Access to care and stage at diagnosis for patients with lung cancer and esophageal cancer: analysis of the Savannah River Region Information System cancer registry data.

ABSTRACT

Background. Disparities have been observed in both the incidences of lung and esophageal cancers and the survival of those patients. Our goals were to determine if race was associated with stage of cancer at diagnosis, and to identify predictors of advanced-stage lung and esophageal cancers.

Methods. All cases of lung and esophageal cancer between 1991 and 1995 in the Savannah River Region Information System cancer registry were studied. Data were analyzed using logistic regression to identify independent predictors of advanced disease at the time of diagnosis.

Results. Among lung cancer patients, histology and distance to nearest hospital predicted diagnosis at an advanced stage. Residence in an area with a high proportion of Medicaid recipients was a predictor of advanced stage in esophageal cancer patients.

Conclusions. In this predominantly rural area, decreased utilization of health services was evident among older, poor, black, rural cancer patients. Further investigation involving prospective data collection from cancer patients is warranted.

**********

LUNG CANCER is the second most frequent malignancy and the leading cause of cancer-related deaths in both men and women. It was estimated that there would be 169,500 new cases of lung cancer and 157,400 lung cancer deaths in the United States in 2001.1 The burden of lung cancer is greater in the black than in the white population of the United States. The Surveillance, Epidemiology, and End Results (SEER) program reported that the age-adjusted incidence of lung cancer was 56.3 per 100,000 among whites and 72.4 per 100,000 among blacks in 1998. The 5-year relative survival for lung cancer between 1992 and 1997 was 14.8% among whites and 11.7% among blacks. (2) Cancer of the esophagus is less common than lung cancer but imposes an especially high burden on the black population. (3,4) It was estimated that there would be 13,200 new cases of esophageal cancer in the United States in 2001, and 12,500 esophageal cancer deaths. According to SEER, the age-adjusted incidence of esophageal cancer in 1998 was 3.6 per 10 0,000 among whites and 7.5 per 100,000 among blacks. The 5-year survival for esophageal cancer patients between 1992 and 1997 was 15.1% among whites and 9.0% among blacks. (2)

The two overarching goals of the US Healthy People 2010 program are to increase the quality and years of healthy life, and to eliminate health disparities. The Healthy People 2010 goal for cancer is to reduce the number of new cases, as well as the disability and death it causes. (5) The overall target is a 21% reduction in cancer deaths, focusing on the more prevalent and preventable cancers. The specific objective for lung cancer is a 22% reduction in deaths; no specific targets are included in Healthy People 2010 for esophageal cancer. (5) The overall goals of the National Cancer Institute (NCI) Strategic Plan to Reduce Health Disparties are to understand the causes of health disparities and to develop effective interventions to eliminate them (6); the objectives are to conduct cancer control and population research to elucidate the causes of cancer-related health disparities, and to define and monitor them. Our study addresses these national priorities.

The Savannah River Regional Health Information System (SRRHIS) is a population-based cancer registry jointly developed by the Medical University of South Carolina in Charleston, and Emory University in Atlanta, Georgia. All incident cancer cases in the 22 counties surrounding the Savannah River (10 counties in South Carolina, 12 in Georgia) occurring from 1991 through 1995 were identified. (4) The SRRHIS documented racial disparities in the incidence of lung and esophageal cancer. (4,7) The incidence of lung cancer is higher in young black men than in white men, but is lower in black women than in white women. The incidence of esophageal cancer is higher in black men than in white men and higher in black women than in white women. Thus, racial disparities reported nationally are also present in this predominantly rural region with a high proportion of African Americans.

For patients with lung or esophageal cancer, the stage of cancer at diagnosis and treatment received are the major determinants of outcome. Stage of cancer at diagnosis varies by the age, sex, and race of the patient, and by the histology of the cancer. In addition, personal illness behavior and health-system factors may influence access to health care, and could be related to the stage of cancer at diagnosis.

Access to care has been defined by the Institute of Medicine as "the timely receipt of personal health care services to achieve the best possible outcome." (8) Reduced access to care is believed to be associated with reduced use of health services, more severe illness, and worse health outcomes. In this study, we adapted a conceptual framework for the investigation of access to health care (8-10) and used data from the SRRHIS cancer registry and the US census to identify factors that may be associated with advanced stage of cancer at diagnosis. We hypothesized that socioeconomic factors (education, marital status, military service), geographic factors (distance to nearest hospital, number of primary care physicians), and health system factors (health insurance) are associated with having advanced-stage lung cancer or esophageal cancer at the time of diagnosis.

METHODS

Subjects

We identified for analysis a total of 3,477 subjects included in the SRRHIS registry with lung cancer (International Classification of Diseases-Oncology [ICD-O] primary-site codes 34.0-34.9) and 323 subjects with esophageal cancer (ICD-O primary-site code 15.0-15.9) diagnosed between 1991 and 1995. For patients with more than 1 cancer recorded in the registry, only data from the first-occurring cancer were included in the analysis, ensuring that all observations were independent.

Data Collection

The cancer registry provided information regarding subject age at diagnosis, sex, race, marital status, and county and zip code of residence, in addition to primary cancer site, histology, and stage. The county and zip codes were linked with data obtained from the 1990 US census regarding per capita income, education, military status, and patient residence in a rural or urban area. Although this type of linking is not a perfect substitute for primary data collection, it has been shown to be a useful method of controlling for confounding factors in other studies of cancer registry data. (11-14)

Per capita income is available directly from the census data files; methods to assign education, military status, and patient residence in a rural or urban area were developed for the study. For each zip code, census data provided tables of the number of persons by level of education, the number who were on active duty in the military or were veterans, and the number who lived in rural and urban areas. From these zip code tables, percentages of residents in the various education, military, and rural/urban groupings were calculated. These percentages thus reflected the probability of having some college education, being or having been active in the military, and residing in a rural area. Patients in the cancer registry were linked to these census data using their zip codes.

The distance to the nearest hospital for each patient was calculated using the MapInfo software package (MapInfo Corp, Troy, NY). All general medical/surgical hospitals in Georgia and South Carolina were geocoded, along with the centroid of each patient's zip code; then the distance from each patient's zip code centroid to each hospital was calculated. The distance to the nearest hospital for each patient in our analyses was defined as the minimum of these distances.

The Area Resource File (ARF) provided information for the calculation of the number of primary care providers per 10,000 persons. (15) Because ARF data were not available for each of the 5 years in which data was collected in the SRRHIS cancer registry, we chose the number of physicians per capita for the year 1994 as the covariate to be used in the analyses.

Cancer Stage

The SRRHIS cancer registry used the SEER staging classification, which categorizes cancers as either localized, regional by direct extension, regional to lymph nodes, regional by direct extension and to lymph nodes, regional not otherwise specified, distant metastases, or unstaged. For the analysis of potential demographic, geographic, and clinical variables that could be considered as independent predictors of distant metastatic disease, the stage of cancer at diagnosis was collapsed into 2 categories. Subjects staged as having distant disease were classified into one group, while those whose stage was localized or regional were classified in the other group. All subjects were included in the calculations of the descriptive statistics of the study sample, regardless of stage. Those subjects whose stage was unknown, however, were not included in the logistic regression.

Histologic Groupings

Histologic groups chosen for analyzing lung cancer were the same as those used in a recent analysis of the 1973 to 1996 SEER data. (16) The 5 lung cancer histology groups were small cell carcinoma, squamous cell carcinoma, large cell carcinoma, adenocarcinoma, and other. Because the stage at diagnosis was least likely to be distant for those subjects with squamous cell carcinoma, that was chosen as the referent category.

Histologic groupings chosen for esophageal cancer were the same as those reported in a paper by Devesa et al (17): adenocarcinoma, squamous cell carcinoma, and other. Because of the small number of those with the "other" histology (n = 12), however, this group of subjects was combined with the adenocarcinoma group for our final analysis.

Model Construction

Logistic regression models were constructed separately for lung and esophageal cancer. Initially, we included histology, age, race, sex, marital status, education, income, military status, physicians per 10,000 persons in county of residence, distance from residence to nearest hospital, and urban/rural status of residence as potential predictors of distant metastatic disease. Because of multicolinearity between several of the variables, however, the list was reduced. Income was highly correlated with education (correlation coefficient [r] = 0.84), as was urban/rural status with distance to nearest hospital (r = 0.62) and physicians per 10,000 persons in county of residence (r = -0.64). The final models for both lung and esophageal cancer therefore included histology, age, race, sex, marital status, education, military status, physicians per 10,000 people in county of residence, and distance from residence to nearest hospital. Using the logistic models, odds ratios (OR) and their respective 95% confidence intervals (CIs) were calculated.

RESULTS

Characteristics of the study sample are shown in Table 1. These descriptive statistics included all subjects, even those whose tumors were unstaged. There were 3,477 incident cases of lung cancer in the 22 SRRHIS counties from 1991 through 1995. The mean age at diagnosis was 66.4 years. Approximately two thirds of patients were male and three fourths were white. Squamous cell carcinoma and adenocarcinoma were the most frequent histologies; small cell carcinoma occurred in 15.4% of the patients. Distant disease at the time of lung cancer diagnosis occurred in 34.9% of cases. Between 1991 and 1995, there were 323 incident cases of esophageal cancer in the 22 SRRHIS counties. The mean age of patients with esophageal cancer was 64.8 years, similar to the mean age of patients with lung cancer. In contrast, approximately three fourths of patients with esophageal cancer were male and approximately 60% were black. Squamous cell carcinoma accounted for approximately two thirds of the cases of esophageal cancer. At the time of diagnosis, distant disease was found in 17% of the esophageal cancer cases. Compared with subjects having lung cancer, those with esophageal cancer were significantly less likely to be white, female, married, earn a high income, or have any college education. Patients with lung cancer were significantly more likely to be diagnosed with distant disease.

There were 3,286 patients with cancers of the lung or esophagus whose cancer stage was known at the time of diagnosis and 514 patients for whom insufficient information was available to assign a stage. Patients with staged disease differed from those with unstaged disease (Table 2). In general, patients with unstaged cancer were older, poorer, less likely to be white or married, and more likely to reside in a rural area with fewer primary care physicians and fewer Medicaid recipients. The unstaged lung cancer patients were more likely to have "other" histologic types of cancer.

Multivariate logistic regression analysis was done on the 3,040 incident cases of lung cancer with information on stage at diagnosis and the variables of interest (Table 3). As expected, histology was statistically significantly associated with advanced stage at diagnosis, with the highest risk associated with small cell carcinoma. After adjusting for histology, race was not associated with advanced stage (for nonwhite race, OR = 0.98; 95% CI, 0.76-1.26). The only variable associated with advanced stage of lung cancer was distance to nearest hospital (for every 10 mile increment, OR = 1.21, 95% CI, 1.04-1.44). Alternately, persons living > 10 miles from a hospital were 1.25 (95% CI, 1.00-1.57) times as likely to be diagnosed with advanced stage than those living < 10 miles away (P = .047). Age, sex, and marital status were not significantly associated with advanced stage of lung cancer at diagnosis. Residence in an area with a high proportion of persons with less than a ninth-grade education, veterans or active-duty military personnel, or Medicaid-eligible residents, and the number of primary care physicians per 10,000 population were associated with advanced stage of cancer at diagnosis.

Multivariate logistic-regression analysis was done using information from the 246 esophageal cancer patients with information on stage of cancer at diagnosis and the variables of interest (Table 4). Neither histology nor race were associated with advanced stage of esophageal cancer (for nonwhite race, OR = 0.73; 95% CI, 0.29-1.89). The only factor significantly associated with advanced-stage esophageal cancer at diagnosis was residence in a county with a large proportion of Medicaid-eligible residents. Similar to the findings for lung cancer, age, sex, and marital status were not associated with advanced stage of esophageal cancer. In contrast to lung cancer, however, residence in an area with a high proportion of persons with less than a ninth grade education, a high proportion of veterans or active duty military personnel, distance to the nearest hospital, and number primary care physicians per 10,000 population were not associated with advanced-stage esophageal cancer either.

DISCUSSION

This study explored the potential relationship of demographic, socioeconomic, and geographic factors with advanced stage of lung cancer or esophageal cancer at diagnosis. The SRRHIS cancer patients with unstaged cancer were more likely to be older, unmarried, black, poor, and residents in rural counties with fewer primary care physicians. Among the approximately 3,000 incident cases of lung cancer and 250 incident cases of esophageal cancer in the predominantly rural counties surrounding the Savannah River region, histology, as expected, was associated with advanced stage of lung cancer at diagnosis. After adjusting for histology, the only factors that were associated with advanced stage at diagnosis were distance to the nearest hospital for lung cancer patients and residence in a county with a high proportion of Medicaid recipients for esophageal cancer patients. None of the other patient-level variables (race, age, sex, marital status), or aggregate-level variables (education, military service, number of pr imary care physicians per 10,000 population) were significantly associated with advanced stage of cancer at diagnosis.

Our analysis has 2 major implications. First, the pattern of available information on cancer stage at diagnosis raises concerns about access to care. There was insufficient information on a substantial proportion of patients (approximate 12% of patients with lung cancer and 30% of patients with esophageal cancer) to assign the stage at diagnosis. This may reflect decreased access to care, patient preferences regarding diagnostic evaluation and treatment, provider practice styles, or incomplete abstraction of data from the medical record by the SRRHIS cancer registry staff. The high-level ascertainment of incident cases of cancers in the SRRHIS cancer registry (4) and the high level of success in identifying incident cases of cancer in persons who sought care across state boundaries in nearby hospitals or cancer centers (18) suggest that problems in obtaining available medical record data is an unlikely explanation for the incomplete information on stage of cancer at diagnosis. Unstaged patients were more like ly to be older, black, poor, and reside in rural areas with low numbers of physicians per capita. All of these traits are consistent with the characteristics of vulnerable patients who may have reduced access to cancer-care services. (19) Limited information on stage, how ever, may also reflect patient preferences for diagnostic evaluation, cancer treatments, or provider practice styles. Previous studies have documented that older age, distance from a cancer center, functional status, (20) marital status, (21) and patient attitudes toward treatment risks (22) are associated with treatment for lung cancer.

The second implication of our study is that, in contrast to findings from other types of cancer, none of the demographic variables were associated with advanced stage of cancer for lung or esophageal cancer patients whose stage of cancer at diagnosis was known. The only factors that were significantly associated with advanced stage of cancer at diagnosis were distance to nearest hospital for lung cancer patients and residence in a county with a high proportion of Medicaid recipients for esophageal cancer patients. Previous studies have documented that distance is associated with referral to university cancer centers for patients with lung cancer. (20) Health insurance was not associated with referral to a university cancer center but was associated with the treatment received. (23) Thus, for lung cancer, demographic and socioeconomic variables appear less likely to explain disparities in survival. Racial disparities in survival of patients with lung cancer appear to be explained by racial disparities in surge ry for patients with resectable disease. (24)

In contrast to our findings in patients with lung cancer, studies of patients with other cancers have documented a variety of demographic and socioeconomic variables associated with advanced-stage cancer at diagnosis. (13,14,25-30) In addition to cancer biology, stage at diagnosis may reflect patients' cancer knowledge, attitude, and beliefs; patients' illness behavior; ability to access cancer-care services; health system factors, such as health insurance; and geographic availability of providers and cancer services. All of these factors may themselves vary by patient age, sex, race, and socioeconomic status. (8-10) Access to preventive services may be a more important factor in stage at diagnosis for those cancers for which there is evidence of the effectiveness of preventive services in reducing incidence or mortality (eg, breast, cervical, and colorectal cancers). For these cancers, socioeconomic status, health insurance, and beliefs about cancer are important factors associated with stage at diagnosis. (14,25,29) The absence of a relationship between demographic and socioeconomic factors in advanced-stage lung and esophageal cancers in our study is, therefore, plausible. Our use of ecologic variables to supplement information available from the cancer registry, and the small number of cases of esophageal cancer suggests caution in interpreting our negative findings.

Although our study did not identify racial disparities in stage of lung and esophageal cancer, it is important to recognize that there are racial disparities in the incidence of both cancers. The highest lung cancer incidence occurs in blacks, (2,3) and the incidence of esophageal cancer (and mortality rates due to esophageal cancer) in coastal South Carolina is among the highest in the United States. (3,31) In the SEER program, the age-adjusted rates of lung and esophageal cancer are higher in the black than in the white population. (2) In the 22 SRRHIS counties, the incidence of both lung and esophageal cancer are also higher in the black than in the white population. (4) A recent analysis of lung cancer in the SRRHIS cancer registry showed a higher incidence of lung cancer in younger black men (less than age 55 years) but not in older black men; the incidence of lung cancer was lower in black than in white women. (7) Racial disparities in the incidence of lung cancer can be explained by education, income, and population density. (32) Socioeconomic differences in incidence of lung cancer can be explained, in part, by differences in smoking habits. (33,34) Racial disparities in incidence of lung cancer are consistent with historical patterns of cigarette smoking. (16) In a recent population-based case-control study of esophageal cancer, four major risk factors--low income, moderate/high alcohol intake, tobacco use, and infrequent consumption of raw fruits and vegetables--accounted for almost all of the squamous cell esophageal cancer in whites, and most of the excess incidence of esophageal cancer in black men. (35)

Racial disparities have also been observed in the survival of cancer patients; lung and esophageal cancer survival rates are lower for blacks than for whites. (2) In an analysis of patients with a diagnosis of lung cancer in the SEER program between 1985 and 1993 with resectable (stage I or stage II) cancer, the rate of surgery was 12.7% lower for black than for white patients. The 5-year survival rate was also lower for black than for white patients. For patients having surgery, however, blacks and whites had similar survival rates, as did black and white patients not treated surgically. (24) Thus, racial disparities in survival for patients with lung cancer appear to be explained by treatment. Surgery and other treatment modalities (radiation therapy and/or chemotherapy) may be the major factors in explaining disparities in outcomes of patients with cancer. Access to treatment for lung cancer may vary due to a variety of patient demographic, socioeconomic, and geographic factors. Greater distance from a canc er center, lower functional status, and age over 75 years have been associated with the use of a university cancer center for patients with lung cancer. (20) Social and economic factors influence the choice of lung cancer treatment. Marital status, medical insurance coverage, and proximity to a cancer treatment center were associated with treatment for lung cancer in patients from a rural, predominantly white population. Patients who were married or who had private medical insurance were more likely to receive surgery for lung cancer. Among lung cancer patients who did not have surgery, patients with private insurance were more likely to receive either radiation therapy or chemotherapy. (23)

Patients' attitudes toward risk and survival also influence their decisions about having surgery or radiation therapy for lung cancer. (22) Thus, for patients with lung or esophageal cancer, racial disparities in incidence of cancer may reflect disparities in lifestyle and health-related behaviors. Racial disparities in survival may reflect disparities in access to treatment and patient attitudes and decisions regarding treatment for cancer. Although several factors are associated with decisions about treatment, after adjusting for treatment, stage of disease, performance status, and clinical factors are the predominant determinants of survival. (36,37) The medical literature suggests that racial and other disparities in the outcomes of lung cancer may be more dependent on decisions to evaluate the stage of disease at diagnosis and decisions about treatment than on racial and other disparities in the outcomes of the treatments that are appropriate for the stage of cancer. We speculate that this may also apply to patients with esophageal cancer.

Analysis of population-based incident cases of lung and esophageal cancer from a predominantly rural area of 2 states that includes a high proportion of black residents is a strength of our study. A second strength is the investigation of 2 distinct cancers known to impose disparate burdens on the black population. A limitation of the study is that some data of interest, such as socioeconomic status, health insurance, distance to nearest hospital, and the number of primary care physicians per 10,000 population, were not directly available from the SRRHIS cancer registry and were obtained from the US census data for persons who reside in the same area. The inclusion of ecologic variables in an analysis introduces bias if the characteristics of the zip code or county differ systematically from the characteristics of the patients with incident cases of lung or esophageal cancer. A further potential limitation is the number of zip codes (132) and counties (22), which may limit the range of the independent variabl es of interest, and thus limit the power of the study to detect significant differences due to the factors for which the ecologic variables are used. Our analysis was appropriately modified to avoid problems of colinearity that often result when group-level variables are analyzed. (38)

We explored the relationship of race and other factors to advanced stage at diagnosis of lung and esophageal cancer. Among incident cases of lung and esophageal cancer, persons with unstaged disease were more likely to be older, black, poor, unmarried, and reside in a rural area with fewer primary care physicians and more Medicaid recipients. Appropriate treatment for these cancers is based on histology and stage of disease. Unfortunately, we do not have direct information on the utilization of health care services for our incident cases of lung and esophageal cancer. Cancer patients with unstaged disease most likely had decreased utilization of the health services necessary for ascertaining the anatomic location and extent of disease. It is likely that the patients with unstaged disease also had decreased use of health services for treatment of their cancer and lower survival rates. The decreased utilization may reflect patient and family preferences or provider recommendations, possibly due to associated co morbid medical conditions. These disparities raise concern about barriers to access of cancer-care services.

The patients with cancer staged at diagnosis were more homogeneous. Race was not significantly associated with advanced stage of lung or esophageal cancer at diagnosis. After adjusting for histology, only distance to nearest hospital for lung cancer patients and residence in an area with a high proportion of Medicaid recipients for esophageal cancer patients were associated with advanced stage at diagnosis. Our analysis was limited by the use of census data when individual patient-level data were not available. Additional prospective studies of factors associated with stage of lung and esophageal cancer are warranted.

TABLE 1

Characteristics of Study Sample, Including Patients With Unstaged Tumors

                                       Lung Cancer
Variable                               (n = 3,477)

Patient-level data
     Age at diagnosis (yrs) *
         Mean [+ or -] SD          66.4 [+ or -] 19.1
         Range                            26-100
     Race (% white) *                     72.6
     Sex (% male) *                       65.5
     Marital status *
         Living with spouse (%)           57.3
     Histology (%) *
         Squamous cell carcinoma          27.8
         Adenocarcinoma                   27.4
         Small cell carcinoma             15.4
         Large cell carcinoma             11.2
         Other                            18.2
    Stage at diagnosis (%)
         Local                            22.3
         Regional                         30.6
         Distant                          34.9
         Unstaged                         12.2
Zip-code-level data
   Per capita income ($) +
         Mean [+ or -] SD         12,736 [+ or -] 5,913
         Range                        3,379-46,353
   Education (%) +
         Completed <9th grade             12.7
            Range                         0-55
   Military status (%) +
         Active or veteran                29.7
            Range                         0-100
   Place of residence (%) +
         Residing in rural area           37.1
            Range                         0-100
   Distance to nearest hospital
     (miles) **
         Mean [+ or -] SD          5.65 [+ or -] 4.70
         Range                         0.02-30.12
County-level data
  Primary care physicians per
        10,000 people ++
        Mean [+ or -] SD            6.9 [+ or -] 3.2
        Range                             0-12
  Medicaid benefits (%) (ss)
        Recipients                        19.2
        Range                           8.4-33.9

                                    Esophageal Cancer
Variable                                (n = 323)

Patient-level data
     Age at diagnosis (yrs) *
         Mean [+ or -] SD          64.8 [+ or -] 11.7
         Range                            28-93
     Race (% white) *                     42.7
     Sex (% male) *                       76.2
     Marital status *
         Living with spouse (%)           47.4
     Histology (%) *
         Squamous cell carcinoma          67.2
         Adenocarcinoma                   24.1
         Small cell carcinoma               -
         Large cell carcinoma               -
         Other                             8.7
    Stage at diagnosis (%)
         Local                            27.6
         Regional                         31.9
         Distant                          17.0
         Unstaged                         23.5
Zip-code-level data
   Per capita income ($) +
         Mean [+ or -] SD         10,435 [+ or -] 6,134
         Range                        3,379-46,353
   Education (%) +
         Completed <9th grade             17.0
            Range                        0-39.7
   Military status (%) +
         Active or veteran                32.1
            Range                         0-82
   Place of residence (%) +
         Residing in rural area           41.2
            Range                         0-100
   Distance to nearest hospital
     (miles) **
         Mean [+ or -] SD          6.03 [+ or -] 5.70
         Range                         0.02-30.12
County-level data
  Primary care physicians per
        10,000 people ++
        Mean [+ or -] SD            6.8 [+ or -] 3.3
        Range                             0-12
  Medicaid benefits (%) (ss)
        Recipients                        19.4
        Range                           8.4-33.9

* Obtained from SRRHIS cancer registry.

+ Determined by linking census data and zip code of patient residence.

** Defined as the distance from the center of a patient's zip code to
the nearest hospital.

++ Obtained from the Area Resource File. (14)

(ss)Obtained from the South Carolina Office of Research and Statistics
and the Georgia Department of Medical Assistance.

SD = Standard deviation.
TABLE 2

Characteristics of Patients with Staged vs Unstaged Lung or Esophageal
Cancer at Diagnosis

                                            Staged
Variable                                  (n = 3,289)

Patient-level data
   Age at diagnosis (yrs)
       Mean [+ or -] SD               65.6 [+ or -] 10.6
   Race (% white)                            70.9
   Sex (% male)                              66.5
   Marital status (% living with
    spouse)                                  57.5
   Histology (%)
      Lung cancer
       Squamous cell carcinoma               28.7
       Adenocarcinoma                        28.9
       Small cell carcinoma                  15.9
       Large cell carcinoma                  11.7
       Other                                 14.8
  Esophageal cancer
      Squamous cell carcinoma                68.8
      Adenocarcinoma/other                   32.2
Zip-code-level data
 Per capita income (S)
    Mean [+ or -] SD                 12,622 [+ or -] 6,001
 Education (% completed <9th grade)          13.0
 Military status (% active or
  veteran)                                   29.7
 Place of residence (% rural area)           36.7
 Distance to nearest hospital
  (miles)
     (Mean [+ or -] SD)                5.7 [+ or -] 4.9
County-level data
 Primary care physicians per 10,000
  people
    Mean [+ or -] SD                   6.5 [+ or -] 3.5
 Medicaid recipients (%)                     19.3

                                           Unstaged
Variable                                   (n = 514)        P Value

Patient-level data
   Age at diagnosis (yrs)
       Mean [+ or -] SD               68.7 [+ or -] 11.3      <.001
   Race (% white)                            64.9              .006
   Sex (% male)                              65.6                NS
   Marital status (% living with                              .0004
    spouse)                                  49.1
   Histology (%)
      Lung cancer                                             <.001
       Squamous cell carcinoma               21.2
       Adenocarcinoma                        19.4
       Small cell carcinoma                  12.5
       Large cell carcinoma                   7.5
       Other                                 41.9
  Esophageal cancer                                              NS
      Squamous cell carcinoma                61.8
      Adenocarcinoma/other                   38.2
Zip-code-level data
 Per capita income (S)
    Mean [+ or -] SD                 12,007 [+ or -] 5,704     .032
 Education (% completed <9th grade)          13.7                NS
 Military status (% active or
  veteran)                                   31.1                NS
 Place of residence (% rural area)           42.7             .0009
 Distance to nearest hospital
  (miles)
     (Mean [+ or -] SD)                5.8 [+ or -] 4.4          NS
County-level data
 Primary care physicians per 10,000
  people
    Mean [+ or -] SD                   5.5 [+ or -] 3.5      <.0001
 Medicaid recipients (%)                     18.5              .012

SD = Standard deviation

NS = not statistically significant
TABLE 3

Results of Logistic Regression Analysis Predicting Distant Stage at
Diagnosis for Lung Cancer Patients

                                        Univariate Analysis

Variable                              OR            95% CI

Patient-level data
 Histology
   Squamous cell carcinoma           1.00          Referent
   Adenocarcinoma                    1.74 *        1.42-2.13
   Small cell carcinoma              3.58 *        2.82-4.51
   Large cell carcinoma              2.09 *        1.61-2.70
   Other                             2.58 *        2.03-3.27
  Race (nonwhite)                    0.93          0.79-1.09
  Sex (female)                       0.91          0.78-1.06
  Age at diagnosis                   0.99          0.99-1.00
  Marital status (married)           0.96          0.83-1.11
Zip-Code-level data
  Education (completed <9th grade)   0.79          0.35-1.75
  Military service (active or        1.18          0.88-1.59
   veteran)
  Distance to nearest hospital       1.19 *        1.03-1.39
   (miles)
County-level data
  Primary care physician per 10,000  1.00          0.98-1.02
   population
  Percent Medicaid-eligible          0.99          0.98-1.00

                                        Multivariate Analysis

Variable                              OR              95% CI

Patient-level data
 Histology
   Squamous cell carcinoma           1.00            Referent
   Adenocarcinoma                    1.77 *          1.44-2.18
   Small cell carcinoma              3.64 *          2.87-4.62
   Large cell carcinoma              2.11 *          1.63-2.73
   Other                             2.62 *          2.06-3.33
  Race (nonwhite)                    0.98            0.76-1.26
  Sex (female)                       0.88            0.64-1.20
  Age at diagnosis                   0.99            0.99-1.00
  Marital status (married)           0.91            0.77-1.06
Zip-Code-level data
  Education (completed <9th grade)   1.06            0.28-3.96
  Military service (active or        1.23            0.66-2.30
   veteran)
  Distance to nearest hospital       1.21 *          1.02-1.44
   (miles)
County-level data
  Primary care physician per 10,000  1.01            0.99-1.04
   population
  Percent Medicaid-eligible          0.99            0.98-1.01

* P < .05

OR = Odds ratio

CI = confidence interval.
TABLE 4

Results of Logistic Regression Analysis Predicting Distant Stage at
Diagnosis for Patients with Esophageal Cancer

                                        Univariate Analysis

Variable                               OR           95% CI

Patient-level data
   Histology
       Squamous cell carcinoma        1.00         Referent
       Adenocarcinoma/other           1.84         0.99-3.42
   Race (nonwhite)                    0.55         0.30-1.01
   Sex (female)                       0.73         0.34-1.56
   Age at diagnosis                   0.99         0.96-1.01
   Marital status (married)           1.87 *       1.02-3.46
Zip-code-level data
   Education (completed <9th grade)   0.10         0.01-1.91
   Military service (active or
    veteran)                          1.35         0.36-5.03
   Distance to nearest hospital
    (miles)                           0.98         0.93-1.04
County-level data
  Primary care physician per
    10,000                            0.98         0.89-1.08
   Percent Medicaid-eligible          1.03         0.98-1.08

                                        Multivariate Analysis

Variable                               OR             95% CI

Patient-level data
   Histology
       Squamous cell carcinoma        1.00           Referent
       Adenocarcinoma/other           1.35           0.64-2.84
   Race (nonwhite)                    0.73           0.29-1.89
   Sex (female)                       1.05           0.31-3.56
   Age at diagnosis                   0.98           0.95-1.01
   Marital status (married)           1.57           0.80-3.08
Zip-code-level data
   Education (completed <9th grade)   0.25           0.00-38.6
   Military service (active or
    veteran)                          1.32           0.12-14.2
   Distance to nearest hospital
    (miles)                           0.99           0.93-1.05
County-level data
  Primary care physician per
    10,000                            0.94           0.84-1.04
   Percent Medicaid-eligible          1.06 *         1.00-1.11

* P<.05

OR = odds ratio

CI = confidence interval.

References

(1.) Greenlee RT, Hill-Harmon MB, Murray T, et al: Cancer statistics 2001. CA Cancer J Clin 2001; 51:15-36

(2.) Ries LAG, Kosary CL, Hankey BF, et al, eds: SEER Cancer Statistics Review, 1973-1996. Bethesda, Md, National Cancer Institute, 1999

(3.) Bates D, Gaston JC, O'Brien P, et al: Carcinoma of the esophagus in South Carolina. J SC Med Assoc 1971; 67:453-436

(4.) Dunbar JB, Lackland DT, Liff J, eds: Cancer incidence Report, 1991-1995, Savannah River Region Health Information System. Charleston, SC, Medical University of South Carolina, 1999

(5.) US Department of Health and Human Services: Healthy People 2010: Understanding and improving Health and Objectives for Improving Health. Washington, DC, US Government Printing Office, 2nd Ed, 2000

(6.) National Cancer Institute: National Cancer Institute strategic plan to reduce health disparities. Available at: http://ospr.nci.nih.gov/healthdisprpt.pdf. Accessed June 5, 2001

(7.) Silverstein MD, Nietert PJ, Ye X, et al: Racial disparities in the incidence of lung cancer: the Savannah River Region Health Information System Cancer Registry, 1991-1995. J Health Care Poor Underserved (In press)

(8.) Institute of Medicine: Access to Health Care in America. Washington, DC, National Academy Press, 1993

(9.) Aday LA, Andersen R: A framework for the study of access to medical care. Health Serv Res 1974; 9:208-220

(10.) Mandelblatt JS, Yabroff KR, Kerner JF: Equitable access to cancer services: a review of barriers to quality care. Cancer 1999; 86:2378-2390

(11.) Hofer TP, Wolfe RA, Tedeschi PJ, et al: Use of community versus individual socioeconomic data in predicting variation in hospital use. Health Serv Res 1998; 33:243-259

(12.) Krieger N: Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based metholology. Am J Public Health 1992; 82:703-710

(13.) Roetzheim RG, Pal N, Tennant C, et al: Effects of health insurance and race on early detection of cancer. J Natl Cancer Inst 1999; 91:1409-1415

(14.) Lannin DR, Mathews HF, Mitchell J, et al: Influence of socioeconomic and cultural factors on racial differences in late-stage presentation of breast cancer. JAMA 1998; 279:1801-1807

(15.) United States Health Resources and Services Administration, Bureau of Health Professions, Office of Data Analysis and Management: Area Resource File. Rockville, Md, US Department of Health and Human Services, 1997

(16.) Wingo PA, Ries LA, Giovino GA, et al: Annual report to the nation on the status of cancer, 1973-1996, with a special section on lung cancer and tobacco smoking. J Natl Cancer Inst 1999; 91:675-690

(17.) Devesa SS, Blot WJ, Fraumeni JF Jr: Changing patterns in the incidence of esophageal and gastric carcinoma in the United States. Cancer 1998; 83:2049-2053

(18.) Lackland DT, Liff J, Haskill JL, et al: The effects of data exchange among geographic cancer registries on the racial diffferences in cancer incidence. J Registry Manage 1996; 23:84-87

(19.) Aday LA: At Risk in America: The Health and Health Care Needs of Vulnerable Populations in the United States. San Francisco, Jossey-Bass Publishers, 1993

(20.) Greenberg ER, Dain B, Freeman D, et al: Referral of lung cancer patients to university hospital cancer centers a population-based study in two rural states. Cancer 1988; 62:1647-1652

(21.) Goodwin J, Hunt W, Key C, et al: The effect of marital status on stage, treatment, and survival of cancer patients. JAMA 1987; 258:3125-3130

(22.) McNeil BJ, Weischselbaum R, Pauker SG: Fallacy of the five-year survival in lung cancer. N Engl J Med 1978; 299:1397-1401

(23.) Greenberg ER, Chute CG, Stukel T, et al: Social and economic factors in the choice of lung cancer treatment. N Engl J Med 1988; 318:612-617

(24.) Bach PB, Cramer LD, Warren JL, et al: Racial differences in the treatment of early-stage lung cancer. N Engl J Med 1999; 341:1198-1205

(25.) Samet J, Hunt W, Key C, et al: Choice of cancer therapy varies with age of patient. JAMA 1986; 255:3385-3390

(26.) Mor V, Masterson-Allen S, Goldberg R, et al: Relationship between age at diagnosis and treatments received by cancer patients. J Am Geriatric Soc 1985; 33:585-589

(27.) Greenfield S, Blanco D, Elashoff R, et al: Patterns of care related to age of breast cancer patients. JAMA 1987; 257:2766-2770

(28.) Mandelblatt J, Andrews H, Kao R, et al: Impact of access and social context on breast cancer stage at diagnosis. J Health Care Poor Underserved 1995; 6:342-351

(29.) Mandelblatt J, Andrews H, Kao R, et al: The late-stage diagnosis of colorectal cancer: demographic and socioeconomic factors. Am J Public Health 1996; 86:1794-1797

(30.) Riley GF, Potosky AL, Klabunde CN, et al: Stage at diagnosis and treatment patterns among older women with breast cancer: an HMO and fee-for-service comparison. JAMA 1999; 281:720-726

(31.) Fraumeni J, Blot W: Geographic variation in esophageal cancer mortality in the United States. J Chronic Dis 1977; 30:759-767

(32.) Baquet CR, Horm JW, Gibbs T, et al: Socioeconomic factors and cancer incidence among blacks and whites. J Natl Cancer Inst 1991; 83:551-557

(33.) Novotny TE, Warner KE, Kendrick JS, et al: Smoking by blacks and whites: socioeconomic and demographic differences. Am J Public Health 1989; 78:1187-1189

(34.) Pierce JP, Fiore MC, Novotny TE, et al: Trends in cigarette smoking in the United States: educational differences are increasing. JAMA 1989; 261:56-60

(35.) Brown LM, Hoover R, Silverman D, et al: Excess incidence of squamous cell esophageal cancer among US black men: role of social class and other risk factors. Am J Epidemiol 2001; 153:114-122

(36.) Stanley K Prognostic factors for survival in patients with inoperable lung cancer. J Natl Cancer Inst 1980; 65:25-32

(37.) Cassileth B, Lusk E, Miller D, et al: Psychosocial correlates of survival in advanced malignant disease? N Engl J Med 1985; 312:1551-1555

(38.) Conner MJ, Gillings D: An empiric study of ecological inference. Am J Public Health 1984; 74:555-559

RELATED ARTICLE: KEY POINTS

* Advanced stage of cancer at diagnosis may indicate limitations in access to oncology services and other health care services.

* Patients with unstaged disease were significantly more likely to be older, black, not living with a spouse, and reside in lower income rural areas with fewer primary care physicians.

* Distance to nearest hospital and histology were associated with advanced stage for lung cancer patients, and residence in an area with a higher proportion of Medicaid recipients was associated with advanced-stage esophageal cancer.

Correspondence to Daniel T. Lackland, DrPH, Medical University of South Carolina, Department of Biometry and Epidemiology, PO Box 250835, Charleston, SC 29425-8060. (Reprints not available.)

From the Center for Health Care Research and the Departments of Medicine, Pediatrics, and Biometry and Epidemiology, Medical University of South Carolina, Charleston.

Supported in part by grants from the US Department of Energy (Grant No. DE-FG09-91SR18217) and the US Department of Defense, Office of Naval Research (Grant No. N00014-99-10784).