Project 1: Hispanic Neighborhoods and Cancer Risks and Outcomes
Methodology
The principal objective of our study is to explore the relationship between social environments and cancer risks and outcomes for Hispanics. We will begin with an extensive secondary analysis of several data sets that are linked to files with information about social, economic and health care context. The purpose of this analysis is to identify specific mechanisms that link community structures and environments with cancer outcomes for Hispanics. The resulting findings will be published in refereed professional journals. They will also inform additional analyses and new data collection by identifying promising hypotheses for future investigation about the precise aspects of Hispanic institutions and behaviors that produce both advantageous and disadvantageous health outcomes for Hispanics.
The first phase of our research (Aim 1) uses data from the SEER cancer registry and vital registration death data linked to census data at the tract level to investigate the relationship between tract concentration of Hispanics and cancer stage, incidence, mortality and survival for the Hispanic population. The second phase examines the relationship between neighborhood environments and health risks and behaviors for Hispanics using a census-linked National Health and Nutrition Examination Survey (NHANES 3) file (AIM 2). Finally, the third phase of the research examines pathways linking neighborhood environments and mortality using data from the National Health Interview Survey-National Death Index file that has been linked to a census contextual file (Aim 3).
Conceptual Model - Multilevel Determinants of Health.
Variables of interest are included in five broad categories (Table 1). The final column lists final outcomes of interest. Our focus will be on the incidence of cancer, survival after diagnosis, and mortality.
Table 1. Conceptual model*
| Predisposing Factors | Stressors & Resources | Behavioral | Physiologic | Outcomes |
|---|---|---|---|---|
Community Environment
Personal Characteristics Age Gender Immigrant Status Education Social Class Acculturation |
Social support br /> [Stressors] [Life events] [Structural strains] Access to Care Insurance coverage Usual source of care |
Health Behaviors Diet Smoking Drinking Exercise Psychological health |
[Physiologic mediators of stress] Blood levels of nutrients [Carcinogens] [Genetic factors] Co morbidity Obesity Tumor size and stage at diagnosis Tumor histology |
Disease (cancer) Survival after diagnosis Site-specific cancer mortality All-cause mortality |
*Factors that will not be measured in the proposed analyses are shown in brackets.
Individual Level Datasets
SEER provides data on cancer incidence, as well as stage at diagnosis, treatment and survival. The vital registration data will give us complete coverage for the states of California and Texas, but are limited in outcomes to cancer morbidity and all-cause mortality rates.
NHIS provides more extensive data on personal characteristics than does SEER, including immigrant status, years in U.S., education, income and acculturation. The outcome with that data set will be cancer prevalence.
HANES data allow us to examine how a large variety of health behaviors and health characteristics vary by neighborhood context.
SEER-Cancer Registry. The Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute collects information on cancer incidence and survival for selected geographic areas in the United States. SEER cancer registry data are distinguished by a high rate of case ascertainment within area. Vital Registration Mortality-Texas and California, 1988-1992; 1998-2002. Mortality detail files for states of Texas and California report underlying cause of death. Deaths resulting from malignant neoplasm are recorded for specific sites using ICD9/10 codes, with detail about age, sex and Hispanic origin. Deaths are reported for census tract of residence at the time of death.
National Health and Nutrition Examination Survey (NHANES 3). NHANES 3 is a National Center for Health Statistics survey that includes 4,500 cases of Hispanic adults in the target age range for our study of persons 25 and older. The majority of this sample consists of Mexicans Americans, who were over-sampled. The survey protocol for NHANES includes an extensive survey instrument covering prevalent medical conditions, diet, health-related behaviors, a medical examination, and laboratory assays of blood and other physical specimens (Table 2).
National Health Interview Survey-National death index link file (NHIS-NDI).
The NHIS is an annual survey administered through in-person interviews to more than 100,000 persons living in more than 30,000 households. The data are collected for the National Center for Health Statistics (NCHS).The sample is representative of the civilian, non-institutional population of the United States, with an over sample of African Americans and Hispanics. NHIS is the primary source of information about health status and disability for this population.
Table 2. Variable Domains, Measures and Values for NHANES 3
| Variable Domain | Measure | Values |
|---|---|---|
| Demographic strata | Age Gender |
Continuous Male/Female |
| Assimilation | Immigrant status Years in U.S.(Foreign only) Age at migration (Foreign only) Language Acculturation (Use) |
United States/Foreign born Years (continuous measure) Age <15/age 15-54/age 55+ Spanish only/English and Spanish/Spanish only |
| Socio-economic status | Years of Schooling Household Income |
0-6 years/7-11years/12+ years $0-$9,999/$10,000-$19,999/$20,000 |
| Residential stability | Years at current address Years in current city |
0-4/5-19/20+ 0-4/5-19/20+ |
| Health status | Self-rated health | Excellent or very good/good/fair or poor |
| Health care access | Health insurance last month Usual place of care Usual physician |
Priv. Only; Pub. only; Priv.+Pub; Other; None Yes/no Yes/no |
| Risk behavior | Leisure Physical Activity per week Ever smoked Current smoker Amount of alcohol use Current alcohol use Age at first intercourse (women) Current sexual activity Weight/body mass index |
0/1/2-3/>3 Ever smoked 100 cigarettes/no Ounces/wk Had 12 drinks in past year/no <15,15-17;18+ Number of partners past year, 0,1,2,3+ <22 / 22-30 / 30-35 / 35+ |
| Social support | Marital status Community support |
Currently Married/Unmarried Continuous measure 0-20 of contact frequency by type |
| Affective disorders | DSM-III diagnosis- (Ages 25-39) | Yes/no |
| Diet | Fiber Ascorbate Folate Total fat Saturated fat Food Insufficient |
Grams per day mg per day mg per day Grams per day Grams per day Yes/no |
| Blood markers | Ascorbate LDL cholesterol homocysteine |
mg/dl mg/dl mcmol/L |
Contextual Data Sets.
The focus of our analyses is the relationship between the characteristics of neighborhood environments and individual behavior and health. We explore these relationships by linking primary data sets containing information about individuals to aggregate-level data sets containing information about characteristics of areas.
United States Census, Summary Files 1,2,3 (1990,2000). Census summary files contain area-referenced data about a wide variety of socio-economic indicators. Summary files 1 and 2 contain data about the age, sex, race and Hispanic origin of the population from the 100% decennial enumeration. Variable domains are listed in Table 3.
Table 3. Variable domains and Measures in United States Census Summary Files 1,2,3
| Domains | Measures |
|---|---|
| Age and Sex | Population counts for age/sex strata |
| Household Structure | % single-parent families; % persons live alone |
| Race and Hispanic Origin | % Hispanic (by sub-group; % White, % African-American of Non-Hispanic |
| Language | % Linguistically-isolated |
| Immigrant status | % foreign-born; % recent foreign-born |
| Migration | % lived in same house 5 years ago (residential stability) |
| Education | % less than 9 years; % high school graduates; % college graduates |
| Employment | % of males/females in labor force; % males/females unemployed |
| Occupation and Industry | % in major occupational/industry categories |
| Income and Poverty | Median income; median income in relation to area median; % in poverty; % on public assistance; Income inequality (GINI index) |
| Home ownership and renting | % of households own vs. rent their home |
| Housing unit characteristics | % with structural limitations (incomplete kitchen, plumbing) |
| Vacancy rates | % housing units not occupied |
| House values and rents | Median gross rent; median house value |
Dartmouth Health Atlas. The Dartmouth Health Atlas Data Viewer contains area-referenced data about physician and hospital resources for the Medicare-enrolled population (Table 4). This data set will be used to measure the availability of physician and hospital resources for the SEER analysis of cancer survival. The area units in the Atlas database are 3,436 hospital service areas that represent local health-care markets for community-based in-patient care.
Table 4. Key medical care resource variables from the Dartmouth Health Atlas
| Variables | Primary Source |
|---|---|
| Acute care medical/surgical hospital beds per 1,000 persons | American Hospital Association Annual Survey |
| Full-time equivalent hospital nurses per 1,000 persons | American Hospital Association Annual Survey |
| Physicians per 1,000 persons | AMA/AOA physician master files |
| Primary care physicians per 1,000 persons | AMA/AOA physician master files |
| Oncologists per 1,000 persons | AMA/AOA physician master files |
Analysis Plan
Specific Aim 1: Investigate the relationship between Hispanic neighborhood environments and cancer incidence and cancer mortality, using census tract level SEER (California, New Mexico) and vital registration (California, Texas) data.
Datasets: SEER, California and New Mexico coverage areas, 1992-1996, 2002-2003; Vital registration mortality, Texas and California, 1988-1992; 1998-2002
Purposes:
- To confirm our preliminary findings in the Hispanic EPESE and in the SEER data for New Mexico that there is an inverse relationship between census tract percent Hispanic (TpH) and cancer prevalence and incidence.
- To investigate variation across subpopulations defined by age group, gender, and immigrant status in the relationship between TpH and cancer incidence and mortality.
- To identify interactions between TpH and contextual and individual covariates of cancer incidence and mortality.
Specific Aim 2: Identify multi-level correlates (contextual and individual) of cancer mortality and cancer prevalence, using the National Health Interview Survey-National Death Index file,
Datasets: NHIS-NDI 1985-1994
Purposes:
- To investigate whether the relationship between TpH and cancer prevalence, cancer mortality and all-cause mortality found for older Mexican Americans is also found in younger subjects and in other Hispanic populations.
- To identify interactions between TpH and contextual and individual covariates of cancer prevalence, cancer mortality and all-cause mortality.
Specific Aim 3: To investigate the relationship between Hispanic neighborhood environments and cancer stage, treatment and survival, using SEER data from New Mexico and California.
Datasets: SEER California and New Mexico coverage areas, 1992-96; 2002-2003:
- 1992-96, these areas include the state of New Mexico and the metropolitan areas of Los Angeles, San Francisco/Oakland, San Jose/Monterrey.
- 2002-2003, these areas include the states of New Mexico and California.
Purposes:
- To explore the effects of neighborhood characteristics (ethnic concentration and poverty) on survival.
- To explore the contribution of delays in diagnosis and inadequate treatment on changes in survival associated with neighborhood characteristics.
- To explore the effect of the availability of health care resources on the association between neighborhood variables and stage at diagnosis and receipt of definitive treatment.
