Insurance status and survival disparities among nonelderly rectal cancer patients in the National Cancer Data Base

Authors


  • This article contains original material. The material is not in any previous abstracts, presentations, reports, or publications.

Abstract

BACKGROUND:

Among patients with colorectal cancer, insurance status is associated with disparities in survival as well as differences in stage and treatment. The role of stage and treatment differences in these survival disparities is not clear because insurance status is also strongly correlated with race/ethnicity, socioeconomic status, and other factors.

METHODS:

The authors used data from the National Cancer Data Base, a national hospital-based cancer registry, to examine insurance status and other factors related to survival among 19,154 rectal cancer patients aged 18 to 64 years. The authors examined the impact of 10 factors on 5-year survival: age, sex, race/ethnicity, histologic grade, histologic subtype, neighborhood education and income levels, facility type, stage, and treatment.

RESULTS:

Adjusted only for age, the hazard ratio (HR) for death at 5 years was 1.00 (referent) among privately insured patients, 2.05 (95% confidence interval [CI], 1.89-2.23) among Medicaid-insured patients, and 2.01 (95% CI, 1.84-2.19) among uninsured patients. After adjustment for all factors other than stage and treatment, the HRs were 1.88 (95% CI, 1.722.04) for Medicaid-insured patients and 1.84 (95% CI, 1.69-2.01) for uninsured patients. After further adjustment for stage and treatment, the HRs were 1.34 (95% CI, 1.22-1.46) for Medicaid-insured patients and 1.29 (95% CI, 1.18-1.42) for uninsured patients.

CONCLUSIONS:

After adjustment for age, further adjustment for 9 other factors reduced the excess mortality among rectal cancer patients without private insurance by approximately 70%. Disparities in stage and treatment accounted for approximately 53% of the excess mortality, whereas factors other than stage and treatment accounted for approximately 17%. Cancer 2010. © 2010 American Cancer Society.

Several studies have shown that colorectal cancer (CRC) patients without private insurance are more likely to be diagnosed with late-stage disease.1-3 Other studies have shown survival disparities by insurance status among CRC patients.4-9 In some of the survival studies, the differences narrowed substantially, but did not disappear, after stratification by stage.9 Treatment disparities may play a role in these findings, but data are extremely limited on the relation between insurance and cancer treatment, and results from the few available studies have not been consistent. For example, only 5 studies have compared the use of recommended treatments among CRC patients with private insurance, public insurance, and no insurance. One found lower use in patients with private insurance,10 3 found higher use in patients with private insurance,4, 11, 12 and 1 found no difference in use by insurance type.13 Moreover, there are demographic factors such as age, sex, race/ethnicity, and poverty status, which may also partly explain the observed survival disparities.14 Thus, the relative importance of stage, treatment, and other factors in explaining insurance-related CRC survival disparities is not clear.

A major factor underlying the paucity of data on insurance and cancer treatment is the finding that information on insurance status is not routinely collected by population-based cancer registries, and treatment data are often incomplete. Most studies of cancer treatment in the United States rely on Surveillance, Epidemiology, and End Results (SEER) and Medicare linked files and include only Medicare-insured patients aged 65 and older. The National Cancer Data Base (NCDB) provides a unique opportunity to study the relation between insurance status and cancer treatment because information on primary payer at diagnosis has been collected by the NCDB for all patients since 1996, and hospital registries reporting to the NCDB are expected to collect data on all initial treatment modalities, including outpatient therapy.

Colorectal cancer is the third most common cause of cancer incidence and mortality in both men and women in the United States, with an estimated 146,970 cases and 49,920 deaths occurring in 2009.15 We were able to identify only 1 published study of insurance-related CRC survival disparities that simultaneously examined the roles of stage, treatment, and demographic factors. This study used state cancer-registry data from Florida.4 To explore the relative contribution of stage, treatment, and 8 other factors in these survival disparities, we used a large national sample to conduct a study of insurance status and survival among 19,154 rectal cancer patients aged 18-64 years. These data were on patients who were diagnosed during the years 1998-2002 and followed through the year 2007. We restricted the study to rectal cancer rather than all colorectal cancers because rectal cancer is staged and treated differently from colon cancer and because previous research is more limited than for colon cancer.

MATERIALS AND METHODS

The NCDB is a national hospital-based cancer registry supported by the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. Approximately 1430 hospitals currently report to the NCDB, which captures approximately 70% of newly diagnosed cancers in the United States. In common with population-based registries, the NCDB contains standardized data elements on patient demographics, tumor characteristics, including stage and histopathologic findings, first course of treatment, and ZIP (postal) code of residence, which is useful for area-based socioeconomic characteristics. In addition, the NCDB contains information on patient insurance status. Reporting facilities are categorized by the CoC Accreditation Program as teaching/research hospitals, comprehensive community cancer centers, community cancer centers, or “other” hospitals (which include facilities operated by the federal government).

Data in the NCDB were collected according to the Registry Operations and Data Standards (ROADS) manual,16 the fifth edition of the American Joint Committee on Cancer (AJCC) Manual for Staging of Cancer,17 and the International Classification of Diseases for Oncology (ICD-O).18 Patients diagnosed with rectal cancer (ICD-O site code C20.9; all histologies except 9750, 9760-9764, 9800-9820, 9826, 9831-9920, 9931-9964, and 9980-9989) from the years 1998 to 2002 were included. Further selection criteria were received first course of treatment at the reporting facility; first malignant primary; invasive tumor behavior; aged 18-64 years with private insurance, Medicaid insurance, or no insurance (included “not insured” and “not insured, self-pay”); AJCC staging system group I, II, III, or IV; and reported by facilities that participated in the NCDB continuously from 1998 to 2006. We did not include patients aged 65 years and older because almost all are covered by Medicare insurance, and we did not include younger Medicare patients because Medicare eligibility under age 65 years is based on the presence of disabling medical conditions, which may modify treatment recommendations and outcomes. We did not include patients who had other forms of public insurance (eg, Indian Health Service) because there were not enough of these patients to analyze separately. A total of 28,667 patients met these initial criteria.

We excluded cases with missing data on vital status or survival time (n = 239), cases with missing data on treatment (surgery, chemotherapy, or radiation therapy; n = 2990), and cases with missing data on 1 or more independent variables (age, sex, race, ethnicity, histologic grade, histologic subtype, neighborhood education level, neighborhood income level, or facility type; n = 6284).

Thus, a total of 19,154 cases were available for analysis. Informed consent was not required because the data used for this study involved secondary analysis of existing data recorded in a manner in which individuals could not be identified (either directly or by identifiers linked to the individuals).

The survival endpoint in this analysis was vital status at 5 years after diagnosis, and survival time was measured in months. Cases were coded as alive, dead, or lost to follow-up, and cause of death was not available. Patients who were lost to follow-up before 5 years were censored as of their last contact, and patients who were followed for more than 5 years were censored at 5 years. On the basis of stage-specific treatment guidelines for rectal cancer,19 we classified all possible treatment combinations into 3 categories: 1) surgery, with or without chemo/radiation; 2) chemo/radiation alone; and 3) no surgery or chemo/radiation. Surgery is recommended for stage I, II, and III patients and selected stage IV patients. Chemo/radiation is recommended for stage II, III, and IV patients, either as neoadjuvant/adjuvant therapy (stages II and III) or as first-line therapy (stage IV). AJCC stage group was based on pathologic TNM (tumor, lymph node, metastasis) variables when these were available and on clinical TNM variables when they were not. By using ZIP code-level data from the 2000 US Census, neighborhood education level was measured by the proportion of residents in the patient's ZIP code without a high school diploma, and neighborhood income level was measured by the median household income in the patient's ZIP code. Both variables were coded using quartiles based on national distributions.

Because preliminary analyses revealed a strong interaction between the effects of stage and treatment, we modeled these 2 effects jointly by creating a single variable with 12 categories (4 stage × 3 treatment categories). To adjust for stage- and treatment-related differences within these 12 categories, we included the following variables in all models: surgical procedure of primary site, surgical margins of primary site, whether chemotherapy and radiation were used alone or together, and pathologic or clinical values for tumor (T), lymph node (N), and metastasis (M) elements of the AJCC staging system. The following variables were also included in all models: insurance status, age, sex, race/ethnicity, histologic grade, histologic subtype, neighborhood education and income levels, and type of reporting facility. After we confirmed the proportional hazards assumption, survival analysis was carried out by using Cox proportional hazards regression. These analyses were performed with the PHREG procedure in SAS software, version 9.2 (SAS Institute, Cary, NC). All statistical analyses used 2-tailed P values, and an α of .05 was used to test for statistical significance.

Because we did not have data on causes of death, we conducted additional analyses by using 5-year relative survival rates to assess the role of death from noncancer causes in our main findings. The relative survival rate is the ratio of the survival observed in the population of interest (cancer patients) to the survival that would be expected if the patients experienced only the mortality of the general population from which they were drawn.20, 21 The relative survival rate represents the probability of surviving a cancer, and 1 minus this rate represents the probability of dying from a cancer. The probability of death from noncancer causes was computed as the difference between total mortality and death from cancer. The expected probabilities for these analyses were obtained from life tables for the US population by single years of age, from birth to 118 years. Age-specific, sex-specific, and race-specific life tables for the year 2000 were used for these analyses.22

RESULTS

In this population of nonelderly rectal cancer patients, insurance status was strongly associated with all of the prognostic factors selected for the study except for histologic grade (Table 1). Patients with Medicaid insurance or no insurance were younger and more likely to be female and nonwhite. Patients without private insurance were much more likely to be diagnosed with stage IV cancer and much less likely to be diagnosed with stage I cancer. There were only small differences in histologic grade and subtype across insurance statuses versus large insurance-related differences in neighborhood education and income level. Patients with Medicaid insurance or no insurance were twice as likely to reside in neighborhoods with the lowest levels of education and income. Patients with private insurance were more likely to be treated in comprehensive community cancer centers, whereas patients with Medicaid insurance or no insurance were more likely to be treated in teaching/research hospitals.

Table 1. Characteristics of Patients by Insurance Status, Nonelderly Rectal Cancer Patients Diagnosed Between 1998 and 2002, National Cancer Data Base (N=19,154)
CharacteristicaInsurance Status
PrivateMedicaidUninsured
n=16,599n=1321n=1234
  • AJCC indicates American Joint Committee on Cancer; CCC, community cancer center.

  • a

    P<.0001 for all differences by insurance status, except histologic grade (P = .14).

  • b

    Using 2000 US Census data for ZIP code of residence at diagnosis.

Mean age at diagnosis, y53.252.152.2
Sex, %   
 Men62.454.359.5
 Women37.645.740.5
Race/Ethnicity   
 White85.262.263.9
 Black7.218.516.2
 Hispanic3.913.512.4
 Asian3.45.66.8
 Other0.30.30.7
AJCC stage group   
 I31.019.517.6
 II24.127.328.7
 III31.128.631.3
 IV13.824.622.5
Histologic grade   
 Well differentiated9.48.99.1
 Moderately differentiated72.370.969.9
 Poorly differentiated17.719.820.1
 Undifferentiated/anaplastic0.60.50.9
Histologic subtype   
 Adenocarcinoma91.386.590.0
 Mucinous adenocarcinoma6.08.65.8
 Other2.85.04.2
Residents without high school degree,b %   
 <14.038.916.219.9
 14.0-19.924.119.420.1
 20.0-28.922.630.427.2
 ≤29.014.434.132.8
Median household income,b %   
 ≥$46,00044.818.824.1
 $35,000-$45,99927.626.427.3
 $30,000-$34,99916.725.822.7
 <$30,00011.029.025.9
Type of reporting facility   
 CCC15.013.215.7
 Comprehensive CCC50.938.539.6
 Teaching/research33.748.244.5
 Other0.40.20.2

Insurance status was also strongly associated with stage-specific treatment for rectal cancer (Table 2). Within every stage, patients with Medicaid insurance or no insurance were substantially less likely to receive standard therapies compared with patients with private insurance. Among patients with stage I cancer, 95% of privately insured patients underwent surgery compared with 83% of uninsured patients. Among patients with stage II and III disease, patients without private insurance were less likely to undergo surgery and were more likely to be treated with chemo/radiation alone. For example, 8% of privately insured stage II patients received chemo/radiation alone versus 19% of stage II patients without private insurance. Among patients with stage IV disease, patients without private insurance were more likely to receive no treatment (no surgery or chemo/radiation). Only 4% of privately insured stage IV patients received no treatment versus 15% of uninsured stage IV patients.

Table 2. Receipt of Stage-Specific Treatment by Insurance Status, Nonelderly Rectal Cancer Patients Diagnosed Between 1998 and 2002, National Cancer Data Base (N=19,154)
Stage-Specific Treatmenta% Patients Receiving Treatment by Insurance Status
PrivateMedicaidUninsured
n=16,599n=1321n=1234
  • a

    P<.0001 for all differences by insurance status.

Stage I   
 Surgery with or without chemo/radiation95.189.983.4
 Chemo/radiation alone3.27.813.8
 No surgery or chemo/radiation1.72.32.8
Stage II   
 Surgery with or without chemo/radiation91.485.979.4
 Chemo/radiation alone7.711.919.2
 No surgery or chemo/radiation0.92.21.4
Stage III   
 Surgery with or without chemo/radiation94.989.489.9
 Chemo/radiation alone4.79.09.6
 No surgery or chemo/radiation0.41.60.5
Stage IV   
 Surgery with or without chemo/radiation62.046.242.2
 Chemo/radiation alone33.645.943.0
 No surgery or chemo/radiation4.48.014.8

In multivariate proportional hazards models, all 10 of the prognostic factors selected for the study were significant independent predictors of death at 5 years (Table 3). After adjustment for age only, patients with Medicaid insurance or no insurance were more than twice as likely to die within 5 years than patients with private insurance. Age-adjusted hazard ratios (HRs) for death at 5 years (and associated 95% confidence intervals [CIs]) for Medicaid-insured and uninsured patients were 2.05 (95% CI, 1.89-2.23) and 2.01 (95% CI, 1.84-2.19), respectively. After adjustment for all factors except stage and treatment, the HRs for Medicaid-insured and uninsured patients were reduced by approximately 17% to 1.88 (95% CI, 1.72-2.04) and 1.84 (95% CI, 1.69-2.01), respectively. (Before adjustment, the average excess mortality in patients without private insurance was approximately 103% [HR, 2.03] vs 86% after adjustment [HR, 1.86]; [86-103]/103 = −0.17 [17% reduction].)

Table 3. Relation Between Insurance Status and Survival Among Nonelderly Rectal Cancer Patients, National Cancer Data Base (N=19,154)
VariableHazard Ratio (95% CI) for Death at 5 Years After Adjustment for
Age OnlyAll Factors Except Stage and TreatmentAll Factors Except Treatment
  • CI indicates confidence interval; CCC, community cancer center.

  • a

    Used 2000 US Census data for ZIP code of residence at diagnosis.

Insurance status   
 Private1.00 (referent)1.00 (referent)1.00 (referent)
 Medicaid2.05 (1.89-2.23)1.88 (1.72-2.04)1.49 (1.37-1.63)
 Uninsured2.01 (1.84-2.19)1.84 (1.69-2.01)1.54 (1.41-1.68)
Age at diagnosis, y   
 <481.00 (referent)1.00 (referent)1.00 (referent)
 49-540.98 (0.94-1.09)1.01 (0.94-1.09)1.02 (0.95-1.09)
 55-590.97 (0.93-1.07)1.00 (0.93-1.07)1.07 (1.00-1.15)
 >601.08 (1.05-1.21)1.13 (1.05-1.21)1.17 (1.10-1.26)
Sex   
 Male1.00 (referent)1.00 (referent)
 Female0.83 (0.79-0.87)0.87 (0.82-0.91)
Race/ethnicity   
 White1.00 (referent)1.00 (referent)
 Black1.32 (1.22-1.44)1.26 (1.16-1.37)
 Hispanic1.09 (0.98-1.22)1.08 (0.96-1.20)
 Asian0.98 (0.85-1.12)0.99 (0.86-1.13)
 Other1.04 (0.67-1.64)0.99 (0.63-1.55)
Histologic grade   
 Well differentiated1.00 (referent)1.00 (referent)
 Moderately differentiated1.46 (1.31-1.62)1.17 (1.05-1.31)
 Poorly differentiated2.93 (2.62-3.28)1.87 (1.67-2.10)
 Undifferentiated/anaplastic4.20 (3.24-5.44)2.66 (2.05-3.45)
Histologic subtype   
 Adenocarcinoma1.00 (referent)1.00 (referent)
 Mucinous adenocarcinoma1.21 (1.10-1.33)1.17 (1.06-1.29)
 Other1.31 (1.16-1.49)1.50 (1.32-1.70)
Residents without high school degree,a %   
 <14.01.00 (referent)1.00 (referent)
 14.0-19.91.10 (1.03-1.19)1.06 (0.99-1.15)
 20.0-28.91.18 (1.08-1.28)1.11 (1.02-1.20)
 ≥29.01.12 (1.02-1.24)1.09 (0.99-1.21)
Median household income,a %   
 ≥$46,0001.00 (referent)1.00 (referent)
 $35,000−$45,9991.07 (1.00-1.15)1.07 (1.00-1.15)
 $30,000−$34,9991.11 (1.02-1.21)1.11 (1.02-1.21
 <$30,0001.11 (1.00-1.23)1.08 (0.98-1.19)
Type of reporting facility   
 CCC1.03 (0.95-1.11)1.15 (1.06-1.24)
 Comprehensive CCC0.98 (0.93-1.04)1.05 (0.99-1.11)
 Teaching/research1.00 (referent)1.00 (referent)
 Other1.07 (0.72-1.61)1.57 (1.05-2.35)
AJCC stage group   
 I 1.00 (referent)
 II 2.25 (2.03-2.50)
 III 3.60 (3.27-3.96)
 IV 19.75 (17.95-21.72)

Additional adjustment for AJCC stage led to an additional 33% reduction in the HRs for Medicaid-insured and uninsured patients, to 1.49 (95% CI, 1.37-1.63) and 1.54 (95% CI, 1.41-1.68), respectively. In these multivariate models, factors associated with poorer survival were older age, male sex, black race, less differentiated histologic grade, lower neighborhood education and income levels, treatment at community cancer centers, and higher stage at diagnosis. The most important of these were histologic grade (HR for grade 4, 2.66; 95% CI, 2.05-3.45) and stage at diagnosis (HR for stage IV, 19.75; 95% CI, 17.95-21.72).

After further adjustment for stage-specific treatment, the HRs for Medicaid-insured and uninsured patients were reduced an additional 20%, to 1.34 (95% CI, 1.22-1.46) and 1.29 (95% CI, 1.18-1.42), respectively (Table 4). Even after adjustment for all factors including treatment, several demographic factors remained significant predictors of mortality. Black patients had a higher risk of death at 5 years (HR, 1.16; 95% CI, 1.07-1.27) as did patients residing neighborhoods with lower levels of education (HR, 1.13; 95% CI, 1.04-1.23) and income (HR, 1.14; 95% CI, 1.04-1.24). In summary, after adjustment for age, further adjustment for 9 other factors reduced the excess mortality among rectal cancer patients without private insurance by approximately 70%. Factors other than stage and treatment accounted for approximately 17% of the excess mortality, whereas disparities in stage and treatment accounted for approximately 53%.

Table 4. Relation Between Insurance Status and Survival Among Nonelderly Rectal Cancer Patients, National Cancer Data Base (N=19,154)
VariableaHazard Ratio (95% CI) for Death at 5 Years After Adjustment for All Factors
  • a

    The model also includes histologic subtype, type of reporting facility, surgical procedure of primary site, surgical margins of primary site, whether chemotherapy and radiation were used alone or together, and pathologic or clinical values for tumor (T), node (N), and metastasis (M) elements of the American Joint Committee on Cancer (AJCC) stage group (results not shown).

  • b

    Uses 2000 US Census data for ZIP code of residence at diagnosis.

Insurance status 
 Private1.00 (referent)
 Medicaid1.34 (1.22-1.46)
 Uninsured1.29 (1.18-1.42)
Age at diagnosis, y 
 <481.00 (referent)
 49-541.05 (0.98-1.13)
 55-591.12 (1.04-1.20)
 %ge;601.20 (1.12-1.29)
Sex 
 Men1.00 (referent)
 Women0.85 (0.80-0.89)
Race/ethnicity 
 White1.00 (referent)
 Black1.16 (1.07-1.27)
 Hispanic1.01 (0.91-1.14)
 Asian/Pacific Islander0.93 (0.81-1.07)
 Other0.77 (0.49-1.21)
Histologic grade 
 Well differentiated1.00 (referent)
 Moderately differentiated1.20 (1.07-1.33)
 Poorly differentiated1.82 (1.62-2.04)
 Undifferentiated/anaplastic2.53 (1.93-3.30)
Proportion of residents without high school degree,b % 
 <14.01.00 (referent)
 14.0-19.91.04 (0.96-1.12)
 20.0-28.91.13 (1.04-1.23)
 ≥29.01.09 (0.98-1.20)
Median household income,b % 
 ≥$46,0001.00 (referent)
 $35,000−$45,9991.06 (0.99-1.14)
 $30,000−$34,9991.14 (1.04-1.24)
 <$30,0001.05 (0.95-1.16)
Stage I 
 Surgery with or without chemo/radiation0.58 (0.34-0.99)
 Chemo/radiation alone1.72 (0.96-3.09)
 No surgery or chemo/radiation1.00 (referent)
Stage II 
 Surgery with or without chemo/radiation0.26 (0.17-0.39)
 Chemo/radiation alone0.81 (0.53-1.25)
 No surgery or chemo/radiation1.00 (referent)
Stage III 
 Surgery with or without chemo/radiation0.53 (0.31-0.91)
 Chemo/radiation alone0.81 (0.46-1.44)
 No surgery or chemo/radiation1.00 (referent)
Stage IV 
 Surgery with or without chemo/radiation0.34 (0.29-0.40)
 Chemo/radiation alone0.69 (0.58-0.81)
 No surgery or chemo/radiation1.00 (referent)

As noted above, it is possible that some of the insurance-related survival differences seen among cancer patients could be due to disparities in risk of death from noncancer causes. In secondary analyses based on 5-year relative survival rates, we found the largest difference in risk of death from cancer by insurance status was 20.1% (23.0% vs 43.1%), whereas the largest difference in risk of death from noncancer causes was only 1.3% (2.4% vs 3.7%). Thus, the disparity in risk of death from cancer by insurance status was approximately 15.5 times larger than the disparity in risk of death from noncancer causes.

DISCUSSION

By using national hospital-based cancer registry data from greater than 19,000 nonelderly rectal cancer patients, we found that the age-adjusted risk of death at 5 years was more than twice as high among patients without private insurance. Disparities in stage and treatment explained approximately 53% of the excess mortality seen among Medicaid-insured and uninsured patients, whereas factors other than stage and treatment explained approximately 17%. Together, 9 factors accounted for approximately 70% of the disparity in age-adjusted risk of death at 5 years seen among patients without private insurance. This appears to be the first study of insurance-related colorectal cancer (CRC) survival disparities in a national patient sample of individuals younger than age 65 years. Our main finding that most of the excess mortality seen among Medicaid-insured and uninsured patients was explained by 2 modifiable factors (stage and treatment) suggests that improving insurance coverage and reducing cost-related barriers to primary care, CRC screening, and high-quality treatment would have a major impact on CRC survival disparities.

The findings of the present study are in general agreement with those of Roetzheim et al,4 who studied stage-specific treatment and survival by insurance status among CRC patients in Florida, and with those of Esnaola et al,12 who used 2003-2005 NCDB data to study the relation between treatment and insurance status among patients with nonmetastatic rectal cancer. Our findings are also consistent with prior studies that found higher tumor grade to be an independent prognostic factor for CRC survival in multivariate analyses.23-26

Although the present study provides some unique findings for insurance status, there is a large body of literature showing that demographic factors such as age, sex, race/ethnicity, and socioeconomic status are significant predictors of treatment and survival among CRC patients. In our study, most of these factors remained significantly associated with treatment and survival in multivariate models that adjust for insurance type, stage, treatment, and other patient, facility, and neighborhood characteristics. For example, we found that compared with white patients, black patients were approximately 16% more likely to die within 5 years of diagnosis; similarly, patients living in neighborhoods with lower levels of education and income had a 13%-14% higher risk of death at 5 years. The survival disparities we observed by race and area socioeconomic status (SES) may reflect geographic and transportation barriers, higher rates of comorbid conditions, cost barriers to medical care, mistrust of the medical system and language and/or cultural issues. Physician biases, such as patient preference, likelihood of patient compliance, and other nonmedical factors, may also influence treatment recommendations and physician-to-patient communication of treatment options.27

Our study has several potential limitations. Because we used data only from hospitals with CoC-approved programs that reported data continuously to the NCDB from 1998 to 2006, the patients and treatment patterns may not be fully representative of all rectal cancer patients in the United States. However, the NCDB includes clinical, treatment, and demographic information for approximately 70% of all incident cancer cases in the United States; such data are not centrally available for the state population-based registries. Another limitation is that the SES variables were based on ZIP code-level characteristics, which are less precise than census tract-level estimates of education level and poverty status, and we did not have individual measures of education and income.

Although data on insurance status collected in the NCDB have not been independently validated, it is highly complete and is abstracted from hospital records where billing data are likely to be accurately recorded. However, nonrandom misclassification is known to occur because a substantial number of patients move from the uninsured category to Medicaid when they are diagnosed with cancer.7 This is because in most states, individuals who develop serious illnesses, including cancer, can qualify for Medicaid as “medically needy” even when their income is higher than the state Medicaid eligibility limit. Although we were not able to adjust for the presence of comorbid medical conditions in this study because the NCDB did not include these until 2003, in a prior study using 2003-2005 data from nonelderly NCDB patients with colorectal cancer, we found that differences in comorbidity explained only approximately 10% of the excess mortality among patients without private insurance.28 Although registry data on chemotherapy may be incomplete, our stage-specific rates of chemotherapy among younger adult rectal cancer patients were higher than those reported by Cronin et al from a national population-based sample that included adults of all ages.29 Finally, there were some aspects of care that we were unable to consider because of limitations in our data.

Despite its limitations, this study provides substantial evidence that poorer survival among cancer patients without private insurance is largely attributable to later stage at diagnosis and inadequate treatment, factors that may be modifiable through healthcare reform. Addressing insurance-related and cost-related barriers to high-quality prevention, early detection, and treatment may not be the only measure that will be needed to eliminate cancer disparities, but our data suggest that these barriers may be particularly important.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary