K.F.R. acquired the data, and in collaboration with S.M.W., designed the study and conceptualized the hypotheses; K.F.R., J.C., and S.M.W. performed the data analysis and interpretation. K.F.R. and J.V.N. drafted and revised the manuscript; critical revision and substantive editing was undertaken by K.F.R. and S.M.W.; K.F.R. provided supervision; J.V.N. provided administrative, technical, and material support.
Presented as a poster at the Academy Health 2010 Annual Research Meeting; June 27-29, 2010; Boston, Massachusetts.
K.F.R. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
In 1999, a multidisciplinary panel of experts in colorectal cancer reviewed the relevant medical literature and issued a consensus recommendation for a 12-lymph node (LN) minimum examination after resection for colon cancer. Some authors have shown racial/ethnic differences in receipt of this evidence-based care. To date, however, none has investigated the correlation between disparities in LN examination and disparities in outcomes after colon cancer treatment.
This retrospective analysis used California Cancer Registry linked to California Office of Statewide Health Planning and Development discharge data (1996-2006). Chi-square analysis, logistic regression, and Cox proportional hazard models predicted disparities in receipt of an adequate examination and the effect of an inadequate exam on mortality and disparities. Patients with stage I and II colon cancers undergoing surgery in California were included; patients with stage III and IV disease were excluded.
A total of 37,911 records were analyzed. Adequate staging occurred in fewer than half of cases. An inadequate examination (<12 LNs) was associated with higher mortality rates. Hispanics had the lowest odds of receiving an adequate exam; however, blacks, not Hispanics, had the highest risk of mortality compared with whites. This disparity was not completely explained by inadequate LN examination.
In 1999, after reviewing the available literature, a multidisciplinary panel of experts in colon cancer concluded that examination of at least 12 lymph nodes (LNs) after colon cancer resection was associated with a survival advantage.1 Several studies conducted since that time have reported a notable increase in compliance with this evidence-based recommendation starting in 2000.2-7 Although direct causation between LN counts and survival has never been established, some have suggested that failure to meet the evidence-based standard may result in understaging of cancers. In this scenario, patients would be erroneously assigned to stage I or II when a more complete LN examination might have more accurately diagnosed stage III disease. Chemotherapy would never be recommended to patients who are inadvertently understaged, and failure to receive chemotherapy in stage III disease is now known to decrease long-term survival.8-10
Racial and ethnic disparities in colorectal cancer have been well documented in the literature.11-22 The explanations for disparities are thought to be multifactorial.23-25 Some investigators have suggested that there may be variation in the quality of care provided to minorities that can negatively impact survival.26-28 For example, there is evidence to suggest that African American patients are less likely to receive adjuvant therapy than white patients with rectal cancer.22 Two recent studies provide evidence that there are racial and ethnic disparities in the receipt of the recommended 12-LN examination after colon cancer resection. Both studies analyzed Surveillance, Epidemiology, and End Results (SEER) data, with or without linkage to Medicare claims data, and showed a significantly lower mean number of LNs examined in black6 and Hispanic29 patients as compared with their white counterparts with stage I to III colon cancer.6 Although the findings suggest that minorities are less likely to receive evidence-based cancer care, the studies are limited in their ability to demonstrate the impact of the findings.
First, both studies included patients with stage III disease. These patients are LN positive and will therefore receive a recommendation for chemotherapy regardless of the total number of LNs examined. By contrast, patients with early stage disease are LN negative and depend on an adequate examination to avoid the consequences of understaging. The 2 studies also used the absolute LN count as the outcome of interest despite strong published evidence to show that the LN ratio, which is the number of positive nodes divided by the total number examined,30-32 is a more robust predictor of survival in patients with stage III disease. Finally, neither study tested the association between disparities in processes of care and disparities in outcomes.
The importance of understanding the correlation between process quality and outcome quality should not be underestimated. If disparities in quality of care correlate with disparities in outcome, then quality improvement efforts can be appropriately targeted to increase health equity. The current investigation analyzes a large series of ethnically diverse patients with LN-negative (stage I and II) colon cancer, treated in California hospitals over an 11-year period. The goal of the study is to determine whether there are disparities in the receipt of an adequate LN examination by race and to correlate the variation in care with disparities in outcomes. The primary hypothesis is that there are racial and ethnic disparities in the receipt of an adequate LN examination, and that this variation from evidence-based care will account for some of the excess mortality observed by certain minority groups after colon cancer treatment.
MATERIALS AND METHODS
Sources of Data
Data for the study were obtained from a large state, all-payer cancer database comprised of a linkage between the California Cancer Registry and the patient discharge abstracts from the California Office of Statewide Health Planning and Development Patient Discharge Database. The linked data set contains information on all patients with stage I and II colon cancer diagnosed and treated in California (1996-2006). The California Cancer Registry is a statewide database containing specific clinical, demographic, and treatment details about all cancers treated in the state. By legislative mandate, all providers treating patients with a primary diagnosis of cancer are required to report the care to the registry. The registry has a very low lost to follow-up rate, because reporting occurs regardless of the treatment modality and whether the patient has received care at multiple facilities within the state. At the time of data disclosure, 5-year outcomes were complete through 2008. California Cancer Registry data are very complete; for example, <3% of patients are missing race data. Demographic and clinical variables contained in the set include age, sex, race/ethnicity, date of diagnosis, date and cause of death, tumor grade, American Joint Committee on Cancer consolidated tumor stage, number of LNs examined, number of positive LNs, census block group socioeconomic status (SES) variables (income, education, employment), and a validated SES composite score.33, 34
The Office of Statewide Health Planning and Development Patient Discharge Database contains records for every discharge from a general acute, nonfederal facility in California. Each record contains International Classification of Diseases 9th Edition, Clinical Modification (ICD-9-CM) codes indicating the primary diagnosis for the index admission and coding for the primary procedure performed during the hospitalization. The database also includes up to 24 secondary diagnoses and an indicator for whether the condition was present on admission. This allows for the distinction of comorbid disease from hospital-acquired conditions and facilitates risk adjustment for mortality models. The patient's insurance type, type of admission (emergent vs nonemergent), disposition at the end of the hospitalization, and location of care are also contained in the database.
Records from the California Cancer Registry were linked to the Office of Statewide Health Planning and Development Patient Discharge Database by the California Cancer Registry staff using a probabilistic linkage algorithm based on the patient's date of birth and Social Security number (SSN). Colon cancer diagnoses reported from 1996 to 2006 were linked to discharge abstracts from the same time interval. Protected health information used for the linkage (including patient name, date of birth, and SSN) were stripped from the records before disclosure to the investigators.
ICD-9-CM codes were used to identify records of patients with a principal diagnosis of colon cancer (ICD-9 codes: 153.xx), excluding cancers of the appendix. Records with concomitant principal procedure coding corresponding to major colon resection (45.7, 45.8) during the index hospitalization were included. Only patients with stage I and II disease, who were diagnosed and treated in an inpatient setting between 1996 and 2006, were included. Subjects who had a primary diagnosis of colon cancer but no coding for surgical therapy were excluded, as there would be no record of LN retrieval or examination. Subjects with stage III and IV disease were excluded, because treatment recommendations for these patients would not be affected by the adequacy of LN examination. Subjects with missing sex or Alaskan Native/American Indian race were excluded because of small cell size in accordance with the California Cancer Registry data user agreement.
The primary outcome of the current study is adequacy of LN examination in early stage colon cancer by race/ethnicity. An adequate exam is defined by documentation as a total of 12 or more LNs examined after resection in patients who had zero positive nodes identified (stage I and II disease). An inadequate exam is defined as <12 nodes examined in the same cohort. The secondary outcome measured is the effect of an adequate nodal exam on disparities in mortality.
Bivariate comparisons and multivariate logistic regression modeling were performed using SAS 9.1 (SAS Institute, Cary, NC). Pearson chi-square was used for bivariate comparisons between race and receipt of an inadequate examination. Multivariate models predicting the odds of receiving an inadequate examination were constructed using forward selection methods and compared associations between patient demographics (age, sex, race), clinically relevant factors (comorbidity score,35 stage of disease at diagnosis, tumor location), and year of admission (model I). The models were adjusted for year of admission to account for known differences in the rate of uptake of evidence-based recommendations across various types of hospitals settings. Models predicting receipt of an adequate exam were further adjusted by SES quintile36 (model II). A baseline Cox proportional hazard model (Cox model I) was constructed to predict the probability of mortality at 5 years and was adjusted for the patient demographics, clinically relevant factors, and year of surgery as described above. Subsequent Cox models were adjusted for adequacy of LN examination (Cox model II) and SES quintile (Cox model III) given the known association with mortality. All tests of significance were 2 tailed. Those with confidence intervals excluding 1 or those with P values <.05 were considered statistically significant.
There were 47,113 cases with a primary diagnosis of stage I or II disease and concomitant coding for a major surgical resection between 1996 and 2006. Of these, 2653 cases were excluded for missing data on sex, race/ethnicity, or stage. Another 6549 records were excluded because of missing or inconsistent information about LN counts or the number of positive nodes. A remaining 37,911 records were retained for analysis.
Patient Demographic and Clinical Characteristics and LN Examination
The cohort demographics as distributed by adequacy of LN examination are depicted in Table 1. Regardless of individual characteristics, fewer than half the patients with stage I or II disease received an adequate examination (43%). Male sex and increasing age were associated with lower rates of adequate examination. Rates of meeting the standard were poor across all racial/ethnic groups; however, the difference between the proportion of black and Hispanic subjects who received an adequate exam and those who did not (17% and 19%, respectively) is larger than the gap for white (14%) and Asian/Pacific Islander populations (16%). Those with lower comorbidity scores and those in higher SES categories had higher rates of adequate exam than those with more comorbidities and lower SES score. Tumors located on the right side of the colon were more often associated with an adequate exam as compared with left-sided tumors. There was no difference between subjects undergoing emergent operations and those undergoing nonemergent procedures. The highest proportion of patients who failed to receive an adequate exam had stage I disease (Fig. 1).
Table 1. Demographics and Clinical Characteristics of All Patients
In multivariate logistic regression models, there was a disparity in meeting the 12-LN recommendation by race/ethnicity. In the baseline model, which was adjusted for patient demographic and clinical factors but not SES, there was no significant difference between blacks, whites, and Asian/Pacific Islander populations in the receipt of an adequate exam (Table 2, model I). Hispanics, however, were >20% less likely to receive an adequate exam (odds ratio [OR], 0.79; confidence interval [CI], 0.73-0.84) (Table 2, model I). Adjusting the model for SES quintile revealed an independent and positive association between SES and receipt of an adequate LN exam (ORhi SES, 1.19; CI, 1.09-1.28); however, the negative association between Hispanic ethnicity and adequate exam persisted (Table 2, model I).
Table 2. Odds of Adequate Staging (≥12 Nodes) by Race/Ethnicity
Patient Characteristic, N=37,911
Model I Patient Characteristics, OR (CI)
Model II Patient Characteristics With Adjustment for SES, OR (CI)
Stage I and II colon cancers, California, 1996 to 2006. Model I is adjusted for sex, age, race, Charlson comorbidity score, location of tumor, type of admission, and year of surgery. Model II is additionally adjusted for SES quintile.
Inadequate LN Exam and Overall Mortality
Subjects with stage I and II disease who received an inadequate LN examination had higher unadjusted mortality rates at 1 and 5 years of follow-up (Fig. 2). This association persisted in multivariate Cox proportional hazard regression models. Inadequate LN examination was associated with a 16% risk-adjusted increased probability of mortality at 5 years (P < .0001) (Table 3, Cox model II).
Table 3. Cox Proportional Hazard Models Predicting 5-Year Mortality Risk After Colon Cancer Surgery
Patient Characteristics, N=37,911
Model I: Mortality by Race, HR (SE)
Model II: Patient Characteristics With Adjustment for Inadequate Staging, HR (SE)
Model III: Patient Characteristics With Adjustment for SES, HR (SE)
Abbreviations: API, Asian/Pacific Islander; HR, hazard ratio; SE, standard error; SES, socioeconomic status.
Stage I and II colon cancers, California, 1996 to 2006. Model I is adjusted for patient characteristics (sex, age, race, Charlson comorbidity score, location of tumor, type of admission, year of surgery). Model II is adjusted for patient characteristics and adequacy of lymph node examination. Model III is adjusted for patient characteristics and SES quintile assignment.
Black vs white
Hispanic vs white
API vs white
II vs I
III vs I
IV vs I
V [high] vs I
Disparities in Mortality
Baseline differences in mortality risk by race/ethnicity are shown in Table 3 Cox model I. The baseline model is adjusted for individual covariates only. Before adjustment for LN exam performance, blacks had the highest hazard of mortality (hazard ratio [HR], 1.09; P = .043) as compared with whites. Hispanics and Asian/Pacific Islander populations appear to have better 5-year survival relative to whites. Adjusting the models to account for adequacy of LN exam reveals a small effect toward reducing disparities in mortality for black patients (HR, 1.08; P = .057). There is also a small increase in the survival advantage for Hispanics, but little or no effect on the estimates for Asian/Pacific Islander populations (Table 3, Cox model II). Adjustment for increasing SES quintile (Table 3, Cox model III) completely eliminates the black/white disparity, reducing the hazard of mortality to a level comparable to whites (HR, 1.07; P = .37).
The current study aimed to determine the presence of disparities in receipt of an adequate LN examination after resection for early stage colon cancer and further to assess the effects of an inadequate examination on disparities in mortality. The results support a strong correlation between inadequate examination (<12 LNs examined) and an increase in both adjusted and unadjusted mortality rates. Although there was a significant difference in the quality of LN examination by race/ethnicity, in particular for Hispanic patients, there was only a small correlation between inadequate examination and racial/ethnic disparities in mortality that were observed in the African American subgroup.
The low rate at which patients with early stage colon cancer meet the minimum standard was not surprising and has been documented in the work of others.6, 37 Our study does not clarify why the overall rate is <50%. No large studies have been conducted to determine the reason for this broad failure to comply with evidence-based standards. Smaller studies have suggested a lack of staff to examine 12 LNs in every specimen,38, 39 whereas others have suggested that this is a function of surgical technique.40 Regardless of where the responsibility lies, the current study raises the importance of actively pursuing an adequate examination to improve overall outcomes. The independent and strong correlation of adequate examination with both unadjusted and adjusted mortality is consistent with others,4, 37, 41, 42 and underscores the recommendations of the College of American Pathologists consensus statement, which recommended that if <12 nodes are found during an examination, additional techniques to identify nodes should be considered.1
The finding of racial/ethnic disparities in receipt of an adequate examination is also not surprising.6, 29 The current results fit into a body of work that shows a consistent lack of evidence-based care delivered to minority populations. In benign disease, others have observed disparities in appropriate use of percutaneous coronary interventions,43, 44 whereas some have shown similar trends in the field of urogynecology and the inappropriate use of hysterectomy.45, 46 Most recently, in colon cancer, McBride et al6 and Cone et al29 demonstrated disparities in the receipt of adequate LN examination. Similar to Cone and colleagues' work, the current study found a significant difference in receipt of an adequate examination for Hispanic patients. However, there was no evidence for a black/white disparity in receipt of adequate examination. Our study further departs from the previous investigations in that adjustment for SES factors did not neutralize the negative effect of Hispanic ethnicity on adequacy of LN exam. One explanation for this difference may be that the SES measures in the current study are derived from census block group data, which are superior to county level data used in other studies. Despite these differences, our study advances the literature by attempting to specifically correlate disparities in process of care for early stage colon cancer with disparities in outcomes of care.
Despite the apparent similarities to previous work, our approach to this analysis is quite novel. We are the first to focus on early stage disease and disparities, which provides an opportunity to intervene on behalf of patients who have the most to gain from the receipt of evidence-based care. If the first 3 LNs in an examination are positive, stopping the exam and concluding that the patient has stage III disease would be accurate, and there would be little theoretical risk of the patient failing to receive recommendations for chemotherapy. Prior studies could not estimate the risk associated with being understaged, because patients with stage III disease who had an inadequate examination were correctly triaged for care despite failure to meet the minimum standard. The concept of the LN ratio attempts to address this issue and raises questions about using the 12-LN denominator in isolation as a measure of quality.30, 32 However, LN ratio is also somewhat limited, because it is only predictive of prognosis in LN-positive (stage III) disease. Our approach incorporates the clinical reality that examining 12 LNs is only imperative if no nodes have been found to be positive (ie, stage I and II disease). Once a single positive LN is identified, treatment recommendations are set (chemotherapy will be recommended). The current study focuses on early stage disease largely because the risk associated with inadequate staging is most relevant in this group.
The current investigation is also novel in that we made an explicit attempt to correlate disparities in process quality with disparities in outcome. No prior studies have tested this association, nor have previous studies compared the effect of clinical care with SES factors simultaneously. Therefore, the findings are novel and inform our understanding of the drivers of colon cancer mortality disparities. Although the current findings identified racial/ethnic disparities in receipt of evidence-based care and a strong correlation between an inadequate exam and mortality, adjustment for an inadequate exam only accounted for a small component of the survival disparity. By contrast, adjustment for SES had a much greater effect on mortality than clinical care factors. These findings support others who have suggested that SES may be a stronger predictor than race/ethnicity in explaining disparities in outcome for cancer and other medical conditions.47-51 There is face validity to this assertion, because the social conditions into which the patient is discharged—including employment, income, and marital status/familial support—may have a direct effect on the ability to obtain care, including cancer follow-up or further treatment.
The current study is a retrospective analysis of administrative data. As a cross-sectional study, our results do not support a causal link between LN examination quality and mortality. It is still unclear why examination of 12 LNs is associated with improved survival. It has been suggested that it is not the examination itself, but rather the potential to achieve accurate staging and eventually obtain treatment appropriate to stage. Alternative explanations include patient or tumor factors that may predispose to LNs that are smaller or more difficult to find, such as a poor immune response to tumor or a tumor with low immunologic antigenicity. Nonetheless, no prior investigators have demonstrated direct causation, and the current study falls in line with others who have simply demonstrated an association.
This study used only California data. Although this may limit generalizability, it is important to note that California is a large and very diverse state, and the data set contains data from all payers. The diversity of payer and patient characteristics in this data set represents an advantage over other data sets that suffer limitations related to patients' age and insurance type, as in the case of Medicare-linked SEER data. Inclusion of all patient ages is particularly relevant for the study of disparities, because some racial and ethnic populations observe higher colon cancer rates at younger ages.52 To corroborate the current findings, future research should involve analysis of national data without restriction on age at the time of diagnosis and treatment.
Finally, although the models were adjusted for patient factors, which can affect LN counts as well as mortality, they were not controlled for the type of surgeon performing the resection. Studies have shown that variation in LN counts is associated with surgeon's training.53 It is possible that the difference in the quality of care by race may be related to use of nonfellowship-trained surgeons; however, there is no reliable way to track this in the current data set. Future investigations should explore the correlation between the type of surgeon and hospital-level characteristics where minorities cluster for care and disparities in quality of care. Most importantly, if disparities in care are identified, they must be tested for their associations with mortality by race.
Despite the limitations, the current findings show that there is widespread failure to meet the 12-LN minimum examination after colon cancer resection in California. Although we did identify a disparity in receiving an adequate exam for some groups, adjusting mortality models for this clinical factor had only a small effect on disparities. This is in contrast to the independent and robust effect of SES. Nonetheless, the results of this study inform our understanding of how this process of care relates to early stage colon cancer mortality and disparities. Therefore, although failing to meet the standard predicts a significantly increased risk of mortality in all populations, our results suggest that initiatives to increase compliance with this evidence-based process of care are unlikely to have a significant impact on disparities.
K.F.R. was supported by a Harold Amos Medical Faculty Development Award from the Robert Wood Johnson Foundation.