SEARCH

SEARCH BY CITATION

Keywords:

  • behavior;
  • carcinoma;
  • epidemiology;
  • gynecology;
  • public health

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND

Cervical carcinoma is the fifth most common cancer among African American women in the U.S. Although the Papanicolaou (Pap) smear is an efficacious screening tool in the early detection of the disease, disparities are known to persist in the utilization of this procedure across socioeconomic groups.

METHODS

Data regarding cervical carcinoma screening and covariates were obtained from the 59,090 Black Women's Health Study participants across the U.S. via a mailed questionnaire in 1995. Logistic regression and multilevel techniques were used to assess the independent effects of the covariates on nonrecent cervical carcinoma screening.

RESULTS

In all, 8.3% of the 40,009 women in the present analysis had not undergone a Pap smear examination within the previous 2 years (nonrecent screening). Lower educational attainment, older age, obesity, smoking, and neighborhood poverty were found to be independently related to increased risk of nonrecent screening. The adjusted odds ratio for nonrecent screening was 1.2 (95% confidence interval [95% CI], 1.1–1.4) for women residing in neighborhoods with 20% or more poverty compared with those in neighborhoods with less than 5% poverty. State of residence was also associated with nonrecent cervical carcinoma screening.

CONCLUSION

These results suggest that among black women, residence in high-poverty (20%) neighborhoods is associated with an increased risk of nonrecent cervical carcinoma screening, independent of individual level risk factors. Cancer 2006. © 2005 American Cancer Society.

Cervical carcinoma is the fifth most common cancer among African American women in the U.S.1 In 2004, it is expected that more than 10,000 women will be diagnosed and nearly 3900 will die from cervical carcinoma.2 The Papanicolaou (Pap) smear is an efficacious screening tool that is able to detect premalignant cervical changes. The recommended screening interval for cervical carcinoma screening, according to the U.S. Preventive Services Task Force Guide to Clinical Preventive Services, is every 1–3 years.3 However, disparities are known to persist in participation in regular cervical carcinoma screening. Younger, nonobese, married, and more educated women with health insurance coverage are most likely to be up to date with their Pap screening.4, 5 Among black women, having a regular source of medical care, access to health insurance, younger age, and good health status have been found to be predictors of having up-to-date cervical carcinoma screenings.6

In addition to individual level factors, neighborhood level disparities also have been described with respect to patterns of cervical carcinoma screening. Results from an ecologic study suggested that cervical carcinoma screening rates were lower in low-income or poverty areas compared with wealthier areas.7 A recent report by the National Cancer Institute found that black women in higher poverty census tracts had a 30% higher incidence of cervical carcinoma compared with black women in lower poverty census tracts. The study also found that women in high-poverty counties had a 71% higher cervical carcinoma mortality rate than women in low-poverty counties.8

To our knowledge, there has been no multilevel study performed to date that examined the separate and independent contributions of individual and neighborhood socioeconomic factors on cervical carcinoma screening behaviors. In the current study we addressed this question in a cohort of black women. In addition to the contribution of neighborhood socioeconomic factors, we were also interested in the potential variations in adherence to cervical carcinoma screening guidelines within different U.S. states. The prevalence of some health behaviors, such as smoking, varies according to the state in which one lives.9 Therefore, we examined individual-, neighborhood-, and state-level predictors of cervical carcinoma screening.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Participants

The Black Women's Health Study (BWHS) was established to assess risk factors for breast carcinoma and other illnesses among American black women. A total of 59,090 women enrolled in the study in 1995 and have been followed subsequently. Women were recruited by mailing questionnaires to 445,000 black women who were subscribers to Essence magazine, members of selected black professional organizations, and respondent's friends and relatives.10, 11 The cohort was comprised of black women from across the U.S. who were ages 21–69 (median, 38 yrs) at the time of study entry.

Of the 59,090 women in the cohort, only those with complete data who had not undergone a hysterectomy or had a history of cervical carcinoma before 1995 were included in the current analysis. Data from a total of 40,009 participants were used. A total of 3595 participants had missing information for cervical carcinoma screening, 2733 had missing covariate values, 9017 had undergone cervical carcinoma or a hysterectomy, 471 were age 65 years or older, and 3265 addresses could not be geocoded either because the address was nonresidential (e.g., a business, post office box, or institution) or because it was not possible to convert the 2000 census tract to a 1990 census tract.

Exposure Measures

We used occupation (professional vs. nonprofessional) and educational attainment to represent individual level socioeconomic status (SES). These variables, along with income and wealth, are commonly used to capture individual SES.12 Data regarding income or wealth were not collected on the 1995 BWHS survey. We also included age group and marital status as individual-level covariates in our regression models.

After conventions established in previous studies of neighborhoods and health, we used census tracts as proxies for neighborhoods.13 Census tracts are administrative divisions of communities into reasonably homogeneous groups of approximately 3000 people. Study subjects were linked to their census tract by geocoding their addresses using 2000 Census data. Because Pap smear use was ascertained in 1995, we linked subjects to their 1990 census tracts using the Census Tract Relationship File; therefore, 2000 Census tracts were converted to 1990 Census tracts. We then merged 1990 Census tract data regarding the percentage of people living in poverty with our 1990 geocoded data. We compared the prevalence of recent Pap smear among the included and excluded women according to several key factors: education level, occupation, marital status, and state of residence. We found the prevalence of Pap smear to be similar among the included and excluded participants according to these factors.

Information concerning known risk factors for cervical carcinoma and possible confounders was obtained from the self-administered 1995 baseline questionnaire. Reproductive and other factors that might increase the level of surveillance were included in the current analysis. These factors included number of pregnancies, smoking status, body mass index (BMI) (weight/height2), and family history of breast carcinoma. Health insurance status, endometriosis, and uterine fibroids were examined in a subanalysis evaluating access to care.

Percent living in poverty at the census tract level was categorized into the following categories to enable comparison with other studies14: < 5% poverty, 5–9.9% poverty, 10–19.9% poverty, and ≥ 20% poverty. At the state level, we examined percent households living in poverty.

Outcome Measure

The 1995 BWHS questionnaire asked “When was your last Pap smear?” Participants chose from the following responses: “Never had one,” “Less than 1 year ago,” “1–2 years ago,” “3–4 years ago,” and “5 or more years ago.” The 1989 Guide to Clinical Preventive Services, which was in effect at the time that data on Pap use were collected at baseline in the BWHS, recommended cervical carcinoma screening should be repeated every 1–3 years. We defined recent cervical carcinoma screening as having undergone a Pap smear within the past 2 years. The Guide also suggested that cervical carcinoma screening may be discontinued at age 65 years if previous screenings have been normal. Therefore, women older than 65 years were excluded from the analysis.

Analysis

We anticipate cervical carcinoma screening to be clustered within the spatial contexts of census-tracts and states. This spatial clustering in the outcome was modeled by explicitly partitioning the different sources of variation. Multilevel statistical techniques provide a technically robust framework with which to analyze the correlated nature of the outcome variable and are pertinent when predictor variables are measured simultaneously at different levels.15 The principles underlying multilevel modeling procedures are now fairly well known.16–18 The fixed and random parameter estimates (along with their standard errors) for the three-level logistic regression model with a structure of 40,009 individuals at Level 1, nested within 12,696 census tracts at Level 2, nested within 44 states at Level 3 were calibrated using predictive/penalized quasi-likelihood (PQL) procedures with second-order Taylor series expansion,19 as implemented within the MLwiN program.20

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Individual-Level Analysis

The overall prevalence of nonreceipt of a Pap smear within the previous 2 years was 8.1%. The prevalence of nonrecent screening increased monotonically after age 40 years (Table 1). The odds ratio (OR) for women ages 60–64 years compared with those ages 20–29 years was 1.9 (95% confidence interval [95% CI], 1.5–2.3). Lower level of educational attainment was strongly associated with nonrecent screening: The OR was 2.6 (95% CI, 2.1–3.2) for women without a high school education compared with women with some graduate school education. Married women had a lower prevalence of nonrecent screening than nonmarried women. Women with body mass index (BMI) < 20 or ≥ 30 had a greater prevalence of nonrecent screening than women with intermediate BMI values. Nulliparous women and women with three or more children were less likely to have had recent screening than women with one or two children. Smokers had a greater prevalence of nonrecent screening than nonsmokers. There were no differences in nonrecent screening noted according to occupation and family history of breast carcinoma. In a subanalysis (data not shown), insurance status, diagnosis of fibroids, and endometriosis were not found to be associated with nonrecent cervical carcinoma screening.

Table 1. Prevalence of No Cervical Carcinoma Screening within the Previous 2 Years and Individual Characteristics (Black Women's Health Study, 1995) (n = 40,009)
 No.Age-adjustedAge-adjusted OR95% CI
Prevalence (%)
  1. OR: odds ratio; 95% CI: 95% confidence interval; BMI: body mass index.

Overall40,0098.1  
Education    
 < High school74414.62.62.1–3.2
 High school graduate531610.21.71.5–1.9
 Some college14,7918.61.41.3–1.6
 College graduate10,0887.41.21.1–1.3
 Some graduate school90706.41 
Marital status    
 Married13,9505.31 
 Living as married158881.41.1–1.7
 Separated18908.51.71.4–2.0
 Divorced59146.61.31.1–1.4
 Widowed9337.81.71.4–2.1
 Single15,73410.52.22.0–2.4
Age    
 20–29 yrs10,8267.31 
 30–39 yrs15,3526.80.90.9–1.0
 40–49 yrs99099.71.41.2–1.5
 50–59 yrs312711.21.61.4–1.8
 60–64 yrs79512.71.91.5–2.3
Occupation    
 Professional17,5377.30.80.8–0.9
 Nonprofessional22,4728.71 
BMI    
 < 2025309.71.41.2–1.7
 20–24.914,1806.51 
 25–29.911,9706.910.9–1.1
 ≥ 3011,329111.71.6–1.9
Parity    
 015,9148.91 
 189155.90.60.5–0.7
 284816.20.70.6–0.8
 ≥ 36699910.9–1.1
Current smoking    
 Yes61877.31.41.2–1.5
 No33,8228.11 
Family history of breast carcinoma    
 Yes7189.81 
 No39,2917.710.8–1.3

In a single-level multivariate model, age, education, marital status, smoking, parity, and BMI all were found to be significant predictors of recent cervical carcinoma screening after adjusting for the other covariates (Table 2).

Table 2. Multilevel Logistic Regression Models of No Pap Smear within the Previous 2 Years
Fixed effectsIndividual covariatesIndividual and census tract covariatesIndividual, census tract, and state covariates
OR95% CIOR95% CIOR95% CI
  1. Pap: Papanicolaou; OR: odds ratio; 95% CI: 95% confidence interval; BMI: body mass index.

Education      
 < High school1.81.4––2.31.81.4––2.31.81.4–2.3
 High school graduate1.51.3–1.81.51.3–1.81.51.3–1.8
 Some college1.31.2–1.51.31.2–1.51.31.2–1.5
 College graduate1.11.0–1.31.11.0–1.31.11.0–1.3
 Some graduate school1.0 1.0 1.0 
Marital status      
 Married1.0 1.0 1.0 
 Living as married1.21.0–1.51.21.0–1.51.21.0–1.5
 Separated1.41.2–1.71.41.2–1.71.41.2–1.7
 Divorced1.21.1–1.41.21.1–1.41.21.1–1.4
 Widowed1.41.1–1.81.41.1–1.81.41.1–1.8
 Single1.91.7–2.11.91.7–2.11.91.7–2.1
Age      
 20–29 yrs1.0 1.0 1.0 
 30–39 yrs1.11.0–1.31.11.0–1.31.11.0–1.3
 40–49 yrs1.91.7–2.11.91.7–2.11.91.7–2.1
 50–59 yrs2.11.8–2.52.11.8–2.52.11.8–2.5
 60–64 yrs2.41.8–3.02.41.8–3.02.41.8–3.0
Occupation1.00.9–1.11.00.9–1.11.00.9–1.1
Current smoking1.21.1–1.31.21.1–1.31.21.1–1.3
Parity      
 01.0 1.0 1.0 
 10.60.6–0.70.60.6–0.70.60.6–0.7
 20.80.7–0.90.80.7–0.90.80.7–0.9
 ≥ 31.00.9–1.11.00.9–1.11.00.9–1.1
BMI      
 < 201.31.1–1.61.31.1–1.61.31.1–1.6
 20–24.91.0 1.0 1.0 
 25–29.91.00.9–1.11.00.9–1.11.00.9–1.1
 ≥ 301.61.4–1.71.61.4–1.71.61.4–1.7
% Census tract      
Poverty      
 < 5%  1.0 1.0 
 5–9.9%  1.10.9–1.21.10.9–1.2
 10–19.9%  1.11.0–1.31.11.0–1.3
 ≥ 20 %  1.21.1–1.41.21.1–1.4

Multilevel Analysis

After taking into account individual-level variables the OR for nonrecent screening was 1.2 (95% CI, 1.1–1.4) for women living in census tracts with ≥ 20% poverty compared with those living in census tracts with < 5% poverty (Table 2). Because there was substantial variation in cervical carcinoma screening noted according to level of education, we also explored the possibility that the individual and neighborhood effects may be different among those with higher education. However, there was no change of estimates (P-interaction > 0.05) when cross-level interactions between individual education and census tract poverty were included in the model (data not shown). In addition, the association of census tract poverty on nonrecent screening persisted when examined within the group of women with at least a college education (< 5% poverty vs. > 20% poverty, OR of 1.3; 95% CI, 1.1–17).

In the model with individual covariates only, the variance between state of residence was found to be significantly associated with the prevalence of recent cervical carcinoma screening (σ2ν0 of 0.041; standard error [SE] of 0.017) (data not shown). The variance of cervical carcinoma screening prevalence between states was not explained by the percent living in poverty (σ2ν0 of 0.031; SE of 0.015) (data not shown). These results suggest that the rate of cervical carcinoma screening is significantly different among states, but that this difference is not explained by the percentage of persons living in poverty.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

To our knowledge, the current study is the first multilevel examination of cervical carcinoma screening behaviors in a national population of black women. The study adds to the literature by demonstrating that census tract and state factors influence cervical carcinoma screening behaviors above and beyond individual factors. Multilevel analysis demonstrated that after adjusting for individual characteristics, the percent of people living in poverty at the census tract level was significantly correlated with recent cervical carcinoma screening prevalence among black women. Two previous studies that assessed the correlation between neighborhood and Pap smear behaviors found an association between living in a poverty area and not being screened for cervical carcinoma.7, 8 Neither of these studies controlled for individual level factors.

The current study findings also reveal an association between the state of residence and cervical carcinoma screening. This association was not explained by the percentage of people in the state living in poverty; therefore, it is not clear what is causing the differences between states. It possible that there is differential access to health care between states due to business or policy decisions; however, we were unable to examine these factors in the current study.

Our findings at the individual level confirm the observations of previous studies that lower educational attainment, older age, obesity, and current smoking are associated with nonreceipt of a Pap smear examination within the previous 2 years. With respect to obesity, previous studies have reported that overweight and obese women are less likely to undergo cervical carcinoma screening.21, 22 One study found that physicians sometimes do not offer Pap smears to obese women and that obese women are sometimes more reluctant to have them.21

We lacked information regarding personal or household income in 1995. This may have resulted in residual confounding (i.e., assigning more influence to the neighborhood level because of incomplete control of individual factors). However, the presence of an association of census tract poverty levels with recent screening among women who had completed college or a higher level of education argues against strong confounding.

The accuracy of self-reported Pap smear in epidemiologic studies has been reported to be low. Studies comparing self-report to medical records have reported that positive predicted values range from 33–75% and that negative predictive values range from 85–94%.23, 24 Studies in low-income or minority populations have reported kappa values for the agreement between self-report and medical records ranging from 0.15–0.34.25, 26 However, because women occasionally seek reproductive care from several sources, medical records may not be a perfect ‘gold standard’ for cervical carcinoma screening.

The BWHS participants are not a random sample of U.S. black women. The women in this study were recruited largely from among women who subscribed to Essence magazine. They are more educated than the general population of black women and the 15% of black women nationally who did not graduate from high school27 are underrepresented in the current study population. It is likely that Pap smear self-report in the BWHS is more accurate than reports among low-income women.25, 26

The results of the current study suggest that individual-, neighborhood-, and state-level characteristics influence cervical carcinoma screening behaviors. Based on these results, we conclude that community outreach programs should focus on high-poverty neighborhoods to decrease the proportion of black women who are not adherent to cervical carcinoma screening recommendations nationally from 17% to the Healthy People 2010 goal of 10%.28

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES
  • 1
    American Cancer Society. Cancer Facts and Figures for African Americans 2003–2004. Atlanta: American Cancer Society, 2003.
  • 2
    American Cancer Society. Cancer Facts and Figures 2004. Atlanta: American Cancer Society, 2004.
  • 3
    U.S. Preventive Services Task Force. Guide to clinical preventive services. Washington, DC: U.S. Preventive Service Task Force, 1989.
  • 4
    Calle E, Flanders D, Thun M, Martin L. Demographic predictors of mammography and Pap smear screening in US women. Am J Public Health. 1993; 83: 5360.
  • 5
    Simoes E, Newschaffer C, Hagdrup N, et al. Predictors of compliance with recommended cervical cancer screening schedule: a population-based study. J Community Health. 1999; 24: 115129.
  • 6
    Mandelblatt J, Gold K, O'Malley A, et al. Breast and cervical cancer screening among multiethnic women: role of age, health, and source of care. Prev Med. 1999; 28: 418425.
  • 7
    Wells B, Horm J. Targeting the underserved for breast and cervical cancer screening: the utility of ecological analysis using the National Health Interview Survey. Am J Public Health. 1998; 88: 14841489.
  • 8
    Singh G, Miller B, Hankey B, Edwards B. Area socioeconomic variations in U.S. cancer incidence, mortality, stage, treatment, and survival, 1975–1999. Bethesda: National Cancer Institute, 2003.
  • 9
    State-specific adult smoking prevalence, smokeless tobacco prevalence and state tax-paid per capita sales of cigarettes, — United States, 1997. MMWR. 1998; 47: 922926.
  • 10
    Rosenberg L, Palmer J, Rao R, Adams-Campbell L. Risk factors for coronary heart disease in African American women. Am J Epidemiol. 1999; 150: 904909.
  • 11
    Russell C, Palmer J, Adams-Campbell L, Rosenberg L. Follow-up of a large cohort of black women. Am J Epidemiol. 2001; 154: 845853.
  • 12
    Duncan GJ, Daly MC, McDonough P, Williams DR. Optimal indicators of socioeconomic status for health research. Am J Public Health. 2002; 92: 11511157.
  • 13
    Krieger N, Zierler S, Hogan J, et al. Geocoding and measurement of neighborhood socioeconomic position: a U.S. perspective. In: KawachiI, BerkmanL, editors. Neighborhoods and health. London: Oxford University Press, 2003: 147178.
  • 14
    Krieger N, Chen J, Waterman P, Soobader M, Subramanian S, Carson R. Geocoding and monitoring US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter?: The Public Health Disparities Geocoding Project. Am J Epidemiol. 2002; 156: 471482.
  • 15
    Goldstein H. Multilevel statistical models, 3rd ed. London: Arnold, 2003.
  • 16
    Subramanian S. The relevance of multilevel statistical methods for identifying causal neighborhood effects. Soc Sci Med. 2004; 58: 19611967.
  • 17
    Subramanian S, Jones K, Duncan C. Multilevel methods for public health research. In: KawachiI, BerkmanL, editors. Neighborhoods and health. New York: Oxford University Press, 2003: 65111.
  • 18
    Subramanian S, Nandy S, Kelly M, Gordon D, Smith G. Patterns and distribution of tobacco consumption in India: cross-sectional multilevel evidence from the 1998-9 national family health survey. BMJ. 2004; 328: 8018006.
  • 19
    Goldstein H, Rasbash J. Improved approximations for multilevel models with binary responses. J R Stat Soc Ser A. 1996; 159: 505513.
  • 20
    Rasbash J, Browne W, Goldstein H, et al. A user's guide to MLwiN. London, UK: Center for Multilevel Modelling, Institute of Education, University of London, 2002.
  • 21
    Wee C, McCarthy E, Davis R, Phillips R. Screening for cervical and breast cancer: is obesity an unrecognized barrier to preventive care? Ann Intern Med. 2000; 132: 697704.
  • 22
    Adams C, Smith N, Wilbur D, Grady K. The relationship of obesity to the frequency of pelvic examinations: do physicians and patient attitudes make a difference? Women Health. 1993; 20: 4557.
  • 23
    McGovern P, Lurie N, Margolis K, Slater J. Accuracy of self-report of mammography and Pap smear in a low-income urban population. Am J Prev Med. 1998; 14: 201207.
  • 24
    Sawyer J, Earp J, Fletcher R, Daye F, Wynn T. Accuracy of women's self-report of their last Pap smear. Am J Public Health. 1989; 79: 10361037.
  • 25
    McKenna M, Speers M. Agreement between patient self-reports and medical records for pap smear histories. Am J Prev Med. 1992; 8: 287291.
  • 26
    Paskett E, Tatum C, Mack D, Hoen H, Case L, Velez R. Validation of self-reported breast and cervical cancer screening tests among low-income minority women. Cancer Epidemiol Biomarkers Prev. 1996; 5: 721726.
  • 27
    U.S. Department of Commerce. Educational attainment in the United States. Washington, DC: Department of Commerce, 1998.
  • 28
    U.S. Department of Health and Human Services. Healthy people 2010: health objectives for improving health. Washington, DC: U.S. Government Printing Office, 2000.