcinoma by area socioeconomic status and race/ethnicity. Surgery 2019;165:751–9.doi:10.1016/j.surg.2018.10.035
      OpenUrl
    25. 25.↵
      Risk adjustment for socioeconomic status or other sociodemographic factors August 2014. Available: https://www.qualityforum.org/Publications/2014/08/Risk_Adjustment_for_Socioeconomic_Status_or_Other_Sociodemographic_Factors.aspx [Accessed 21 Dec 2018].
    26. 26.↵
      2. Joynt KE ,
      3. De Lew N ,
      4. Sheingold SH , et al
      . Should Medicare value-based purchasing take social risk into account? N Engl J Med 2017;376:510–3.doi:10.1056/NEJMp1616278
      OpenUrl
    27. 27.↵
      2. Joynt KE ,
      3. Harris Y ,
      4. Orav EJ , et al
      . Quality of care and patient outcomes in critical access rural hospitals. JAMA 2011;306:45–52.doi:10.1001/jama.2011.902
      OpenUrlCrossRefPubMedWeb of Science
    28. 28.↵
      2. Joynt KE ,
      3. Jha AK
      . Characteristics of hospitals receiving penalties under the hospital readmissions reduction program. JAMA 2013;309:342–3.doi:10.1001/jama.2012.94856
      OpenUrlCrossRefPubMedWeb of Science

    Footnotes

    • Contributors Study conception and design: RJE, CJRS, MFM, ABB, SMK, TY, DDO, KB and RPM. Acquisition of data: RJE and CJRS, Feinglass Analysis and Interpretation of data: RJE, CJRS, JF and RPM. Drafting of manuscript: RJE and RPM. Critical revision: RJE, CJRS, JF, MFM, ABB, SMK, TY, DDO, KB and RPM.

    • Funding This study was funded by Center for Strategic Scientific Initiatives, National Cancer Institute grant no: K07CA216330; American Cancer Society grant no: IRG-18-163-24; National Heart, Lung and Blood Institute grant no: K08HL145139; Agency for Healthcare Research and Quality grant no: 5T32HS000078, K12HS026385.

    • Competing interests None declared.

    • Patient consent for publication Not required.

    • Ethics approval This study was considered non-human subjects research by the Northwestern institutional review board and exempt from approval.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data availability statement Data may be obtained from a third party and are not publicly available.

    Linked Articles