We used data from the 2000 NHIS, Cancer Control Module (CCM) (22), for our analysis. The NHIS contains health information obtained from noninstitutionalized, civilian U.S. households with a maximum of one adult and one child. The CCM is part of the adult sample level of the NHIS. In 2003, the American Cancer Society modified its cervical cancer screening recommendations (23). At the time we developed this manuscript, the 2005 NHIS data were not yet available. However, since the 2003 cervical screening guidelines were not fully implemented at the time of the 2005 NHIS data collection, we do not anticipate extensive differences in the results. Of the 38,633 households sampled in the entire 2000 NHIS, 72.1% (32,374) of adults aged 18 or older responded to the CCM questionnaire. Our sample consists of women aged 55 or older without a history of hysterectomy (N = 3301).

The dependent variable was dichotomous: individuals either had or had not had a Pap test in the past 36 months. American Cancer Society recommendations for cancer screening at the time of the interview included at least one Pap test every 3 years for women 18 years of age or older with a cervix (24,25). The analysis included the following potential predictors: 1) age, 2) education level, 3) income, 4) marital status, 5) number of mammograms in the past 6 years, 6) talking with or visiting a health professional in the past 12 months, 7) talking with or visiting an OB/GYN in past 12 months, 8) having a usual source of preventive or routine care, 9) residence in a metropolitan statistical area (MSA) or non-MSA area, and 10) race/ethnicity.

The NHIS used a complex stratified multistage probability design that was age-adjusted and that oversampled African Americans and Hispanics using the 2000 census population. Data were analyzed using descriptive analysis and logistic regression in SPSS (SPSS Inc, Chicago, Illinois) and SAS 9.1 (SAS Institute Inc, Cary, North Carolina) to adjust for the complex sampling and using person weights to adjust for census-level data on sex, age, and race/ethnicity. Using data from women who provided complete information on all variables studied, we assessed the relationship between receiving a Pap test within the past 3 years and the women's demographics, whether they had received other cancer screening (e.g., mammograms), and whether they had visited a health professional or OB/GYN within the past year.

The central Appalachia region, comprising mainly West Virginia, Kentucky, Ohio, and Tennessee, shares a cultural, economic, and resource base and has disproportionately high rates of cervical cancer (19). To better understand this cervical cancer burden, we studied two central Appalachian counties with comparable population sizes, economic structures, and health care capacity. The Kentucky county had 42,421 people, of which 30% lived below the federal poverty guideline and 39% had less than a high school education. The West Virginia county had 37,710 people, of which 24% lived below the federal poverty guideline and 63% had less than a high school education.

To assess what rarely or never-screened middle-aged or older central Appalachian women consider determinants of Pap test use, two local, trained middle-aged women conducted in-depth interviews consisting of structured and semistructured questions. These women were community and social workers who have lived all of their lives in the Appalachian communities in which they conducted the interviews and had extensive experience interviewing Appalachian residents. To familiarize the interviewers with research, we had them participate in extensive training sessions in which they were taught about cancer prevention, cervical cancer, and Pap tests.

The interviewers recruited the women for the study through snowball sampling (26) from a community college cafeteria, a low-income housing project, and a senior citizens center. Since identifying rarely or never-screened women is a sensitive and difficult task, we relied on the trustworthiness and reputations of the local interviewers who would discreetly ask women from these locations several screening questions. Screening included questions about age and residence (if not known) and questions about disease prevention, including when they last had a doctor's checkup and when they last had a mammogram and Pap test. Inclusion criteria were rarely (3 or more years since last Pap test) or never receiving screening for cervical cancer, residing in central Appalachia, and being middle aged or older (55 or older). Of women who were eligible and who were asked by the interviewer to participate, approximately 20% refused to do so.

Consistent with standard qualitative sampling, including theoretical saturation, participants were recruited and interviewed and data were analyzed simultaneously (27). On completion of 23 interviews, no new data emerged, and we considered our data collection complete (interviews with two additional women already had been scheduled and served as further verification).

For women who were interested in participating in the survey, an appointment was arranged for an interview, generally at the participant's home. The interviewers explained the project, administered informed consent documents, and conducted semistructured interviews. Interview guides were shaped by ecological theories of behavior that focus on the individual, social networks, the provider and health care system, and the community environment (28). All procedures were approved by the University of Kentucky's Institutional Review Board. At the end of the session, participants were provided with a $25 honorarium and information about accessing screening services, including the National Breast and Cervical Cancer Early Detection Program, with which most were unfamiliar.

The tape-recorded sessions were transcribed for subsequent analysis. Two trained researchers independently did line-by-line or axial coding, in which a label or code is affixed to chunks of text (29). To ensure rigorous, systematic, and comparable analysis, we compiled the codes and themes into a codebook. To enhance verification, two researchers coded identical text portions, establishing an intercoder reliability rating of .85 or greater (30). Researchers also met to discuss concerns or discrepancies.

We screened NHIS data for appropriateness of multivariate analysis. Analyses detected no multicollinearity among predictor variables, including variables with highly skewed distribution (having a usual source of health care and having seen a health professional in the past year). For the final analyses, we excluded 1078 cases because of missing data. The only apparent differences between the initial sample (N = 3301) and the final sample (N = 2223) were that the original sample had a lower percentage of women reporting receipt of a Pap test in the past 3 years (78.6% compared with 84.7% in the final sample) and a higher percentage of women having seen an OB/GYN in the past year (72.5% compared with 66.6% in the final sample). Descriptive statistics for the final sample are presented in Table 1.

The variables used in a logistic regression analysis to predict whether someone received a Pap test within the past 3 years were being Hispanic, non-Hispanic black, or non-Hispanic other; marital status; income; residing in an MSA; categorical age; education level; number of mammograms in the past 6 years; having seen a health professional in the past year; having seen an OB/GYN in the past year; and having a usual source of health care. As shown in Table 2, being married, living in a metropolitan area, having seen an OB/GYN in the past year, having a usual source of health care, being younger, and having a higher number of mammograms in the past 6 years were associated with a higher likelihood of receiving a Pap test within the past 3 years. Race/ethnicity, income, and education level did not significantly contribute to the final model.

Sample

The 25 participants, two-thirds of whom were white and one-third of whom were black, ranged in age from 55 to 79, with a median age of 62. Educational attainment was limited, with 42% receiving 12 years of education or fewer. Annual household incomes were similarly modest; 22% had incomes of less than $5000, 35% had incomes of $5000 to $14,999, 25% had incomes of $15,000 to $24,999, and 18% had incomes of $25,000 or higher. Two women had private health insurance, five lacked health insurance, and 18 (72%) mentioned Medicare, Medicaid, or dual eligibility as their primary source of health insurance. The median number of years since their last Pap test was 8, with a range of 3 years to never for having been screened.

Determinants of cervical cancer screening

In-depth interviews revealed extensive convergence with the NHIS data. Participants described how older age, living in a geographically isolated place, and limited social ties converge to thwart Pap testing. A younger participant noted:

Discussing growing older and more isolated, another woman echoed this perspective:

Being older conferred numerous challenges to Pap test receipt, including competing attention from more pressing health conditions and having more limited incomes. Being unmarried, as was the case for approximately one-third of this sample, added to these challenges by limiting health vigilance or access (i.e., no one pressuring you to take care of yourself or transporting you to an appointment) and by perceived insufficient need to undergo a gynecological check-up because of limited sexual activity.

Living in a geographically remote environment further compounded the challenges of being screened, particularly for a health activity given low priority. Women discussed numerous tangible community challenges (i.e., bad roads, inadequate transportation, lack of health care professionals). A 65-year-old woman who did not drive discussed the lack of public transportation services:

Participants also described health care systems barriers to getting a Pap test. Negative opinions of preventive health services seem to stem from several sources: insufficient and inconsistent availability of health care professionals, inadequate health insurance, lack of confidentiality with medical care, and negative attitudes toward formal health care.

The scarcity of health care professionals in rural and underserved communities leads to a constellation of associated problems, as noted among our participants and in the literature (31). Participants described too few doctors (leading to inadequate hours for clinics, a dire shortage of specialists, few female physicians, and physicians who are viewed as "good doctors" being overworked), too little privacy (one woman reported that the test results of a very sensitive problem were known at her worksite before she even returned to her job that afternoon), and constant turnover of health care professionals, especially among international medical graduates and residents.

Many women seldom make appointments with physicians unless they are in great discomfort or believe their health to be in danger, suggesting that the doctor is a last resort when experiencing pain or disability. As described by the participant below, who had not "been to a doctor's office except to get something for bronchitis trouble I had, and that was about 5 years ago," this lack of care tends to become habitual:

Participants voiced concern about how health care providers chastise them for not seeking regular preventive care, which only made them reluctant to seek care, both to prevent health problems and to treat emerging problems. One woman explained how a nurse yelled at her for not having regular blood pressure checks:

Another woman noted that she seldom visited a health care provider,

Others described the possible humiliation they might face during such an examination, particularly being smokers or overweight, behaviors they acknowledged as problematic and likely to meet with disapproval from health care professionals. One woman who last had a Pap test 14 years ago noted,

Participants' narratives defy a simple conclusion that the women lack an appreciation for prevention or lack transportation. Instead, tangible barriers to prevention combined with a historical and community context that may relegate prevention to low-priority status thwart Pap testing.

Our quantitative and qualitative findings suggest that being unmarried or living alone, living in an isolated place, and being older tend to decrease Pap test use. Married women were 43% more likely to be current on their Pap tests compared with all other women, similar to findings by Hewitt et al (7). Narratives from rarely or never-screened women help explicate this finding: unmarried women may not have anyone to encourage them to seek health care and might also experience greater logistical barriers, including a lack of transportation (especially salient for women who do not drive) and a lack of health insurance. Some women also discussed their perceived lack of need for Pap tests due to their current lack of sexual activity, an incorrect perception that may heighten their risk of cervical cancer.

Women from MSAs were 52% more likely to have received a Pap test in the past 3 years than were women from rural areas, a finding that is both refuted (6) and corroborated (13) by previous research. Our in-depth interviews point to several explanatory factors, including structural limitations (e.g., insufficient health care services, lack of health insurance).

With each additional 10-year age band among women aged 55 or older, the likelihood of having regular screenings decreased by 35%, a finding consistent with previous research (7,8,17). Interviews revealed that advanced age places a greater burden on women to manage other, seemingly more important health concerns ("aches and pains" along with serious chronic diseases), in addition to factors such as social and geographical isolation and greater financial constraints. These results are not consistent with a previous study on colorectal cancer screening in Appalachia in which older age was associated with a higher likelihood of having a recent screening (33). However, colorectal cancer and cervical cancer are very different conditions with different ages of screening initiation. In addition, colonoscopy is now covered by Medicare, removing a major financial barrier for older adults.

Both the qualitative and quantitative portions of this study indicate that a key determinant of receiving a Pap test is undergoing other types of preventive screening. Although intuitive, this finding has only received a small amount of attention in the literature (13). Lyttle and Stadelman (34) found similar results, with women who received regular mammograms being more likely to receive regular colorectal cancer screenings. The qualitative interviews in our study suggested that competing health, social, and economic concerns converged with histories of self-reliance, traditional health practices (e.g., complementary and alternative medicine), and unconventional health beliefs that fall outside of formal medical encounters. Participants, many of whom have multiple, demanding, and costly chronic conditions, discussed the need to prioritize their use of time and economic and social resources. Preventive health screenings may be relegated to a secondary concern, not out of ignorance, fatalism, or lack of motivation — assumptions frequently leveled at traditionally underserved populations — but because of self-care traditions and prioritization (31). Previous research confirms that those self-care practices established earlier in life tend to remain in people's health repertoire, particularly when they continue to operate in resource-scarce conditions (35). The women in our study, and most likely many women who do not receive Pap tests, have developed health strategies consistent with their culture and circumstances (poverty and shortages of health care professionals meant mother was the doctor) and with their health priorities (managing multiple chronic conditions, including pain and some disability).

Having contact with an OB/GYN was the strongest predictor of Pap use in this sample of women, increasing the likelihood of being current by nearly eight and a half times and corroborating others' results (7,11). As suggested by our qualitative participants, those individuals who "pop over to their doctors on a regular basis and say, 'Here I am — ain't I healthy or what?' " not only have a preventive health orientation but also have the means to get to providers' offices and to overcome a lack of health care professionals and an inadequate continuity of care. Participants do not mistrust or refute the legitimacy of the medical establishment; however, factors such as lower socioeconomic status and traditions of self-reliance may encourage women to give low priority to prevention (31,36).

Race/ethnicity, income, and education level did not significantly predict current Pap test screening. However, Hewitt and colleagues (7) suggest that individuals aged 25 or older with less than a high school education and with a lower income are less likely to receive regular screenings. Additionally, Swan and colleagues (6) found that higher income and higher levels of education are associated with receiving regular Pap tests among women 25 and older. These results suggest that distinct factors may play a larger role in screening uptake among middle-aged and older women.

One limitation of self-report surveys such as the NHIS is that many women overreport Pap test use (37). Another limitation is that many women in the sample did not respond to one of the variables in the study, which may influence the generalizability of the quantitative results. A limitation of our quantitative analyses was that two variables (having a usual source of health care and having seen a health professional in the past year) had little variability, perhaps explaining why seeing a health professional in the past year did not significantly predict Pap testing.

Finally, while the predictive and explanatory powers of this study were enhanced by complementary methods, the data came from separate data sets. We cannot necessarily explain the decisions of the women in the quantitative portion of this study through the narratives of the qualitative participants. However, qualitative insights help to explain patterns observable through the quantitative data, while the quantitative data allow us to examine associations among variables with confidence in their reliability and generalizability. We cannot definitively state that the factors and circumstances influential for the 25 women interviewed are generalizable to other locales in Appalachia or in the rural United States.

Our results suggest that a vicious cycle pertains to underscreening of vulnerable populations. Older, unmarried women living in rural areas who do not have a usual source of preventive or routine health care, who do not visit or speak with an OB/GYN each year, and who do not receive other preventive care such as mammograms are less likely than other women to receive regular Pap tests. Rarely or never-screened women explain this pattern in terms of logistical challenges (e.g., not having adequate transportation), perceptual barriers (e.g., worry that health care professionals will chastise them for being overweight), and lack of social support (e.g., a supportive person who is concerned about the woman's health). When they do overcome these barriers, women report feeling stigmatized in the medical encounter, which undermines their use of preventive services.

The best predictor of screening is the recommendation of a health care professional (38–40). In the absence of supportive family and friends, health professionals must provide encouragement in addition to consistent and clear recommendations to patients of all ages, with special attention to middle-aged or older women. Kind words and sensitivity to the environmental and personal constraints faced by many traditionally underserved women would most likely facilitate screening. In addition, expanded support should be provided for programs such as the National Breast and Cervical Cancer Early Detection Program, which assist lower-income and rarely or never-screened women in receiving needed cancer prevention services.