Socioeconomic and demographic disparities in treatment for carcinomas of the colon and rectum
The current study examined the relationship between socioeconomic and demographic factors and type of treatment for carcinomas of the colon and rectum. The National Institutes of Health and the National Cancer Institute recommend surgery followed by adjuvant chemo- and/or radiotherapy for Stage III colon and Stages II and III rectal carcinomas.
The authors linked Washington State's cancer registry and hospital discharge records and U.S. census data to assess socioeconomic and demographic factors related to treatment, controlling for clinical factors.
Compared to colon carcinoma patients under age 65 years, patients aged 75–84 years and 85 years or older were at higher risk for a treatment plan of surgery without adjuvant therapy (adjusted odds ratio [OR] = 2.5, 95% confidence interval [CI] = 1.3–4.7; OR = 14.1, CI = 6.3–31.4, respectively). Risk of no adjuvant therapy was more than doubled for patients in zip codes in the lowest quartile of per capita income compared to the top three quartiles (OR = 2.3, CI = 1.5–3.4) and for those with Medicare compared to private insurance (OR = 2.2, CI = 1.3–3.8). Older patients with rectal carcinoma were also at higher risk of a treatment plan that did not include adjuvant therapy.
The current findings suggest disparities in the provision of recommended medical procedures related to socioeconomic and demographic factors. Cancer 2002;95:39–46. © 2002 American Cancer Society.
The U.S. Department of Health and Human Services Healthy People 2010 initiative1 articulates two overarching public health goals for the next decade. One of these goals is the elimination of health disparities due to race, income and rural residency. Healthy People 2010 describes many factors related to poorer health among low income or rural populations, including access to high quality health care. Recent studies on treatment for breast, lung, and colorectal carcinomas suggest that medical treatment for cancer varies with socioeconomic and demographic factors independent of clinical presentation.2–9 Older age is a consistent predictor of poorer quality of care.3, 5–7 Studies have also identified race,2, 4, 8 education,5, 7 income,4 marital status,9 and type of medical insurance9 as important in explaining variations in cancer treatment.
Colorectal carcinoma is the second leading cause of cancer death in the United States. The fact that there is a recognized consensus regarding initial treatment for Stage II rectal carcinoma (cancer that has spread to surrounding tissue but not to lymph nodes or distant tissue) and Stage III rectal and colon carcinoma (cancer that has spread to regional lymph nodes) makes it possible to study quality of treatment using administrative databases. For Stage III colon and Stages II and III rectal carcinoma, the National Institutes of Health 1990 Consensus Conference recommended surgery followed by adjuvant chemo- and/or radiotherapy.10 The National Cancer Institute supports this recommendation.11 The goal of the current study was to show whether in Washington state income, type of medical insurance, age, race, and/or residence in a rural area were associated with an initial treatment plan that included adjuvant therapy.
Cancer Registry Data
The Washington State Cancer Registry (WSCR) collects information on all Washington state residents diagnosed with cancer. The WSCR contains identifying information, including name, gender, and date of birth. It also includes information on race and residence at diagnosis, primary site, stage at diagnosis, planned first course of treatment, and hospital where treatment was given.
Stage at diagnosis is recorded following two different coding schemes depending on where the cancer diagnosis report originates. Reports originating in the Seattle/Puget Sound Surveillance, Epidemiology and End Results (SEER) program operated by the Fred Hutchinson Cancer Research Center use extent of disease SEER coding standards. This program covers 13 counties in the Puget Sound area and submits 70% of the WSCR cancer reports. Reports from the rest of the state use the American Joint Commission on Cancer (AJCC) staging criteria. We converted the SEER staging to the AJCC staging using SEER criteria.12 We then identified all patients in the WSCR with Stage II or III rectal or Stage III colon carcinoma diagnosed in 1996 or 1997 whose WSCR records indicated they received surgery. Patients were classified as having an initial treatment plan that included adjuvant therapy if the chemo- or radiotherapy treatment fields in the WSCR indicated that therapy had been administered, recommended, or refused by the patient or guardian.
Hospital Discharge Data
Since surgery for Stage II rectal and Stage III rectal and colon carcinomas requires inpatient hospitalization, we used probabilistic linkage software, Automatch version 3.0 (Ascential Software, West Borough, MA),13 to link the cancer cases to Washington's hospitalization data. We obtained the hospitalization data from the Comprehensive Hospital Abstract Reporting System (CHARS), which is compiled from uniform billing forms for all state-licensed hospitals in Washington. The system does not include information for hospitalizations in military, Veterans Administration, or out-of-state hospitals. CHARS contains information on the primary diagnosis, up to eight concurrent diagnoses, and primary and secondary payers of hospital expenses. The diagnoses are coded to the International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM). The patient identifiers include the first two letters of the first and last names, gender, date of birth, and zip code of the patient's residence. Linkage was based on these identifiers, diagnosis, hospital, and month and year of hospital admission and diagnosis.
To develop a measure of comorbidity, the noncancer diagnoses included in the CHARS records containing surgery for colon or rectal carcinoma were weighted following the ICD-9-CM adaptation of the Charlson comorbidity index.14, 15 We made three modifications to the index. First, diagnoses related to colorectal carcinoma were excluded from the index, since all patients in the study had colon or rectal carcinoma. Second, we did not weight the patients' comorbidity based on their ages because we controlled for age in the analysis. Third, we included diagnoses at the time of surgery for all conditions because conditions at the time of surgery may affect subsequent care.
Fluorouracil (5FU) is the most common agent used in adjuvant chemotherapy for colon and rectal carcinoma. We identified poor nutritional state, depressed bone marrow function, infection, and hypersensitivity to 5FU as contraindications to use of 5FU.16 There is no ICD-9-CM code for hypersensitivity to 5FU. We converted the other three conditions into ICD-9-CM codes (Table 1) so that we could determine whether the hospitalization that included the patient's surgery included diagnoses with contraindications to chemotherapy. We did not identify any contraindications for adjuvant radiation therapy.
Table 1. International Classification of Diseases, 9th Revision Clinical Modification Codes Used to Identify Contraindications to Fluorouracil
|Poor nutritional state|| |
|Effects of hunger||994.2|
|Adult neglect nutritional||995.84|
|Depressed bone marrow function|| |
|Diseases of white blood cells||288|
|Infections and parasitic diseases||001–139|
|Central nervous system||320, 322–326, 373, 360.0–.1, 374.4, 375.3, 376.0–.1, 380.1|
|Respiratory system||460–466, 473–475, 480–487|
|Digestive system||566–567, 569.5 –.6, 575.0, 576.1|
|Genitourinary system||590, 599.0, 614–616|
|Skin & subcutaneous system||680–686|
|Miscellaneous||771, 958.3, 996.6, 998.5, 999.3|
The primary payer in CHARS is the entity that has the principal responsibility for paying the hospital bill. We divided primary payer into five groups: Medicare, Medicaid, other government payer (e.g., Washington State Labor and Industries, Indian Health Services, military), private insurance (both fee for service and managed care plans), and self-pay.
U.S. Census Data
We used the patients' zip codes as recorded in WSCR combined with 1990 U.S. Census data to determine rural residence and to assign patients to quartiles of income. Specifically, we used 1990 U.S. Census population data and TIGER line files from Maptitude version 4.0 (Caliper Corp., Newton, MA) to identify the population density of the zip code in which each patient resided. We defined areas with less than 100 people per square mile as rural. We also used 1990 U.S. Census data from Maptitude version 4.0 to determine the per capita income for each patient's zip code. Table 2 includes the data source for each variable in the analysis.
Table 2. Data Sources
|Type of cancer||WSCR (100% concordant with CHARS)|
|Type of treatment|| |
| Surgery||WSCR (100% concordant with CHARS)|
| Adjuvant therapy||WSCR|
|Gender||WSCR (99.7% % concordant with CHARS)|
|Race||WSCR (not available in CHARS)|
|Age group||WSCR (99.6% concordant with CHARS)|
|Zip code of residence||WSCR (93.3% concordant with CHARS)|
| Urban-rural designation of zip code||1990 US Census|
| Per capita income of zip code||1990 US Census|
|Contraindications for 5FUb||CHARS|
|ACOS/COC approved hospital||CHARS|
|Stage at diagnosis||WSCR|
We analyzed data for rectal and colon carcinomas separately. We first calculated crude odds ratios and 95% confidence intervals for surgery with no adjuvant therapy for each of the socioeconomic, demographic, and clinical factors. We then conducted logistic regression analyses. For these analyses, we combined levels of a variable when the strength of the association was the same across different levels in the univariate analyses. Since age had the strongest association with treatment in the univariate analyses, the first set of regression models included age and each of the other the factors that was statistically significantly associated with treatment in the univariate analyses. The final model initially included all of the factors that were associated with treatment in the first set of models. We then dropped factors if they were not statistically significantly associated with treatment and did not substantially alter the odds ratios of factors that were associated with treatment. We used a P value of 0.05 or less to determine statistical significance. We evaluated the final models for overdispersion by fitting a quasi-likelihood model to the data, and we examined residuals to look for outliers and overly influential observations and to evaluate the fit of the model.17, 18
The WSCR includes 1,192 reports of Stage II and III rectal and Stage III colon carcinomas diagnosed in 1996 and 1997. We excluded 51 records where WSCR recorded a military, Veterans Administration, or out-of-state hospital; 35 records where the adjuvant therapy field was unknown; and 8 records that did not include surgery in the first course of treatment. Of the remaining 1,098 records, we were able to match 1,021 (93.0%) to the CHARS data set. Approximately 98% of the matches were exact matches. For 2% of the matches, there were differences between the two data sets for one of the variables used for matching. One of the records had a zip code that was not included in the 1990 census data, resulting in complete data for 1,020 records, 388 for rectal and 632 for colon carcinoma. These records represent 1,018 people. One of the two people who were counted twice had primary diagnoses of both colon and rectal carcinomas. The other person had two primary diagnoses of colon carcinoma.
The initial treatment plan did not include adjuvant therapy for 38.1% of the patients with colon carcinoma and for 26.6% of the patients with rectal carcinoma. Over 90% of the study group was white and approximately 80% lived in nonrural areas. The rectal carcinoma patients were somewhat younger than the patients with colon carcinoma, but there was not a substantial difference in the zip code per capita income of the two groups. Medicare was the largest primary payer, followed by private insurance. We identified very few people with contraindications to 5FU. Approximately 85% of the study group attended hospitals with approved American College of Surgeons/Committee on Cancer programs (Table 3).
Table 3. Type of Treatment and Characteristics of Study Patients
|Type of treatment|| || |
| Surgery only||241 (38.1%)||103 (26.6%)|
| Surgery and adjuvant therapy||391 (61.9%)||285 (73.5%)|
|Gender|| || |
| Female||323 (51.1%)||155 (40.0%)|
| Male||309 (48.9%)||233 (60.1%)|
|Race|| || |
| American Indian/Alaska Native||3 (0.5%)||7 (1.8%)|
| Asian/Pacific Islander||11 (1.7%)||12 (3.1%)|
| Black||11 (1.7%)||6 (1.6%)|
| White||592 (93.7%)||359 (92.5%)|
| Unknown||15 (2.4%)||4 (1.0%)|
|Age in years|| || |
| Less than 65||184 (29.1%)||152 (39.2%)|
| 65–74||167 (26.4%)||110 (28.4%)|
| 75–84||191 (30.2%)||106 (27.3%)|
| 85 and older||90 (14.2%)||20 (5.2%)|
|Residence|| || |
| Rural zip code||112 (17.7%)||84 (21.7%)|
| Nonrural zip code||520 (82.3%)||304 (78.4%)|
|Zip code per capita income|| || |
| Quartile 1: $6,411-$11,881||154 (24.4%)||99 (25.5%)|
| Quartile 2: $11,884-$13,951||167 (26.4%)||94 (24.2%)|
| Quartile 3: $13,980-$16,413||157 (24.8%)||92 (23.7%)|
| Quartile 4: $16,431-$34,478||154 (24.4%)||103 (26.6%)|
|Primary payer|| || |
| Medicare||399 (63.1%)||212 (54.6%)|
| Medicaid||10 (1.6%)||16 (4.2%)|
| Other government insurance||10 (1.6%)||2 (0.5%)|
| Private insurance||207 (32.8%)||147 (37.9%)|
| Self pay||6 (1.0%)||11 (2.8%)|
|Comorbiditiesa|| || |
| None||440 (69.6%)||299 (77.1%)|
| One or more||192 (30.4%)||89 (22.9%)|
|Contraindications for 5FUb|| || |
| Yes||8 (1.3%)||2 (0.5%)|
| No||624 (98.7%)||386 (99.5%)|
|ACOS/COC approved hospital|| || |
| Approved||551 (87.2%)||327 (84.3%)|
| Not approved||61 (12.8%)||61 (15.7%)|
|Stage|| || |
| Stage II|| ||212 (54.6%)|
| Stage III||632 (100%)||176 (45.4%)|
In subsequent analyses, we combined people of American Indian, white, and unknown races. In the WSCR data, people with race coded as unknown are likely to be white. A linkage between Washington state mortality data and the WSCR showed that 112 of 113 people reported as unknown race in WSCR between 1992 and 1996 were reported as white on the death certificate (unpublished data, Washington State Department of Health). The same linkage showed that almost 30% of people classified as American Indian on the death certificate were classified as white in WSCR. Other reports also show that American Indian race is underreported in the WSCR.19, 20 Because it is likely that many American Indians are reported as white in the current data, and because of the small number of people reported as American Indian, we included those reported as American Indian with people reported as white and unknown. Results in the final logistic regression model were not substantively different when we excluded those reported as American Indians.
For colon carcinoma, the odds of initial treatment recommendation of surgery without adjuvant therapy were statistically significantly higher for women compared to men; for those in age groups over 65 years compared to those less than 65 years; for rural compared to nonrural residents; for people living in zip codes in the lowest quartile of per capita income compared to those living in the highest quartile; for people with Medicare as their primary payer compared to those with private insurance; and for those with one or more contraindications to 5FU compared to those with no contraindications (Table 4).
Table 4. Crude Odds Ratios and 95% Confidence Intervals for the Proportions with an Initial Treatment Recommendation of Surgery without Adjuvant Therapy
|Gender|| || |
| Female||1.6 (1.2–2.2)||1.6 (1.0–2.5)|
|Racea|| || |
| Non-API and non black||1.0||1.0|
| API||0.2 (0.0–1.0)||0.9 (0.2–3.4)|
| Black||0.2 (0.0–1.0)||1.4 (0.2–7.6)|
|Age in years|| || |
| Less than 65||1.0||1.0|
| 65–74||2.0 (1.2–3.3)||3.0 (1.5–5.8)|
| 75–84||3.9 (2.4–6.2)||6.5 (3.5–12.0)|
| 85 and older||22.0 (12.2–39.5)||51.8 (19.9–134.6)|
|Residence|| || |
| Non-rural zip code||1.0||1.0|
| Rural zip code||1.6 (1.1–2.4)||1.1 (0.7–2.0)|
|Zip code per capita income|| || |
| Quartile 4: $16,431-$34,478||1.0||1.0|
| Quartile 3: $13,980-$16,413||0.8 (0.5–1.3)||0.7 (0.4–1.3)|
| Quartile 2: $11,884-$13,951||1.0 (0.6–1.6)||0.7 (0.4–1.3)|
| Quartile 1: $6,411-$11,881||2.0 (1.2–3.1)||0.7 (0.4–1.3)|
|Primary payer|| || |
| Private insurance||1.0||1.0|
| Medicare||4.1 (2.8–6.0)||4.0 (2.3–6.8)|
| Medicaid||2.9 (0.8–10.2)||1.0 (0.2–4.6)|
| Other government||1.8 (0.5–7.3)||6.7 (0.6–78.5)|
| Self pay||0.9 (0.1–7.6)||1.5 (0.3–7.4)|
|Comorbiditiesb|| || |
| One or more||1.3 (0.9–1.9)||1.2 (0.7–2.0)|
|Contraindicationsc|| || |
| None||1.0||(0 cell: not computed)|
| One or more||11.7 (2.2–61.7)|| |
|Hospital|| || |
| ACOS/COC approved||1.0||1.0|
| Not approved||1.1 (0.7–1.8)||0.7 (0.4–1.4)|
|Stage|| || |
| Stage III|| ||1.0|
| Stage II|| ||2.3 (1.4–3.6)|
The findings for rectal carcinoma were similar except that there were no differences between rural and nonrural residents and there were too few people with contraindications to assess this factor. Those with Stage II rectal carcinoma were more than twice as likely to have an initial treatment recommendation that did not include adjuvant therapy compared to those with Stage III rectal carcinoma (Table 4).
In the final logistic regression model, colon carcinoma patients were at higher risk of not having adjuvant therapy as part of the recommended initial course of treatment if they were age 75 and older, living in zip codes in the lowest quartile of per capita income, and were covered by Medicare as their primary payer (Table 5).
Table 5. Odds Ratios and 95% Confidence Intervals for the Proportions with an Initial Treatment Recommendation of Surgery Without Adjuvant Therapy Adjusted for Other Factors in the Final Logistic Regression Model
|Age in years|| || |
| less than 65||1.0||1.0|
| 65–74||1.2 (0.6–2.4)||3.2 (1.6–6.4)|
| 75–84||2.5 (1.3–4.7)||6.4 (3.3–12.6)|
| 85 and older||14.1 (6.3–31.4)||59.8 (15.3–234.2)|
|Zip code per capita income|| || |
| Quartiles 2-4: $11,884-$34,478||1.0|| |
| Quartile 1: $6,411-$11,881||2.3 (1.5–3.4)|| |
|Primary payer|| || |
| Private insurance or self pay||1.0|| |
| Medicare||2.2 (1.3–3.8)|| |
| Medicaid||3.2 (0.8–12.5)|| |
| Other government insurance||2.0 (0.5–8.4)|| |
|Stage|| || |
| Stage III|| ||1.0|
| Stage II|| ||2.5 (1.5–4.3)|
Patients with rectal carcinoma were at higher risk of not having a recommendation of adjuvant therapy as part of the initial course of treatment if they were age 65 or older. Patients with Stage II rectal carcinoma were less likely to have a recommendation of adjuvant therapy compared to patients diagnosed at Stage III (Table 5).
An examination of residuals showed that the models provided a good fit to the data. Correction for overdispersion had almost no effect on the parameter estimates in either model.
The current study suggests that several socioeconomic and demographic factors are related to treatment of Stage III colon and Stages II and III rectal carcinoma. Older age was the demographic factor most strongly related to the lack of a recommendation for adjuvant therapy. This finding is consistent with that of other investigators who identified older age as a risk factor for sub-optimal treatment of cancer.3, 5–7 Neither the National Institutes of Health Consensus Conference10 nor the National Cancer Institute11 limits the recommendation for adjuvant therapy for Stage III colon and Stages II and III rectal carcinomas to younger age groups. In relation to colon carcinoma, the National Cancer Institute notes that several studies have indicated that the toxic effects of several regimes of chemotherapy are the same for patients 70 years old and older as for younger patients.11
Colon carcinoma patients in the current study who lived in a zip code in the lowest quartile of per capita income were more than twice as likely as other patients to lack a recommendation for adjuvant therapy. In addition to the current study, only one4 of five3–7 studies found an association between sub-optimal cancer treatment and living in an area of relatively low income. However, a lack of standardization in the methods used to define income and to control for other potentially important variables, such as education, makes these disparate findings difficult to interpret. The one study that found an association used a continuous measure of median family income of the patient's census tract.4 This study did not include a measure of education in the analysis. Three of the studies finding no association between income and treatment for cancer used the median income in the patient's zip code analyzed as a dichotomous variable,5 a continuous variable,6 or a categorical variable in ten $10,000 increments.7 The other study finding no association between income and treatment for cancer divided census tracts into quartiles based on the proportion of households below the federal poverty level.3 All four of these studies included a measure of education based on the overall educational level in the area in which the patient lived. The findings in relation to education were not consistent. Two studies3, 6 found no association between sub-optimal cancer treatment and educational level; two studies5, 7 found lower educational levels associated with increased risk of sub-optimal treatment. However, in these latter studies, it was not clear if an initial association between income and treatment was attenuated by the inclusion of education in the final model, since neither the results of univariate analyses nor the criteria for inclusion in the final model were provided.
We found that colon carcinoma patients who had Medicare as the primary payer were more than twice as likely as patients with private insurance to lack a recommendation of adjuvant therapy. We identified several studies comparing treatment for colorectal carcinoma in health maintenance organizations and fee-for-service settings, but these studies are not comparable to the current study. One of these studies did not discuss insurance coverage.21 Two studies included Medicare patients, but did not distinguish between those with and without additional insurance coverage and did not compare Medicare patients to those with other forms of insurance.9, 22 Although people age 65 and older in the U.S. are almost universally covered by Medicare, in the current sample, almost 13% of people age 65 years and older had private insurance as the primary payer of hospital expenses.
Studies of differences in cancer treatment by race have primarily focused on differences between black and white cancer patients. These studies have not yielded consistent findings. Two studies of lung carcinoma2, 4 and one study of breast carcinoma8 concluded that black cancer patients received sub-optimal treatment compared to white patients. These differences were not found in another study of breast carcinoma6 nor in a study of followup treatment for colorectal carcinoma.3 The current study did not find differences in treatment between racial groups. However, the number of nonwhite patients was small, and it is likely that many of the patients reported as white in the WSCR were of other races. Therefore, we are not able to draw conclusions about potential differences in treatment by race. This issue may merit additional study given that the mortality rate for colorectal carcinoma among blacks in Washington is higher than that for whites.20
An external audit of the SEER region of the WSCR was conducted for cases diagnosed in 1995.23 This audit found an overall discrepancy rate of 2.0% for colorectal carcinoma. The audit found a 5.0% error rate in extent of disease such that correction of the error would result in a change of stage. A similar audit of 1996 cases originating outside the SEER region identified an error rate of 11.5% for stage of diagnosis for colorectal carcinoma patients.24 An unknown proportion of these errors would result in a change of stage. These audits did not include treatment and did not provide sufficient information to determine whether the errors in stage were differential or non-differential. However, we have no reason to believe that these or other abstracting errors are differentially associated with the socioeconomic and demographic factors considered in the current study.
The extent to which a recommendation for or receipt of adjuvant therapy is underreported in the WSCR bears further investigation. There are several factors that could contribute to underreporting, such as inconsistencies in abstracting medical records (such that no adjuvant therapy is indicated in instances where unknown should be reported), delays in decisions about the provision of adjuvant therapy, and receipt of adjuvant therapy in settings not associated with the facility in which the surgery occurred. We do not know the extent to which these factors may differentially affect reporting of adjuvant therapy in older patients, patients living in zip codes with lower per capita incomes, and those who have Medicare as their primary source of insurance.
While we partially accounted for contraindications to adjuvant therapy, it is likely that we did not identify all patients with contraindications. We were only able to identify patients with contraindications to 5FU from the hospitalization record at the time of surgery. Our ability to assess comorbidity was also limited to conditions that were reported in the hospital record. The extent to which older patients, patients living in zip codes with lower per capita incomes, and patients with Medicare as their primary source of insurance are likelier than other patients to have contraindications to adjuvant therapy or comorbidities that are not reported in the hospital record bears further investigation.
Since we excluded patients from federal and out-of-state hospitals, we are only able to generalize these results to patients treated in Washington state. We were unable to match 7.0% of the WSCR records to the CHARS dataset. There were no differences between the matched and unmatched records related to treatment or income group. A greater proportion of the unmatched records were for patients less than 65 years old and for patients with rectal carcinoma (5.9% in patients 65 and older versus 9.7% in those younger than 65; 4.3% and 11.4% for colon and rectal carcinomas, respectively), perhaps indicating more outpatient surgery for younger patients and those with rectal carcinoma.
We based income on the 1990 census and used relatively large geographic areas. Thus, we do not know how well the income variable reflects either 1996-1997 relative income in the patient's zip code or the patient's personal income. An analysis using per capita income from smaller geographic units, such as census blocks, may more accurately reflect the patient's income. However, we have no reason to suspect differential misclassification of relative per capita income based on treatment recommendations. Additionally, we did not assess the relationship between educational level and adjuvant therapy. Education is related to income and may be another important factor in the receipt of adjuvant therapy.
Studies indicate an overall reduction in mortality for surgery plus adjuvant chemotherapy compared to surgery alone for Stage III colon carcinoma in the range of 22–33%.25–28 These studies included patients aged 18 to 86 years old. One set of reports showed no significant differences between recurrence or survival in those over 60 years old compared to those 60 or younger at three28 or five years,25 although the P value for survival at five years approached statistical significance (0.064). The authors did not report whether age modified the effect of treatment on survival. Another study noted that three year survival was better for those under 65 years of age compared to older patients in both the treatment and control groups but did not report on whether there was reduced mortality in older patients compared to younger patients in the treatment group.26
Given the importance of adjuvant therapy in saving lives, additional study is needed to confirm or refute the association of a lack of a recommendation of adjuvant therapy with age, living in a lower income area, and having Medicare as the primary payer of hospital expenses. If these findings persist, it is necessary to delineate the relative importance of having a low income versus living in a low income neighborhood and why patients on Medicare do not seem to be receiving appropriate treatment in order to determine appropriate public health actions aimed at reducing disparities in treatment. A study of factors considered by physicians recommending treatment for colon carcinoma may be helpful in elucidating the role of socioeconomic and demographic factors in treatment recommendations.
These findings suggest that the experience of lower income and older patients and those relying on Medicare as their primary source of insurance is consistent with the findings of the Institute of Medicine's National Cancer Policy Board. The Board concluded that “for many Americans with cancer, there is a wide gulf between what would be construed as the ideal and the reality of their experience with cancer care.”29 The current study lends support to the Board's suggestion that Medicare could be an important vehicle in instituting changes to quality of care for cancer and that issues related to quality of care for cancer need to be addressed at both the state and national level. Hospitals could also be an important vehicle for change, since hospitals could use their cancer registry information to determine whether patients are getting proper treatment for Stages II and III rectal and Stage III colon carcinoma.
The current study shows that public health agencies have the ability to examine potential disparities in the provision of health care using administrative and surveillance databases. The goals of public health agencies include both assuring access to medical care and eliminating disparities in health related to age, race/ethnicity, and socioeconomic status. One method of achieving these goals is through reducing disparities in the provision of recommended medical procedures related to socioeconomic and demographic factors.