Health Insurance and Stage at Diagnosis of Laryngeal Cancer: Does Insurance Type Predict Stage at Diagnosis? | Head and Neck Cancer | JAMA Otolaryngology–Head & Neck Surgery | JAMA Network
[Skip to Navigation]
Table 1. Characteristics of Patients Diagnosed as Having Early- vs Advanced-Stage Laryngeal Cancer (National Cancer Database, 1996-2003)
Characteristics of Patients Diagnosed as Having Early- vs Advanced-Stage Laryngeal Cancer (National Cancer Database, 1996-2003)
Table 2. Characteristics of Patients Diagnosed as Having T1, T2, T3, and T4 Laryngeal Cancer (National Cancer Database, 1996-2003)
Characteristics of Patients Diagnosed as Having T1, T2, T3, and T4 Laryngeal Cancer (National Cancer Database, 1996-2003)
Table 3. Adjusted Odds Ratios (ORs) for Patients With Advanced- vs Early-Stage Laryngeal Cancer (National Cancer Database, 1996-2003 [n = 51504])
Adjusted Odds Ratios (ORs) for Patients With Advanced- vs Early-Stage Laryngeal Cancer (National Cancer Database, 1996-2003 [n = 51504])
Table 4. Multivariate Analysis Adjusted Odds Ratios (ORs) for Patients With T2, T3, or T4 vs T1 Laryngeal Cancer (National Cancer Database, 1996-2003 [n = 51504])
Multivariate Analysis Adjusted Odds Ratios (ORs) for Patients With T2, T3, or T4 vs T1 Laryngeal Cancer (National Cancer Database, 1996-2003 [n = 51504])
Original Article
August 2007

Health Insurance and Stage at Diagnosis of Laryngeal Cancer: Does Insurance Type Predict Stage at Diagnosis?

Author Affiliations

Author Affiliations: Department of Otolaryngology, Emory University (Dr Chen), and Department of Health Services Research, American Cancer Society (Drs Chen, Halpern, and Ward and Ms Schrag), Atlanta, Georgia; and American College of Surgeons, Chicago, Illinois (Mr Stewart).

Arch Otolaryngol Head Neck Surg. 2007;133(8):784-790. doi:10.1001/archotol.133.8.784

Objective  To examine whether patients with no insurance or Medicaid are more likely to present with advanced-stage laryngeal cancer.

Design  Retrospective cohort study from the National Cancer Database, 1996-2003.

Setting  Hospital-based practice.

Participants  Patients with known insurance status diagnosed as having invasive laryngeal cancer at Commission on Cancer facilities (N = 61 131) were included. Adjusted and unadjusted logistic regression models analyzed the likelihood of presenting at a more advanced stage.

Main Outcome Measures  Overall stage of laryngeal cancer (early vs advanced) and tumor size (T stage) at diagnosis.

Results  Patients with advanced-stage laryngeal cancer at diagnosis were more likely to be uninsured (odds ratio [OR], 1.97; 95% confidence interval [CI], 1.79-2.15) or covered by Medicaid (OR, 2.40; 95% CI, 2.21-2.61) compared with those with private insurance. Similarly, patients were most likely to present with the largest tumors (T4 disease) if they were uninsured (OR, 2.92; 95% CI, 2.60-3.28) or covered by Medicaid (OR, 3.97; 95% CI, 3.56-4.34). Patients who were black, between ages 18 and 56 years, and who resided in zip codes with low proportions of high school graduates or low median household incomes were also more likely to be diagnosed as having advanced disease and/or larger tumors.

Conclusions  Individuals lacking insurance or having Medicaid are at greatest risk for presenting with advanced laryngeal cancer. Results for the Medicaid group may be influenced by the postdiagnosis enrollment of uninsured patients. It is important to consider the impact of insurance coverage on stage at diagnosis and associated morbidity, mortality, quality of life, and costs.

Laryngeal cancer is diagnosed in nearly 10 000 men and women in the United States annually and is among the most common cancers in the upper aerodigestive tract.1 Localized laryngeal cancer can be treated equally well with either radiation or conservation laryngeal surgery and has a favorable 5-year relative survival rate of 79%.2,3 In contrast, treatment options for advanced-stage laryngeal cancer are either total laryngectomy followed by radiotherapy or chemotherapy and radiation (chemo-RT). These treatment options produce equivalent survival, yet chemo-RT is associated with lower posttreatment morbidity and loss of function.4-6 Five-year survival rates among patients with advanced-stage laryngeal cancer (those with regional extension of tumor and/or lymph node involvement) are much lower than the survival rates for patients with early-stage disease, ranging between 32% and 44%.3 Furthermore, one-third of patients with advanced disease require a laryngectomy for salvage, resulting in a complete loss of voice and sometimes impairment of swallowing. These adversely affect quality of life with respect to nutrition, social functioning, and personal hygiene.7

Because stage at diagnosis is a key factor influencing prognosis and treatment, it is important to consider factors associated with late diagnoses. In laryngeal cancer, hoarseness, dysphagia, otalgia, and voice changes are all common symptoms that can lead to detection of early-stage disease. It is plausible that individuals without health insurance or with other barriers to health care access would be less likely to seek medical attention for these symptoms and thus present at a later stage with worse survival and fewer options for treatment.

To our knowledge, no prior study has investigated the relationship between stage at diagnosis of laryngeal cancer and insurance status. Information on insurance status is not routinely compiled or collected by population-based cancer registries. The National Cancer Database (NCDB) provides a unique opportunity to study the relationship between insurance status and stage at diagnosis because information on insurance status has been collected for all patients since 1996. The purpose of the present study is to examine the relationship between the patient's insurance status at the time of diagnosis and American Joint Commission on Cancer (AJCC) overall stage8 and T stage at presentation among patients diagnosed as having laryngeal cancer from 1996 to 2003 reported to the NCDB.


The NCDB is a national hospital-based cancer registry jointly sponsored by the American College of Surgeons and the American Cancer Society. Prior to 1997, submissions of records of patients with cancer to the NCDB were voluntary and open to all cancer facilities in the United States. Beginning in 1997, data collection was mandated as a requirement of the Commission on Cancer approved programs. Since 1999, approximately 75% of newly diagnosed cancers in the United States have been captured in the database.9

Cases of invasive squamous cell carcinoma of the larynx diagnosed from 1996 to 2003 were extracted from the NCDB using the appropriate International Classification of Diseases–Oncology, Third Edition (ICD-O-3) site and histology codes (n = 76 529). Data were abstracted using coding guidelines documented in the Registry Operations and Data Standards (ROADS)10 manual for cases diagnosed from 1996 to 2002 and the Facility Oncology Registry Data Standards (FORDS)11 manual for the diagnosis year 2003; the fourth, fifth, and sixth editions of the AJCC Manual for Staging of Cancer,8 and the second and third editions of the WHO International Classification of Disease for Oncology. For cases diagnosed prior to 2001, reported International Classification of Diseases–Oncology, Second Edition (ICD-O-2) tumor morphology codes were recoded to the ICD-O-3 codes.12 Site codes selected were C10.1 (anterior surface of epiglottis), C13.1 (aryepiglottic fold), C32.0 to C32.3 (glottis, supraglottis, subglottis, and laryngeal cartilage), C32.8 (overlapping lesion of larynx), and C32.9 (larynx, not otherwise specified [NOS]). The captured histology codes included 8045, 8051, 8052, 8070 to 8078, 8083, 8084, 8090 to 8094, 8097, 8123, 8147, 8560, and 8570 for squamous cell and/or basal squamous cell cancer.

From the patients with the correct site and histology codes (n = 76 529), only patients who were treated at the reporting institution (n = 72 262), who had stage I to IV at presentation (n = 65 541), and who were 18 years or older (n = 65 521) were included. Patients with unknown insurance status were excluded (n = 4390 [6.7%]), leaving 61 131 in the final study population. Institutional review board approval was not required for this study, which did not include individual identifiers.

The primary independent variable was the FORDS data element for primary payer/insurance at diagnosis. The FORDS codes were grouped into the following categories: Medicaid, Medicare (which included both Medicare alone and Medicare with supplement), uninsured (which included FORDS codes for not insured–NOS, not insured–charity write-off, and not insured–self-pay), other government funded plans (Department of Veterans Affairs [VA], Indian Health Service, Public Health Service, welfare, state-funded NOS, and federally funded NOS), and private insurance plans (health maintenance organization [HMO], preferred provider organization [PPO], managed care NOS, private insurance, Tricare/CHAMPUS [Civilian Health and Medical Program of the Uniformed Services], military, and insured NOS). The plans in the private insurance category were grouped together because these plans represent either privately purchased insurance (purchased by the individual, a family member, and/or employer) or insurance provided by the military that functions in a similar manner as private insurance (Tricare/CHAMPUS). Generally, all patients 65 years and older are eligible for Medicare enrollment regardless of comorbidity. Persons younger than 65 years who are either disabled or have renal failure or amyotrophic lateral sclerosis are also covered by Medicare.13 Since those with Medicare insurance who are younger than 65 years may be substantially different than those 65 years and older, the Medicare category was dichotomized by age (18-64 years and ≥ 65 years).

Other independent variables included sex (male and female), race (white, black, Hispanic, and other, which included Asian American/Pacific Islander and Native American), age (categorized as covariates into quartiles of 18-56, 57-64, 65-72, and ≥ 73 years), US census region of residence (West, Midwest, Northeast, and South),14 education (proportion in patient's zip code of residence without a high school degree based on national quartiles of the 2000 US census as ≥ 29%, 20%-28.9%, 14%-19.9%, and < 14%), and median household income in patient's zip code of residence based on national quartiles of the 2000 US census as less than $30 000, between $30 000 and $34 999, between $35 000 and 45 999, and $46 000 or more.

Type of treatment facility (community hospital, community cancer facility, and teaching/research facility) was also included as an independent variable because prior analyses have demonstrated variation in stage by facility type. The 3 types of treatment facilities are described by the Commission on Cancer. Community hospitals treat at least 300 cancer cases a year and have a full range of services for cancer care, but patients need referral for portions of their treatment. Community cancer centers are facilities that offer the same range of services as the community hospitals but have at least 750 annual cancer cases and conduct weekly cancer conferences. Teaching/research facilities differ from community cancer facilities in that the teaching/research facilities have residency programs and ongoing cancer research. Of the 39 National Cancer Institute (NCI) Comprehensive Cancer Programs, 29 participate in the Commission on Cancer approvals program and are included among teaching/research facilities in this study.

Two separate analyses were performed. The first analysis used a dichotomized-dependent variable describing stage at diagnosis, categorized as early (stages I and II) or advanced (stages III and IV).8 In the second analysis, the second dependent variable was the reported T stage at diagnosis, categorized as T1, T2, T3, and T4.8

Analyses were performed with SAS version 9.1 (SAS Statistical Institute, Cary, North Carolina). Statistical testing was conducted via the SAS procedure for logistic regression (PROC LOGISTIC). Initial (unadjusted) logistic regression models analyzed the likelihood of presenting with advanced vs early stage and of presenting at more advanced T stage based on each independent variable separately (ie, 1 at a time). Additional regression analyses controlled for all independent variables found to be statistically significant in these univariate models. Proportional logistic models were not justified for examining the T stage at diagnosis as the dependent variable because the score test result for the proportional odds assumption was significant (ie, regression lines for cumulative logits were not parallel). Therefore, we used a generalized logistic model (the SAS “glogit” link option for PROC LOGISTIC) to compare patients with T2, T3, or T4 stages at diagnosis separately with the referent T1 group. Patients with missing data for the significant covariates were excluded from the multiple logistic regression models, leaving 51 504 (84.3%) patients eligible for these analyses. We assessed whether the adjusted models (ie, including all significant independent variables) had a significantly better fit than the unadjusted models (ie, only including insurance status). Goodness of fit was tested by computing the change in deviance (Δ D) from the adjusted models to the unadjusted models and assessing the significance using the likelihood ratio statistic. Statistical significance was defined as P < .01. Marginal level of significance was defined as .01 < P < .05.


A total of 61 131 patients with laryngeal cancer were eligible for analyses. There were 32 665 (53.4%) with early-stage disease and 28 466 (46.6%) with advanced-stage disease. The distribution of T stage included 22 693 (37.1%) with T1 disease, 15 111 (24.7%) with T2 disease, 13 541 (22.2%) with T3 disease, and 9786 (16.0%) with T4 disease. Table 1 compares the characteristics of patients presenting with early- vs advanced-stage disease. Table 2 presents similar results, illustrating patient characteristics by T stage. Patients with advanced-stage disease or more advanced T stage were more likely to be uninsured or have Medicaid or other government-funded plans than were those with early-stage disease. The proportion of women having T4 disease (compared with T1 disease) increased compared with men; that is, 19.9% of patients with T4 lesions were female (an increase from 17.4% female among patients with T1 disease). Conversely, the proportion of men having T4 disease (compared with T1 disease) decreased; ie, 80% of all patients with T4 disease were male, a decrease from 82.6% male among patients with T1 disease. The proportion of patients who were black increased as T stage and overall stage increased. A higher proportion of patients who were younger than 65 years, lived in the South census region of the United States, and lived in zip codes with lower median incomes and lower proportion of high school graduates also had more advanced T and overall stage. Patients with advanced disease were more likely to be treated at a teaching/research facility.

Table 3 presents corresponding logistic regression results for analyses relating insurance type to the likelihood of presenting with advanced disease controlling for patient sex, age, race, treatment facility type, zip code–based education and income categories, and US census region. The likelihood ratio test statistic for the adjusted model was statistically significant (Δ D = 30 191.4 [P < .05]). After controlling for covariates, patients who are uninsured (odds ratio [OR], 1.97; 95% confidence interval [CI], 1.79-2.15) or who have Medicaid insurance (OR, 2.40; 95% CI, 2.21-2.61) were more likely to present with advanced-stage disease than those with private insurance. Patients with insurance through other government plans (OR, 1.34; 95% CI, 1.22-1.48) were also more likely to present with advanced-stage disease than were those with private insurance. Patients with Medicare insurance, regardless of age, were also more likely to present with advanced-stage disease than were those with private insurance, although the increased risk of advanced disease among Medicare patients was less than that among uninsured or Medicaid patients. Patients who were female, black, between ages 18 and 56 years, or resided in zip codes with low proportions of high school graduates or with median household annual income less than $46 000 were also at increased risk of presenting with advanced-stage laryngeal cancer. Patients who lived in the Midwest and South census regions were at marginally decreased risk of advanced disease. Patients treated at teaching/research facilities were more likely to present with advanced disease.

Table 4 presents the final logistic regression model predicting T-stage disease by insurance type while controlling for sex, age, race, treatment facility type, zip code level education and income, and US census region of the country. The likelihood ratio test statistic for the adjusted model was again statistically significant (Δ D = 34 933.3 [P<.05]). Similar to the results in Table 3, uninsured patients and those with Medicaid insurance were more likely than patients with private insurance to present at more advanced T stages, and this association was strongest for the most advanced stage (T4) (OR, 2.92 and 3.97, respectively). The relationship between other covariates (sex, race, age, treatment facility type, and socioeconomic factors) and advanced T-stage presentation was similar to the relationship between these covariates and presentation with advanced overall stage. One exception is that census region of residence was largely unrelated to likelihood of advanced T-stage presentation.

Because the lack of private insurance (ie, being uninsured or having Medicaid, Medicare, or other government insurance) or being a member of a racial/ethnic minority group (statistically significant only for blacks) were both associated with a greater likelihood of advanced disease at presentation (Table 3), we explored the interaction of insurance type and minority racial/ethnic status (data not shown). In general, most interactions between insurance type and race were not statistically significant; those that achieved statistical significance were marginal (.01 < P < .05). The odds of presenting with advanced disease were lower among younger Hispanics with Medicare (OR, 0.61 [P = .04]), Hispanics with insurance from other government plans (OR, 0.60 [P = .03]), and patients of “other race” with Medicaid (OR, 0.63 [P = .04]) than among whites with comparable insurance. Overall, we conclude that among patients with a particular type of insurance (or lack of insurance), differences by race have little impact on the likelihood of presenting with advanced stage.

We also explored the relationship between laryngeal subsites (supraglottic larynx, glottic larynx, and subglottic larynx) and stage at presentation. Fewer than 1000 cases of subglottic cases were available, and thus no meaningful relationships were found owing to lack of power. Again, all insurance types were strongly associated with stage of diagnosis for supraglottic cancer and glottic cancers with the exception of Medicare (age <65 years), which was not associated with stage of presentation for supraglottic subsites. A significant limitation of this subsite analysis is the difficulty with coding subsite of origin for laryngeal lesions (eg, transglottic cancers).


Our principal finding was that for patients diagnosed as having laryngeal cancer between 1996 and 2003 at American College of Surgeons Commission on Cancer–approved hospitals, health insurance status was an important predictor of advanced stage and larger tumor size. Compared with patients with private insurance, those who were uninsured or had Medicaid insurance had a higher likelihood of being diagnosed as having advanced stage and with T2, T3, and T4 disease than those with private insurance. Individuals with Medicare insurance also had increased risk of presenting with advanced disease and larger tumors compared with patients with private insurance, although this association was not as strong as that for uninsured and Medicaid patients. Other factors associated with more advanced disease at diagnosis were sex (women at higher risk), age (younger patients at higher risk), race (African Americans at highest risk), treatment facility type (teaching/research facilities at higher risk), and residence in a zip codes with a higher proportion of individuals without a high school degree and low household income.

Insurance type, sociodemographic characteristics, and other factors differed significantly between those with early-stage and advanced-stage laryngeal cancer. Our analysis controlled for sex, race, age, and treatment facility type in the logistic regression models to adjust for differences in patient mix between early- and advanced-stage cancer groups. Ecological data on education attainment and household income were included to control for area-level factors; individual-level socioeconomic characteristics were not available. Census area of residence was included to control for variations in health care access, cultural differences, or health care utilization in the different regions of the United States.15,16

In multivariate analysis, the type of health insurance remained the strongest predictor of stage at diagnosis and tumor size, with the OR for larger primary tumors (T stage) at diagnosis increasing in a stepwise fashion from T1 to T4 for all insurance types relative to private insurance. The relative risk of being diagnosed as having T4 disease was highest for uninsured and Medicaid patients (OR, 2.92 and 3.97, respectively). Individuals who were female, younger than 56 years, and black were also at higher risk of presenting with larger tumors than were their demographic counterparts.

Insurance type may be related to likelihood of having a regular health care provider and seeking care for symptomatic disease (ie, the delay between first seeking care and definitive diagnosis). Lack of health insurance is strongly associated with barriers to access for health care in other studies. For example, the National Health Interview Survey found that in 2002-2003, 46.7% of uninsured adults aged between 18 and 64 years had no usual source of health care, compared with 9.3% of adults with health insurance.17 In 2003, 38.1% of adults younger than 65 years with no health insurance reported no visits to physician's offices, emergency departments, and home visits in the past 12 months, compared with 12.8% of insured adults.17 Individuals without health insurance are also less likely to report use of cancer preventive services such as mammography, Papanicolaou testing, and colorectal cancer testing.17,18 A prior study that linked Florida cancer registry records with hospital discharge records to obtain insurance status found that patients without insurance coverage or with Medicaid insurance were more likely to be diagnosed as having late-stage cancer at a several sites (eg, colon-rectum, melanoma, breast, and prostate) than were patients with commercial indemnity insurance.19 One recent study reported that regardless of income, uninsured individuals use recommended health care services less often than those who are insured.20

Little has been published about factors that affect stage at diagnosis among patients with laryngeal cancer patients. Hoarseness, dysphagia, otalgia, and voice changes commonly appear in early disease; despite these early symptoms, some patients present with advanced stage.3 Investigators have reported that patients with fee-for-service insurance are more likely to use cancer screening tests18—other evidence of how insurance status may affect stage at diagnosis. The relationship between insurance and late-stage disease may also reflect a delay in seeking medical care by uninsured or underinsured symptomatic patients because of a lack of available qualified providers or other barriers. In such cases, by the time an uninsured or Medicaid recipient receives medical attention, the disease is advanced.

There are limitations to this study. First, some patients who were categorized as having Medicaid may have presented with no insurance coverage; the application for Medicaid coverage may have commenced on diagnosis, and coverage may be extended retroactively to the date of diagnosis. Previous studies suggest that patients who enroll in Medicaid at the time of cancer diagnosis may have more advanced-stage disease than do patients enrolling in Medicaid prior to diagnosis,21 although after controlling for stage at diagnosis, there were no significant differences in survival between early and late Medicaid enrollers.22 However, because information on date of Medicaid enrollment is not available in the NCDB, we were unable to adjust for this and may thus be undercounting the actual number of uninsured patients at the time of diagnosis. Similarly, no information is available on the type of insurance coverage prior to cancer diagnosis or the consistency of this coverage. Many individuals in the United States may experience periods of being uninsured, and these periods are generally longer among lower-income individuals.23 The impact of lack of consistent insurance coverage on cancer stage at diagnosis is unknown.

We also did not have information on the distance traveled to treatment facility, which may relate to access to care. We did not have individual level information on income, education, or other socioeconomic status measures and relied on estimates based on zip codes. We also lacked information on the time delay between presentation of symptoms and the diagnosis of laryngeal cancer. This information would have helped to define where lapses or potential gaps in access in care occur when individuals are not insured or underinsured.

In conclusion, our analyses provide the first assessment, to our knowledge, of the strong association between medical insurance and stage of laryngeal cancer at diagnosis among a large, generalizable cohort. Insurance coverage is a highly modifiable factor that affects not only tumor associated morbidity and mortality but also quality of life and economic costs.

Back to top
Article Information

Correspondence: Amy Y. Chen, MD, MPH, Department of Health Services Research, American Cancer Society, 1599 Clifton Rd NE, Atlanta, GA 30329 (

Submitted for Publication: October 24, 2006; final revision received February 27, 2007; accepted March 25, 2007.

Author Contributions: Dr Chen had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Chen, Schrag, Halpern, and Ward. Acquisition of data: Stewart and Ward. Analysis and interpretation of data: Chen, Schrag, and Ward. Drafting of the manuscript: Chen, Schrag, and Halpern. Critical revision of the manuscript for important intellectual content: Chen, Schrag, Halpern, Stewart, and Ward. Statistical analysis: Chen, Schrag, and Halpern. Administrative, technical, and material support: Chen, Stewart, and Ward. Study supervision: Chen and Ward.

Financial Disclosure: None reported.

Previous Presentation: This study was presented at the American Head and Neck Society Annual Meeting; August 20, 2006; Chicago, Illinois.

American Cancer Society Cancer Facts & Figures 2006.  Atlanta, GA: American Cancer Society Inc; 2006
Mendenhall  WMWerning  JWHinerman  RWAmdur  RJVillaret  DB Management of T1-T2 glottic carcinomas.  Cancer 2004;100 (9) 1786- 1792PubMedGoogle ScholarCrossref
Surveillance Research Program National Cancer Institute SEER*STAT software, version 6.2.3. Accessed 2006
Brizel  DMAlbers  MEFisher  SR  et al.  Hyperfractionated irradiation with or without concurrent chemotherapy for locally advanced head and neck cancer.  N Engl J Med 1998;338 (25) 1798- 1804PubMedGoogle ScholarCrossref
Forastiere  AAGoepfert  HMaor  M  et al.  Concurrent chemotherapy and radiotherapy for organ preservation in advanced laryngeal cancer.  N Engl J Med 2003;349 (22) 2091- 2098PubMedGoogle ScholarCrossref
Department of Veterans Affairs Laryngeal Cancer Study Group, Induction chemotherapy plus radiation compared with surgery plus radiation in patients with advanced laryngeal cancer.  N Engl J Med 1991;324 (24) 1685- 1690PubMedGoogle ScholarCrossref
Gillison  MLForastiere  AA Larynx preservation in head and neck cancers: a discussion of the National Comprehensive Cancer Network Practice Guidelines.  Hematol Oncol Clin North Am 1999;13 (4) 699- 718PubMedGoogle ScholarCrossref
Greene  FPage  DFleming  I  et al AJCC Cancer Staging Manual. 6th ed. Chicago, IL: Springer; 2002
Commission on Cancer Site by stage distribution of cases reported to the NCDB: diagnosis year 2003.  2004. Accessed 2006
Commission on Cancer Volume II: Registry Operations and Data Standards (ROADS). 1998. Accessed 2006
Commission on Cancer Volume II: Registry Operations and Data Standards (ROADS). 1996. Accessed 2006
Fritz  APercy  CJack  A  et al International Classification of Diseases for Oncology, ICD-O. 3rd ed. Geneva, Switzerland: World Health Organization; 2000
Centers for Medicare and Medicaid Services Medicare. 2006. Accessed 2006
United Status Census Report Census regions and divisions in the United States. 2000. Accessed 2006
Skinner  JWeinstein  JNSporer  SMWennberg  JE Racial, ethnic, and geographic disparities in rates of knee arthroplasty among Medicare patients.  N Engl J Med 2003;349 (14) 1350- 1359PubMedGoogle ScholarCrossref
Wennberg  JE The Dartmouth atlas of health care. Accessed 2006
National Center for Health Statistics Health, United States 2005 With Chartbook on Trends in the Health of Americans.  Hyattville, MD: US Government Printing Office; 2005
Potosky  ALBreen  NGraubard  BIParsons  PE The association between health care coverage and the use of cancer screening tests: results from the 1992 National Health Interview Survey.  Med Care 1998;36 (3) 257- 270PubMedGoogle ScholarCrossref
Roetzheim  RGPal  NTennant  C  et al.  Effects of health insurance and race on early detection of cancer.  J Natl Cancer Inst 1999;91 (16) 1409- 1415PubMedGoogle ScholarCrossref
Ross  JSBradley  EHBusch  SH Use of health care services by lower-income and higher-income uninsured adults.  JAMA 2006;295 (17) 2027- 2036PubMedGoogle ScholarCrossref
Bradley  CJGiven  CWRoberts  C Late stage cancers in a Medicaid-insured population.  Med Care 2003;41 (6) 722- 728PubMedGoogle Scholar
Bradley  CJGardiner  JGiven  CWRoberts  C Cancer, Medicaid enrollment, and survival disparities.  Cancer 2005;103 (8) 1712- 1718PubMedGoogle ScholarCrossref
Short  PFGraefe  DR Battery-powered health insurance? stability in coverage of the uninsured.  Health Aff (Millwood) 2003;22 (6) 244- 255PubMedGoogle ScholarCrossref