Influence of Socioeconomic Status and Hospital Type on Disparities of Lymph Node Evaluation in Colon Cancer Patients

Abstract

BACKGROUND

A minimum of 12 dissected lymph nodes (LNs) has been recommended as a consensus guideline for resections in colon cancer patients. This study assessed the influence of both socioeconomic status (SES) and hospital type on compliance with this colon LN dissection guideline and examined the time trend for ≥12 LNs dissected.

METHODS

Stage I to III incident colon cancer cases diagnosed from 1996 to 2007 were obtained from the Louisiana Tumor Registry. A composite census tract-level SES score was created to serve as a surrogate for individual-level SES. Hospitals performing colon resections were categorized into 5 groups according to the Commission on Cancer Accreditation Program. Multiple logistic regression analyses were used.

RESULTS

Of 10,460 colon cancer cases diagnosed during the study period, 43.9% had ≥12 LNs dissected. Patients residing in less affluent SES areas were less likely to receive a dissection of ≥12 nodes than those residing in more affluent areas. SES was no longer significant after adjusting for race, sex, age, stage, grade, anatomic subsite, diagnosis year, and hospital type. In contrast, hospital type was significantly associated with the number of LNs dissected, even after adjusting for other factors. Patients diagnosed from 2002 to 2007 were twice as likely (95% confidence interval, 1.84–2.17) to have ≥12 LNs dissected than those diagnosed from 1996 to 2001 after adjustment.

CONCLUSIONS

In Louisiana, hospital type is an independent significant predictor of adequate LN evaluation for colon cancer. Training and education are needed to reduce this disparity in the facilities with consistently lower LN yield in their dissections.

INTRODUCTION

The status of lymph node (LN) involvement is an important determinant for properly staging colon cancer. A minimum of 12 LNs dissected and examined has been recommended as a guideline for colon and rectal cancer surgery for more than a decade.^1–3 Data from various sources have suggested that the dissection of a sufficient number of LNs results in more accurate cancer staging^4–6 as well as improved survival.^7–10 Despite all supporting evidence that evaluating an adequate number of LNs is beneficial for colon cancer patients, data from the Surveillance, Epidemiology, and End Results (SEER) program for the period 1988 to 2001 showed that 63% of patients with locoregional colon cancer had <12 LNs dissected. Although an increasing trend of adequate LN dissections was observed during the study period, only 44% of colon cancer patients had ≥12 LNs dissected in the year 2001.¹¹ In 2007, the dissection of ≥12 LNs was endorsed by the National Quality Forum through joint collaboration with the standard setters as a quality of care measure for resected colon cancer.¹²

Demographic and clinical variables such as age, sex, tumor stage, anatomic subsite, tumor grade, and diagnosis year have been significantly associated with the number of LNs examined among colon cancer patients.^11,13 Socioeconomic status (SES) was also recognized as an important factor contributing to the disparities in LN evaluation.¹³ In addition, Bilimoria et al reported that patients who received colon resections at academic institutions, National Cancer Institute (NCI)-designated Comprehensive Cancer Centers, or high-volume hospitals were more likely to have an adequate number of LNs examined.¹⁴ These studies, however, either did not assess the associations of both SES and hospital characteristics or used large geographic areas (such as county level) that are not sufficiently socioeconomically homogenous to classify SES characteristics.¹⁵

The objectives of our study were to assess the influence of both census tract-level SES and hospital type on the disparities of LN dissections in Louisiana as well as to examine the time trends for having an adequate number (≥12) of LNs dissected. SES influences access to care and choice of facility, whereas the hospital type impacts the practice and delivery of care and treatment strategies. Louisiana has a large proportion of African Americans (about 30%), and its population, regardless of race and ethnicity, is also socioeconomically disadvantaged as measured by poverty level, income, education, and other indicators. In addition, Louisiana has a public hospital network that provides free health care to the indigent, uninsured, and medically underserved residents. The large public hospitals also serve as teaching hospitals to 3 medical schools in the state; therefore providing a unique setting to evaluate LNs dissected for colon cancer cases.

MATERIALS AND METHODS

Patient Selection

Colon cancer cases were obtained from the Louisiana Tumor Registry, a statewide population-based cancer surveillance system, a registry of the NCI SEER program, and the Centers for Disease Control and Prevention's National Program of Cancer Registries. All first primary invasive colon cancer cases (International Classification of Diseases for Oncology 3rd edition site codes: C18.0–C18.9)¹⁶ diagnosed between 1996 and 2007 among those aged 20 years or older were included in the study. This study was restricted to microscopically confirmed stage I to III resected colon cancer. Familial adenomatous polyposis, carcinoids, sarcomas, and lymphomas of the colon were excluded. We further excluded colon cancer cases with non-specified or unknown number of LNs examined (358 cases), race other than white and black (74 cases), and census tract based only on post office box addresses (814 cases).

Description of Variables

The outcome variable, number of LNs dissected, was grouped as <12 nodes or ≥12 nodes based on the recommendation of consensus guidelines.^1–3,12 The independent variables of interests were census tract-level SES and hospital type. Hospitals, where colon resections were performed, were categorized according to the definition of the American College of Surgeon's Commission on Cancer Accreditation Program¹⁷: teaching hospital cancer programs, community hospital comprehensive cancer programs, community hospital cancer programs, public hospitals, and others (non-Commission on Cancer/nonpublic hospitals including surgical clinics). Public hospitals include 7 non–Commission on Cancer-accredited public hospitals, 3 Veterans Affairs hospitals, and 2 military base hospitals.

Because individual SES information is not available in medical records, the primary source of the registry data, we geocoded patients' street addresses at diagnosis to census tract and developed a census tract-level SES score as a surrogate for individual-level SES, using a similar approach previously published.^18,19 Three socioeconomic measures at the census tract level were obtained from the 2000 Census; these include: (1) education—the proportion of population aged ≥25 years with less than a high school educational attainment; (2) poverty—the proportion of total population with income in 1999 below the federal poverty level; and (3) income—median household income in 1999. Because these measures of SES status are highly correlated, a principal component analysis²⁰ was carried out (previous arcsine transformation of education and poverty, and logarithm transformation of income) to obtain an SES index. Louisiana has 1106 census tracts; of these 1074 had at least 1 colon cancer patient in this study. The first principal component explained 87.8% of the variability in the 3 SES measures and therefore sufficed as a composite score of SES. This component is a weighted average with slightly larger weight for income (−0.665) than for poverty (0.584) and for education (0.464). Lower scores of the first principal component mean higher SES. Census tracts were then grouped into quintiles of the first component scores, and each patient was assigned to an SES quintile according to the tract where they lived at the time of diagnosis. The first quintile represented the most affluent SES group, and the fifth quintile represented the least affluent (most deprived). Patients with census tract coded to Zip Code only were assigned randomly to tracts within their counties.

Independent clinical variables were American Joint Committee on Cancer (AJCC) stage (I–III), anatomic subsite, histological grade, and year of diagnosis. Because different staging systems were used to stage cancer cases between 1996 and 2007, we converted stage at diagnosis to AJCC sixth edition²¹ to have comparable stage coding across these systems as well as to resolve cases with unknown AJCC stage. For cases diagnosed between 1996 and 2003, staging was converted from SEER Extent of Disease²² to the sixth edition AJCC stage, and for cases diagnosed between 2004 and 2007, the sixth edition AJCC stage was derived from the Collaborative Staging System. Discrepancies between the converted or derived AJCC stage and the originally coded AJCC TNM stage were manually reviewed and resolved. Anatomic subsites included right colon (C180–C183), transverse colon (C184), left colon (C185–C187), and overlapping or colon not other specified (C188–C189); histological grade included well differentiated, moderately differentiated, poorly differentiated/undifferentiated/anaplastic, and unknown. The year of diagnosis was grouped into 2 periods: 1996 to 2001 and 2002 to 2007. Patient demographic variables included in the analysis were race (white and black), sex (male and female), and age (≤49, 50–59, 60–69, 70–79, and ≥80 years).

Statistical Analysis

Distribution of cases by demographics (race, sex, and age), clinical variables (AJCC stage, histological grade, colon subsite, year of diagnosis, number of LNs examined), and hospital type were compared by SES, using the chi-square test. Univariate logistic regression was performed to assess the unadjusted association of demographic variables, clinical variables, SES group, and hospital type with number of LNs dissected (<12 and ≥12). Multiple logistic regression analyses were carried out to estimate adjusted associations and to identify statistically significant predictors. Identification of significant predictors was based on a stepwise selection procedure with entry and stay significance levels of .05; only 2-factor interactions were considered. The proportions of cases with nodal dissection ≥12 over time were assessed with the Cochran-Armitage test for linear trend. All analyses were carried out using SAS version 9.2 (SAS Institute, Cary, NC).

RESULTS

Of the 10,460 eligible colon cancer cases, the majority were white (73.0%), 60 years or older (75.0%), and with moderately differentiated grade tumor (68.4%) (Table 1). More than 50% of these patients had <12 LNs dissected, and 4% had no nodes removed at all. The median number of LNs dissected was 10 (interquartile range, 6–16). About 40% of patients received colon resections at non-Commission on Cancer/nonpublic hospitals and only 6.4% at nonteaching public hospitals.

Table 1.

Demographic and Clinical Characteristics of Stage I-III Colon Cancer Patients by SES Level, Louisiana, 1996–2007

			SES Level (Quintile)b
	ALLa		1st (Most Affluent)		2nd		3rd		4th		5th (Least Affluent)
	N	%	N	%	N	%	N	%	N	%	N	%	P c
Total	10460	100.0	2330	22.3	2357	22.5	2086	19.9	1933	18.5	1754	16.8
Race													<0.0001
White	7635	73.0	2130	27.9	2015	26.4	1665	21.8	1161	15.2	664	8.7
Black	2825	27.0	200	7.1	342	12.1	421	14.9	772	27.3	1090	38.6
Sex													0.0046
Male	5095	48.7	1159	22.7	1132	22.2	1058	20.8	957	18.8	789	15.5
Female	5365	51.3	1171	21.8	1225	22.8	1028	19.2	976	18.2	965	18.0
Age at diagnosis													0.6178
20–49	868	8.3	196	22.6	216	24.9	161	18.5	157	18.1	138	15.9
50–59	1742	16.7	381	21.9	380	21.8	331	19.0	346	19.9	304	17.5
60–69	2589	24.8	562	21.7	569	22.0	525	20.3	482	18.6	451	17.4
70–79	3104	29.7	705	22.7	708	22.8	650	20.9	541	17.4	500	16.1
≥ 80	2157	20.6	486	22.5	484	22.4	419	19.4	407	18.9	361	16.7
AJCC stage													0.0024
Stage I	2529	24.2	607	24.0	596	23.6	503	19.9	435	17.2	388	15.3
Stage II	4238	40.5	906	21.4	920	21.7	888	21.0	826	19.5	698	16.5
Stage III	3693	35.3	817	22.1	841	22.8	695	18.8	672	18.2	668	18.1
Histological grade													0.1158
Well Differentiated	707	6.8	139	19.7	149	21.1	154	21.8	135	19.1	130	18.4
Moderately Differentiated	7158	68.4	1557	21.8	1610	22.5	1442	20.1	1336	18.7	1213	16.9
Poorly or Undifferentiated	2043	19.5	490	24.0	470	23.0	392	19.2	360	17.6	331	16.2
Unknown	552	5.3	144	26.1	128	23.2	98	17.8	102	18.5	80	14.5
Subsite													0.0168
Right	4926	47.1	1170	23.8	1137	23.1	945	19.2	901	18.3	773	15.7
Transverse	943	9.0	183	19.4	202	21.4	204	21.6	177	18.8	177	18.8
Left	4408	42.1	941	21.3	973	22.1	897	20.3	827	18.8	770	17.5
Overlapping or Unknown	183	1.8	36	19.7	45	24.6	40	21.9	28	15.3	34	18.6
Diagnosis year													0.0121
1996–2001	5094	48.7	1147	22.5	1181	23.2	965	18.9	908	17.8	893	17.5
2002–2007	5366	51.3	1183	22.0	1176	21.9	1121	20.9	1025	19.1	861	16.0
Nodes examined													0.0239
0–11	5865	56.1	1251	21.3	1304	22.2	1177	20.1	1129	19.2	1004	17.1
≥ 12	4595	43.9	1079	23.5	1053	22.9	909	19.8	804	17.5	750	16.3
Hospital type													<0.0001
THCP	1011	9.7	189	18.7	205	20.3	184	18.2	182	18.0	251	24.8
COMP	2985	28.5	1097	36.8	747	25.0	391	13.1	421	14.1	329	11.0
CHCP	1665	15.9	323	19.4	361	21.7	412	24.7	304	18.3	265	15.9
Public	664	6.4	62	9.3	121	18.2	144	21.7	167	25.2	170	25.6
Non-CoC/Non-public	4135	39.5	659	15.9	923	22.3	955	23.1	859	20.8	739	17.9

Abbreviations: SES, socioeconomic status; THCP, teaching hospital cancer program; COMP, community hospital comprehensive cancer program; CHCP, community hospital cancer program; CoC, Commission on Cancer.

^aPercentage presented in column.

^bPercentage presented in row.

^cP value chi-square test for association.

Colon cancer patients were distributed about evenly throughout the 5 SES levels, with approximately 45% of the patients residing in the 2 most affluent SES areas. All demographic and clinical factors except age and histological grade were significantly associated with SES. More than half (54.3%) of the white patients lived in the 2 most affluent areas (first and second quintiles) compared with about ⅕ of black patients (19.2%). Female patients were more likely to live in the less affluent SES areas than male patients. In general, patients residing in more affluent SES areas were more likely to be diagnosed with early stage disease and with right side colon tumors. The type of hospital performing the surgery was highly associated with SES group (P < .0001), with about half of patients in the most affluent area receiving colon resections at community hospital comprehensive cancer programs compared with only 19% of patients in the least affluent SES area (Table 1). In contrast, patients residing in the least affluent SES area were more likely to undergo colon resections at teaching hospital cancer programs and public hospitals.

Whereas race and sex were not significantly associated with adequate number LNs being dissected, younger patients, patients with advanced stage, tumors with poorly or undifferentiated histological grade, or cancer in the right colon tended to have a higher proportion of ≥12 LNs dissected than their counterparts (Table 2). Compared with colon resections performed in the earlier time period (1996–2001), those performed in the later time period (2002–2007) were 89% more likely to have ≥12 LNs dissected. Patients residing in more affluent areas were more likely to receive adequate nodal dissections (≥12 nodes). Hospital type was also highly associated with the number of LNs dissected. Patients who received colon resections at public hospitals or non-Commission on Cancer/nonpublic hospitals were about half as likely to have ≥12 LNs dissected as those treated at teaching hospital cancer programs or community hospital comprehensive cancer programs.

Table 2.

Distributions and Odds Ratios (Unadjusted and Adjusted) of Number of Nodes Dissected for Stage I-III Colon Cancer Patients, Louisiana, 1996–2007

	Nodes Examined
	0–11		12+		Unadjusted		Adjusteda
	N	%	N	%	Odds Ratio	(95% CI)b	Odds Ratio	(95%CI)
Race
White	4267	55.9	3368	44.1	1.00	(referent)	1.00	(referent)
Black	1598	56.6	1227	43.4	0.97	(0.89–1.06)	0.87	(0.78–0.96)
Sex
Male	2894	56.8	2201	43.2	1.00	(referent)	1.00	(referent)
Female	2971	55.4	2394	44.6	1.06	(0.98–1.15)	1.06	(0.97–1.15)
Age at diagnosis
20–49	392	45.2	476	54.8	1.00	(referent)	1.00	(referent)
50–59	893	51.3	849	48.7	0.78	(0.67–0.92)	0.77	(0.65–0.92)
60–69	1436	55.5	1153	44.5	0.66	(0.57–0.77)	0.62	(0.53–0.73)
70–79	1814	58.4	1290	41.6	0.59	(0.50–0.68)	0.52	(0.44–0.61)
≥ 80	1330	61.7	827	38.3	0.51	(0.44–0.60)	0.41	(0.35–0.49)
AJCC stage
Stage I	1776	70.2	753	29.8	1.00	(referent)	1.00	(referent)
Stage II	2290	54.0	1948	46.0	2.01	(1.81–2.23)	1.99	(1.78–2.22)
Stage III	1799	48.7	1894	51.3	2.48	(2.23–2.76)	2.32	(2.07–2.61)
Histological grade
Well Differentiated	487	68.9	220	31.1	1.00	(referent)	1.00	(referent)
Moderately Differentiated	4034	56.4	3124	43.6	1.71	(1.45–2.02)	1.54	(1.29–1.84)
Poorly or Undifferentiated	945	46.3	1098	53.7	2.57	(2.15–3.08)	1.92	(1.57–2.33)
Unknown	399	72.3	153	27.7	0.85	(0.67–1.09)	0.93	(0.72–1.21)
Subsite
Right	2309	46.9	2617	53.1	1.00	(referent)	1.00	(referent)
Transverse	529	56.1	414	43.9	0.69	(0.60–0.79)	0.65	(0.56–0.75)
Left	2921	66.3	1487	33.7	0.45	(0.41–0.49)	0.43	(0.39–0.47)
Overlapping or Unknown	106	57.9	77	42.1	0.64	(0.48–0.86)	0.60	(0.44–0.83)
Diagnosis year
1996–2001	3261	64.0	1833	36.0	1.00	(referent)	1.00	(referent)
2002–2007	2604	48.5	2762	51.5	1.89	(1.75–2.04)	1.99	(1.84–2.17)
SES level
1st (Most affluent)	1251	53.7	1079	46.3	1.00	(referent)	1.00	(referent)
2nd	1304	55.3	1053	44.7	0.94	(0.84–1.05)	1.03	(0.91–1.16)
3rd	1177	56.4	909	43.6	0.90	(0.80–1.01)	1.06	(0.93–1.20)
4th	1129	58.4	804	41.6	0.83	(0.73–q0.93)	0.96	(0.84–1.10)
5th (Least affluent)	1004	57.2	750	42.8	0.87	(0.76–0.98)	1.06	(0.92–1.23)
Hospital type
THCP	495	49.0	516	51.0	1.00	(referent)	1.00	(referent)
COMP	1449	48.5	1536	51.5	1.02	(0.88–1.17)	0.98	(0.84–1.14)
CHCP	920	55.3	745	44.7	0.78	(0.66–0.91)	0.69	(0.59–0.82)
Public	408	61.4	256	38.6	0.60	(0.49–0.73)	0.54	(0.43–0.66)
Non-CoC/Non-public	2593	62.7	1542	37.3	0.57	(0.50–0.66)	0.54	(0.46–0.63)

Abbreviations: SES, socioeconomic status; THCP, teaching hospital cancer program; COMP, community hospital comprehensive cancer program; CHCP, community hospital cancer program; CoC, Commission on Cancer.

^aAdjusted for all independent variables listed on table.

^b95% Confidence interval.

After adjustment for all independent variables in the model, the significance of the associations remained unchanged except for race and SES (Table 2). Adjusted odds ratios (ORs) for those significant independent variables were similar to the unadjusted ones. Blacks were slightly less likely than whites (OR, 0.87; 95% confidence interval, 0.78–0.96) to have adequate LNs dissected. SES was no longer significant after adjusting for other predictors (P = .5446), and sex remained not significant (P = .2041). The final model included race, age at diagnosis, AJCC stage, histological grade, anatomic subsite, year of diagnosis, hospital type, and 3 significant interaction terms: AJCC stage and histological grade (P = .0033), AJCC stage and anatomic subsite (P = .0007), and hospital type and diagnosis year group (P = .0353). Regarding the interaction of stage by grade, the odds of having ≥12 LNs dissected for patients with stage III disease were about the same among the 3 tumor histological grades (except unknown), but for patients with stage I and II disease, the OR increased from well-differentiated to poorly differentiated tumor. For interaction of cancer stage by anatomic subsite, the decreasing odds of having ≥12 LNs dissected from right colon to left colon was steeper for stage I than more advanced stages. As for hospital type by diagnosis time period, steady decreasing odds of ≥12 LNs dissected were observed from teaching hospital cancer programs to non-Commission on Cancer/nonpublic hospitals in the earlier time period, but mixed in the later period.

There was an overall increasing trend of ≥12 LNs dissected over time (Fig. 1). The trend was leveled between 1996 and 2001, followed by a gradual increase between 2002 and 2005, and a sharp increase after 2005. We observed a similar pattern by hospital type until 2003; thereafter, a substantial increase occurred except for teaching hospital cancer programs in 2005 and 2006 (Fig. 2). All these trend analyses are highly significant, with P < .001.

Figure 1.

Overall trend of proportion of ≥12 lymph nodes (LNs) dissected for stage I to III colon cancer patients in Louisiana from 1996 to 2007 is shown.

Figure 2.

Trend of proportion of ≥12 lymph nodes (LNs) dissected for stage I to III colon cancer patients by hospital type in Louisiana from 1996 to 2007 is shown. CHCP, community hospital cancer program; CoC, Commission on Cancer; COMP, community hospital comprehensive cancer program; THCP, teaching hospital cancer program.

When examining the compliance of dissecting ≥12 LNs by hospital type and time period, we observed different patterns for the 2 time periods (1996–2001 and 2002–2007). During the earlier period, there was a steady increase in the ORs of receiving ≥12 LNs dissected by hospital type from public hospital to teaching hospital cancer programs when compared with non-Commission on Cancer/nonpublic hospitals (Fig. 3), with ORs ranging from 1.16 to 2.19, all statistically significant except for public hospitals. In the later period, more uniformity was apparent, and ORs remained significant for community hospital comprehensive cancer programs and teaching hospital cancer programs only.

Figure 3.

Odds ratios of ≥12 lymph nodes dissected among hospital types by diagnosis time period for stage I to III colon cancer patients in Louisiana from 1996 to 2007 are shown. CHCP, community hospital cancer program; CoC, Commission on Cancer; COMP, community hospital comprehensive cancer program; THCP, teaching hospital cancer program.

DISCUSSION

It is well recognized that LN-positive (stage III) colon cancer patients benefit from postsurgery adjuvant chemotherapy,^23–25 particularly for patients with ≥12 LNs removed.²⁵ In Louisiana, the proportion of patients having adequate LNs dissected is lower than all SEER registries combined during the years 1996 to 2007, 44% versus 51%, respectively (SEER public data). Consistent with previous studies using SEER data,^11,13 we found that age at diagnosis, tumor stage, histological grade, and anatomic subsite are highly associated with the number of LNs examined. Overall, the associations were reduced after adjusting for other factors. Our study did not find gender disparity in the number of LNs dissected. Whites were 15% more likely to have ≥12 LNs dissected than blacks after controlling for other demographic variables, clinical variables, hospital type, and SES.

Although SES is associated with adequate LN dissection in univariate analysis, after controlling for race and/or hospital type (regardless of age and clinical variables) SES is no longer significant in our study. These results are inconsistent with recent findings of McBride et al¹³ and could result from differences in methodology for creating the composite SES (ad hoc scoring vs principal components), differences in the selection of geographic level SES (county level vs census tract level), or the high correlation between SES and race and between SES and hospital type in our data. With regard to the last, Louisiana residents of deprived SES can receive free care in public hospitals, some of which are affiliated with medical schools and serve as teaching hospitals, where quality of care and compliance with clinical practice guidelines are emphasized. Thus, to further explore whether SES may be of significance, we carried out a stratified analyses by race, hospital type, and diagnosis time period. After we controlled for other predictors, SES persistently showed no impact on the number of LNs dissected except for the community hospital cancer program hospitals (P = .0366).

According to previous studies, hospital type and volume were highly related to compliance with the extent of LN evaluation for colon, gastric, and pancreatic cancer.^14,26 Our finding is generally comparable to these studies. Patients undergoing colon cancer resections at teaching hospital cancer programs or community hospital comprehensive cancer programs are more likely to have adequate LN dissections than at other facilities. The trend analysis shows an overall increase toward adequate LN dissections over time, especially after 2004, when a sharp increase was noted in most hospital types. This may have resulted from the Commission on Cancer's 2004 requirement that at least 90% of all cancer pathology reports use College of American Pathologists cancer protocols in American College of Surgeons-Commission on Cancer hospitals.²⁷ The decrease in disparities of having ≥12 LNs dissected among the 5 hospital types in the later time period (2002–2007) as compared with the earlier time period (1996–2001) may reflect an increased general awareness of the benefit of adequate LN evaluation accompanied by better compliance and documentation to improve the quality of care in Louisiana.

In addition, we found that the proportion of adequate LN dissections for patients undergoing colon resections in teaching hospital cancer programs decreased considerably in 2005 and 2006, then improved in 2007. This noticeable change is most likely because of the impact of Hurricane Katrina. In Louisiana, the 3 major teaching hospital cancer programs located in the New Orleans area performed about 73% of colon resections in all teaching hospitals in the state. Immediately after Katrina, the shortage in both physicians and pathology laboratory staff might have resulted in either fewer LNs being removed/examined or limited documentation in pathology reports from 2005 and 2006. Although all 3 Veterans Affairs hospitals have Commission on Cancer-approved cancer programs, these facilities provide care mainly to military veterans. In examining the association between hospital type and number of adequate LN dissections, no significant difference was found when grouping them into either the community hospital comprehensive cancer program or the public hospital categories because of the small case counts.

Several strengths should be noted for our study. The primary strength is the use of census tract-level SES, which tends to distinguish more homogenous populations with regard to socioeconomic characteristics than county-level SES. In addition, we created a composite SES score, thus providing a better, more refined surrogate for individual SES. Furthermore, the inclusion of hospital type as a determinant allows for an assessment of whether compliance with treatment guidelines may differ by hospital training programs and/or practice culture.

Our study is subject to a few limitations. The effect of surgeon specialty and volume are not evaluated, because surgeon characteristics are not routinely collected in a population-based cancer registry. Studies have shown that surgeon specialty and volume may affect treatment decisions, as well as a patient's health outcome.^28,29 Hence, surgeon characteristics may also influence the number of LN dissections. In addition, the patient's health insurance coverage is not included in this analysis, because such information is often incomplete in medical charts. The type of health insurance that a patient has, or the lack of insurance, can limit the patient's choice of health care facility. A study conducted on female breast cancer patients found that selection of treatment and choice of hospital were significantly affected by the patient's health insurance coverage.³⁰ The type of health insurance also influenced the treatment selection for colorectal cancer patients.³¹ However, given that Louisiana has a network of 10 public hospitals that offer free health care to residents who have no insurance, we anticipate insurance status would be less impacted. Another limitation is the lack of information in pathology practices. Recent studies have shown that the practice of a pathology laboratory and thoroughness of the pathologist's examination could affect the counts on the number of LNs examined and the finding of positive LNs.^32,33 Reese's study³³ also suggested that a trained pathology assistant can improve the number of LNs harvested from colorectal resection specimens, leading to better quality of pathological evaluation.

In conclusion, we have demonstrated that for Louisiana, hospital type is an independent and more significant determinant than SES for the adequate LN evaluation of colon cancer. The Louisiana public hospital network offers care that reduces health disparities among SES disadvantaged residents. Compliance with colon treatment guidelines regarding the adequate number of LNs dissected in Louisiana has steadily increased over time. Nevertheless, the proportions of adequate LN dissections performed during the 2002 to 2007 time period were still <50% in community hospital cancer programs, public hospitals, and non-Commission on Cancer/non-public hospitals. Therefore, further effort is needed to develop effective training and education to ensure proper surgical techniques and pathologic examinations and documentation especially in the facilities with consistently lower LN yield in their dissections to increase compliance with standard clinical guidelines and thus reduce disparity.