Adjuvant Chemotherapy Among Medicaid-Enrolled Patients Diagnosed With Nonmetastatic Colon Cancer

Abstract

Objective

It has been suggested that low-income populations may not receive adjuvant chemotherapy for colon cancer, although factors associated with its receipt have not been well-elucidated. This article describes the characteristics associated with chemotherapy among a Medicaid-insured population diagnosed with colon cancer.

Methods

A retrospective cohort design among 692 Medicaid-insured individuals diagnosed with regional colon cancer was conducted. Logistic regression analyses assessed patient, hospital, and community characteristics associated with chemotherapy. Data were derived from the N.C. Central Cancer Registry, N.C. Medicaid Claims, the American Hospital Directory, and the US Census.

Results

Forty-two percent received chemotherapy. Persons <65 years of age, diagnosis and treatment at different facilities, and living in a community with a higher percentage of persons in poverty were associated with receipt of chemotherapy. Individuals <65 years at time of diagnosis and had a score of 1< on the Charlson Comorbidity index were 16% less likely to receive chemotherapy as those less than 65 years of age with no comorbid conditions. Receipt of chemotherapy among those 65 to 74 and those 75 and older did not differ appreciably by comorbidity status.

Conclusion

Patient age was important in predicting who received adjuvant care, although the impact of comorbidity on chemotherapy was more pronounced among those <65 years of age.

Data from the National Initiative for Cancer Care Quality indicates that 78% of individuals diagnosed with stage II or III colon cancer receive recommended care.¹ Whereas most people diagnosed with colon cancer will undergo surgery, receipt of adjuvant therapy for regional colon cancer is considerably lower.^1,2 In a recent review of 22 published studies on the use of chemotherapy for stage III colon cancer, an estimated 39% to 71% received adjuvant chemotherapy.²

Etzioni et al found that older age and comorbidity are the most consistent predictors of underutilization of chemotherapy among patients diagnosed with regional disease. Twenty of 22 studies reviewed in this manuscript were population-based studies, with the majority using Surveillance, Epidemiology and End Results (SEER) cancer registry and Medicare claims linked data. Whereas data from SEER-Medicare provide an excellent resource for understanding national patterns of cancer care, the findings on age and comorbidity should be interpreted with some caution. SEER-Medicare includes persons aged 65 and older, which limits the interpretation of the finding that “older people” are less likely to receive chemotherapy.

Comorbidity is the most rational explanation for suboptimal treatment among the elderly, however, comorbidity does not fully explain treatment disparities between young and old.³ Ong et al reported that 95% of stage III patients with colorectal cancer under the age of 75 were referred for adjuvant therapy compared with only 67% of those 75 years of age or older, independent of comorbidity.⁴ Additionally, Townsley et al reported that elderly patients with cancer were less likely than younger patients to receive any cancer treatment.⁵ In a study of rural patients with colon cancer in North and South Carolina, 1 of 3 patients diagnosed with colon cancer ages 30 to 49 received adjuvant chemotherapy, comparable with those ages 75+ (44%). The group most likely to receive chemotherapy was ages 50 to 74 (87%). As Etzioni et al highlight, the majority of studies have not investigated the interplay between age and comorbidity and receipt of chemotherapy among patients with regional colon cancer, an important gap that we address in this study.

Another important shortcoming in the literature on colon cancer treatment is the over-reliance on the Medicare-insured population. In many ways, SEER-Medicare is an ideal database for exploring cancer care because the data are population-based, there are sufficient cancer cases for analyses, and cancer treatment differences cannot be attributed to health insurance status. However, Medicare is generally comparable with private insurance in terms of reimbursement (or it has been until recent years) and offers no information about socioeconomic status of the individual patient.

An alternative approach to population-based cancer research is to use Cancer Registry data linked to Medicaid claims. Such databases have shown that low-income women with breast cancer often experience later diagnosis, inadequate treatment, and poorer survival.^6–8 Comparable linked studies for colon cancer have yet to be published. However, such data offers the same advantage as SEER-Medicare (ie, a homogenously insured population using a population-based registry), while controlling for socioeconomic status at the individual level. All patients insured by Medicaid are means-tested and qualify for Medicaid under strict income-eligibility criteria.

A few studies have shown that individuals insured by Medicaid receive poorer quality care and have poorer colorectal cancer outcomes than those insured privately or through Medicare.^3,9–11 In a study by Harlan et al among elderly individuals who are insured by Medicaid, only 51.1% received cancer care according to guidelines.³ This is compared with 58.3% of the elderly covered by Medicare and 62.7% covered by private insurance. Bradley et al also found that Michigan Medicaid patients diagnosed with colon cancer were approximately half as likely as Medicare patients to initiate or complete chemotherapy.¹²

Thus, our goal in conducting this research study was to address the before-mentioned limitations in the literature on receipt of chemotherapy among poor individuals diagnosed with regional colon cancer by using a population-based cohort of the Medicaid-insured. We explore patient, hospital and community factors correlated with chemotherapy use among persons diagnosed with regional disease.¹³

MATERIALS AND METHODS

Study Design

The research team conducted a retrospective cohort study among the Medicaid population in North Carolina. The Wake Forest University Health Sciences and Davidson College Institutional Review Boards approved the research in accord with assurances filed with and approved by the US Department of Health and Human Services; both Boards waived the need for informed consent. In addition, the study underwent ethical review at the North Carolina Division of Medical Assistance and the North Carolina Central Cancer Registry.

Sample

North Carolina residents diagnosed with first primary, SEER-staged regional colon cancer (C180 C181 C182 C183 C184 C185 C186 C187 C188 C189 C199) in 1999–2002 insured by Medicaid were included, and were treated at the reporting facility. Cases were excluded if they met the following criteria: the case had no valid social security number (n = 1) or could not be matched to Medicaid claims data based on available information (n = 166); or were reported to the registry by an out-of-state facility, or had an unknown reporting facility (n = 14). The final sample size included individuals diagnosed with SEER local or regional colon cancer in 1999–2002 and who had a Medicaid claim 6 months after diagnosis (n = 692).

Data Sources

The North Carolina Central Cancer Registry is mandated by state law to register all incident cancer cases and first courses of treatment, and follows the requirements of the North American Association of Central Cancer Registrars. North Carolina Central Cancer Registry data are geocoded against TeleAtlas GC Geocode Layer (2004), using the ESRI ArcGIS geocoding engine.

North Carolina Medicaid covers all adults participating in cash assistance programs (eg, Work First Family Assistance, Supplemental Security Assistance), those 65 and older at 100% below the federal poverty level, and the disabled. All claims for payment from North Carolina Medicaid are filed via paper from physicians’ offices and hospitals, which are subsequently key-punched and converted into automated files. In North Carolina, Medicaid is almost entirely fee-for-service with one small managed care program. Medicare files in North Carolina crossed over to Medicaid during this study period, which ensures dual-eligibility patients have complete claims data in the Medicaid files.

Data Merge

The Medicaid eligibility file was matched to the North Carolina Central Cancer Registry data by social security number. Data were de-identified prior to data analysis. Using the facility identification number and hospital name, data were then merged with American Hospital Directory data. Poverty and residence data from the Census 2000 were matched by block group or zip code. The latter was used only when geocoded street address was missing (11%).

Measures

The main outcome of interest in this analysis was receipt of chemotherapy, which was defined using Medicaid claims. We considered chemotherapy to have been administered if, within 365 days of diagnosis, at least one of the following Healthcare Common Procedure Coding Systems procedure codes appeared in any claim or at least one of the of the following the international classification of diseases, 9th revision (ICD-9) codes appeared in any claim: Jxxx, 964xx, or 965xx series; G0345 to G0363; Q0083 to Q0085; RC331 to RC335; S9329; and W8222. ICD-9 codes V58.1x, V66.2, V67.2, and V99.25. Surgery was assessed using the cancer registry (yes if within 180 days of diagnosis) and radiation therapy (yes, if within 365 days since diagnosis) from Medicaid claims.

Potential factors associated with the outcome were grouped into patient, health services, and community characteristics.

Patient Characteristics

Patient characteristics derived from the cancer registry included: age at diagnosis (<65, 65–74, ≥75 or <65 and ≥65, depending on sample size), sex (male, female), and race (white, black, other race). Precancer comorbidity was defined using diagnostic ICD-9 codes from Medicaid claims 1 year prior to cancer diagnosis. Each individual was assigned a Charlson Comorbidity Index score, which is a well-established measure of mortality risk for comorbid conditions based on ICD-9codes.^14,15 Charlson et al developed the score to evaluate prognosis based on age and comorbid conditions. With each increased level of the comorbidity index, the cumulative mortality attributable to comorbid disease increases in a step-wise fashion. All individuals were assigned a noncancer Charlson Index score based on hospital assigned ICD-9 codes from Medicaid/Medicare billing claims.

Comorbidity was categorized into 0, 1+ based on the distribution of the data.

Health Services Characteristics

The health services characteristics included total surgery volume (divided into tertiles) based on surgical volume distribution from the American Hospital Directory; whether the reporting facility was a member of the University Health Consortium (yes, no); and whether patients were diagnosed and treated at the reporting facility (class of case 1) or patients diagnosed elsewhere but treated at the reporting facility (class of case 2).

Community Characteristics

Poverty was determined by downloading census block group (88%) match followed by 5-digit and 3-digit zip code level data (12% match) from the US Census and then merged with the geocoded street address using geographical identifier. For analytical purposes, poverty was categorized into tertiles. Urban residence was determined by merging patients’ geocoded street address data and urban area data using ArcMap (Version 9.2, ERSI, 2006). The resulting dichotomous (yes, no) data were exported and converted to a SAS data for analysis.

Statistical Analysis

Descriptive statistics of patient, health services, and community characteristics and treatment were computed. The bivariate relationship between patient (age, comorbidity, race, gender), health services (class of case, tertiles of hospital surgery volume, hospital a member of a university consortium), and community characteristics (tertiles of poverty, urban residence) and receipt of chemotherapy was assessed using a Cochran-Armitage test for trend. Receipt of chemotherapy was regressed on these same characteristics using logistic regression modeling. Odds ratios and 95% confidence intervals were computed from age-adjusted models (model 1) and multivariable models (model 2). Multilevel models were considered to account for the correlation of the persons within the same hospital and geographical location. However, these models could not be fit due to the small numbers of participants within each cluster, and were therefore abandoned. A 2-sided alpha level of 0.05 was used to indicate statistical significance. All data were analyzed using SAS (version 9.1, Cary, NC).

RESULTS

Characteristics of the Sample

There were 692 Medicaid-insured individuals diagnosed with regional colon cancer between 1999 and 2002 (Table 1). Almost all patients were treated with surgery (98%), whereas only 42% received chemotherapy. The sample was approximately 66% female, 42% black, and 26% aged under 65 years and 45% aged 75 years or older. Fifty-seven percent had at least one comorbid condition. Most of the patients were diagnosed and treated at the same facility (92%) and 19% were treated at a university health consortium hospital. About 49% lived in urban areas.

TABLE 1.

Characteristics of the Sample of Medicaid-Insured Individuals Diagnosed With Regional Colon Cancer Between 1999–2002 (n = 692)

	n	%
Treatment received
Surgery	679	98.3
Chemotherapy	293	42.3
Radiation	37	5.3
Patient characteristics
Female	454	65.6
Race
White	386	55.8
Black	288	41.6
Other/UK	18	2.6
Age
<65 yr	177	25.6
65–74 yr	206	29.8
≥75 yr	309	44.7
Charlson comorbidity index
0	295	42.6
1+	397	57.4
Hospital characteristics
Surgery volume
Lowest tertile	233	33.7
Middle tertile	226	32.7
Highest tertile	233	33.7
Class of case
Diagnosed and treated at reporting facility	634	91.6
Diagnosed elsewhere, treated at reporting facility	58	8.4
Treated at university health consortium facility	131	18.9
Community characteristics
Poverty
Lowest tertile	236	34.1
Middle tertile	233	33.7
Highest tertile	223	32.2
Urban	336	48.6

Factors associated with receipt of chemotherapy in unadjusted models include: age less than 65 years (P < 0.001, compared with those 65–74 and ≥75 years), having a Charlson Comorbidity score of 0 (P = 0.005, vs. 1 or greater), being male (P = 0.03), diagnosis of cancer at one facility, but first course of therapy at another facility (P < 0.001 vs. diagnosis and treatment at the same facility), and living in the highest tertile poverty region (P = 0.008, versus living in the middle or lowest tertile of poverty).

In the multivariable model, persons 65 to 74 years old and those 75+ were significantly less likely to receive chemotherapy compared with those under 65 years old (Table 2). Diagnosis of cancer at one facility and receipt of first course of therapy at another facility remained significant. These individuals were 57% less likely to receive chemotherapy after adjusting for age and other covariates. Living in the highest tertile of poverty also remained significant, with those individuals having an odds of receiving chemotherapy 2 times greater than those in the lowest poverty tertile.

TABLE 2.

Factors Associated With Receipt of Chemotherapy Among Medicaid-Insured Individuals Diagnosed With SEER-Staged Regional Colon Cancer Between 1999–2002 (n = 674)^*

	Model 1		Model 2
	Age-Adjusted		Multivariable
	OR	95% CI	OR	95% CI
Patient characteristics
Female	1.12	0.78–1.62	1.15	0.78–1.69
Race
Black (vs. white)	1.07	0.76–1.52	0.88	0.60–1.30
Age (vs. <65)
65–74 yr	0.47	0.31–0.71	0.48	0.30–0.75
≥75 yr	0.10	0.06–0.15	0.10	0.06–0.16
Charlson comorbidity index
0 (vs. 1+)	1.07	0.76–1.52	1.12	0.78–1.60
Hospital characteristics
Surgery volume (vs. highest tertile)
Lowest tertile	1.19	0.79–1.81	1.15	0.73–1.83
Middle tertile	1.38	0.91–2.09	1.33	0.84–2.12
Class of case (diagnosis and treatment at same facility)	0.41	0.22–0.79	0.43	0.22–0.85
Treated at university health consortium facility	0.99	0.65–1.52	1.03	0.63–1.68
Community characteristics
Poverty (vs. lowest tertile)
Middle tertile	0.96	0.63–1.45	0.93	0.60–1.44
Highest tertile	1.81	1.19–2.74	2.00	1.25–3.20
Rural	1.12	0.79–1.57	1.25	0.85–1.83

^*Dropped 18 cases (“other” race) in the multivariable model due to small sample size. Results reported are from logistic regression modeling.

Because of the potential moderating effect of age on the relationship between comorbidity and receipt of chemotherapy we also explored the bivariate effect of age and comorbidity on receipt of chemotherapy (Table 3). Individuals who were under age 65 at the time of diagnosis and had a score of 1+ on the Charlson Comorbidity index, were 16% less likely to receive chemotherapy as those less than 65 years of age with no comorbid conditions. Receipt of chemotherapy among those 65 to 74 and those 75 and older did not differ appreciably by comorbidity status.

TABLE 3.

Percentage of People Receiving Chemotherapy Diagnosed With Regional Colon Cancer by Age and Comorbidity Status

	Age
Comorbidity	<65	65–74	75+
Yes (1+)	89 (62.9%)	99 (55.6%)	209 (18.7%)
No (0)	117 (75.2%)	78 (47.4%)	100 (18.0%)

DISCUSSION

Whereas the majority of Medicaid insured persons received surgery (98%), less than half of individuals diagnosed with regional disease received adjuvant chemotherapy. According to a recent review of 12 articles comparing factors contributing to chemotherapy usage, rates range from 39% to 71%, with recent reports estimating 48% to 62%.² Thus, whereas the North Carolina Medicaid population appears to have abysmally low rates of chemotherapy utilization, rates are similar to other published studies.

Overall, we found that older age and greater comorbidity were significant correlates with receipt of chemotherapy in the unadjusted models, consistent with prior literature.^16–18 However, in this population the effects of comorbidity was only seen in those aged <65 years. Comorbidity did not influence receipt of chemotherapy use in patients 65 and older, which is a surprising finding and an important contribution to the literature.

This may indicate that comorbid conditions are more likely to influence treatment decision-making regarding adjuvant therapies in younger patients compared with older patients. This raises important questions for evaluation and treatment of older adults with colon cancer. For example, are older adults presenting with comorbidity or functional impairment not measured by the Charlson Comorbidity Index? This suggests that improved assessment measures are needed for clinical trials involving older adults. Or, is decision making influenced predominantly by chronologic age rather than comorbidity? A better understanding of this issue is needed to evaluate and potentially improve treatment decision-making for this population.

Racial/ethnic disparities in receipt of chemotherapy were not observed in this study which controls for socioeconomic status by including only persons insured by Medicaid. Our findings are consistent with a study which linked California Cancer Registry and discharge records, controlling for insurance status.¹⁹ However, race/ ethnic differences have been observed in SEER-Medicare and other population-based studies that have relied on community-level characteristics to control for SES.^16–18 Medicaid populations tend to be diverse, typically over-representing minorities in their sample, allowing for sufficient power to test for racial/ethnic differences.

Hospital and community characteristics included in this study did little to independently explain receipt of chemotherapy after adjusting for age. Class of case and poverty were the only nonpatient centered correlates of chemotherapy in the age-adjusted and multivariable models for regional disease. The “Class of Case” variable in the models suggest that individuals who are referred elsewhere for treatment rather than remaining the local community were more likely to receive guideline-consistent care. It is possible that individuals from high poverty areas are traveling to communities with greater health care resources and are, subsequently, getting more guideline-consistent adjuvant therapy. The finding that those living in communities with the highest tertile of poverty were more likely to get chemotherapy appears to support this hypothesis. However, these findings must be interpreted cautiously given the small percentage of the population with regional disease (8.4%) defined as class of case “2.”

This study adds to the current literature by focusing on treatment patterns among low-income populations diagnosed with colon using a population-based cohort of registry and Medicaid claims data. The use of registry-claims databases are well-established (SEER-Medicare, for example) and Cancer registries are potentially the best source of cancer care data for population research because of their information on disease status and nearly complete ascertainment of incident cancer cases following national standards for cancer registry data common elements and formats.^20,21 Further, these data allow us to investigate treatment for persons under age 65 and to evaluate the influence of comorbidity by age.²

Although there are advantages to using such data there are limitations that must be acknowledged. First, the data are limited to the demographic characteristics of the Medicaid-insured. In North Carolina, those insured by Medicaid are predominately white and African American, which limits generalizability to other minority populations. Thus, inferences about cancer care for non-African American minorities cannot be drawn from the results. Second, interpretations in the differences in treatment cannot be attributed to patient preferences or provider recommendation.

The study is also limited by its external validity due to inclusion of only one state in the analysis. State-specific analyses are preferred in Medicaid-based research given that Medicaid is partially state-funded and largely state-controlled insurance for the poor. Differences exist across states in terms of Medicaid eligibility and administration. However, the emerging literature on cancer disparities among Medicaid recipients in multiple states suggests that such studies have collective value.^9–12 When considered together, studies of Medicaid-insured cancer care offer important insight into the quality of care that Medicaid patients receive irrespective of the state in which a person resides.

Overall, we believe the most important findings of this research are 3-fold: (1) the very low utilization (42%) of chemotherapy among the Medicaid population diagnosed with regional colon cancer, (2) the strong independent relationship of age to receipt of chemotherapy regardless of comorbidity status in those over 65 years, and (3) the role of comorbidity among those <65 years on whether adjuvant therapy is administered for those with regional disease. These contributions have implications for treatment decision-making and referral patterns for physicians who care for individuals diagnosed with colon cancer.