| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Patrick Fox, PhD, Institute for Health & Aging, University of California, San Francisco, Laurel Heights Campus, San Francisco, CA 94143-0646. E-mail: pf1965{at}itsa.ucsf.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Mammograms • Older women • Breast-cancer screening
In spite of a higher incidence of breast cancer, higher rates of diagnosis in the later stages of the disease, and increased likelihood of mortality in older women, mammography rescreening rates are lower for older women than for younger women (Fink, Shapiro, & Roester, 1972; Glanz et al., 1992; Scaf-Klomp, van Sonderen, Stewart, van Dijck, & van den Heuvel, 1995; Song & Fletcher, 1998). An impact of early detection through regular mammography use is the possibility of a more optimistic prognosis (McCarthy et al., 2000) and less invasive early-stage treatment choices (King, Rimer, Balshem, Ross, & Seay, 1993), in that obtaining regular mammography screening has been shown to reduce breast cancer mortality rates (American Cancer Society, 2001).
Mammography guidelines currently used in the United States do not specify an upper age limit for screening (American Academy of Family Physicians, 2001; American Cancer Society, 2000; American College of Obstetrics & Gynecology, 2000; U.S. Preventive Services Task Force, 2002). Findings regarding the benefit of screening among women aged 70 and older are unclear, largely because of the lack of randomized trials that include women over the age of 69 (Van Hoeyweghen, 2001).
Among studies examining the benefits of screening in older women, one report showed clear mortality reductions among women 50 to 69 years of age, but the report could not assess the effectiveness of screening among older women as a result of data limitations (Fletcher, Black, Harris, Rimer, & Shapiro, 1993). Other studies have found screening mammography to be associated with a 43% decrease in the risk of detecting metastatic breast cancer in women between the ages of 66 and 79 (Smith-Bindman, Kerlikowske, Gebretsadik, & Newman, 2000); a reduction of breast cancer related deaths among women up to 74 years of age (Humphrey et al., 2002); and mortality reductions among women 85 years of age and older (McPherson, Swenson, & Lee, 2002). Others have concluded that screening mammography is only minimally beneficial among women 69 years and older because of the small gain in life expectancy and competing mortality risks, and they recommend that this fact play an important role when older women make screening decisions (Kerlikowske, Salzmann, Phillips, Cauley, & Cummings, 1999). In response to these issues, some researchers have emphasized the heterogeneity among older women and have established protocols for screening decisions accordingly (Mandelblatt et al., 2000).
Individual Factors That Influence Mammography Rescreening
Mammography rescreening rates have been found to be significantly lower among women in lower socioeconomic status (SES) groups and among minority women (Calle, Flanders, Thun, & Martin, 1993; Caplan, Wells, & Haynes, 1992). Adherence to mammography screening among women between the ages of 51 and 75 has been shown to be associated with (a) having a higher income; (b) being Caucasian; (c) being between the ages of 51 to 64 and having had breast symptoms or a family history of breast cancer; (d) seeing a gynecologist as one's regular physician; (e) having had breast symptoms or a history of breast cancer; (f) having had a higher frequency of clinical breast exams; and (g) having had a recent doctor's visit (Zapka, Stoddard, Maul, & Costanza, 1991). Another study of women between the ages of 50 and 75 that examined factors associated with repeat versus one-time mammography use found that regular visits to a gynecologist, believing the lifetime risk of breast cancer is at least 10%, and feeling vulnerable to getting the disease were associated with reports of repeat mammography screening (Taylor, Taplin, Urban, White, & Peacock, 1995). A study that included low-income, multiethnic women found that although repeat screening was very low, women who had a higher number of visits to primary and specialty clinics were more likely to engage in repeat screening (Bastani et al., 1995). Other findings suggest that women who have a prior mammogram are more likely to have a subsequent mammogram, a result that was especially relevant to older women and African American women (Parker, Sabogal, & Gebretsadik, 1999).
Provider Factors That Influence Mammography Rescreening
Research suggests several reasons for the lack of breast cancer screening or rescreening among older women, with the absence of a physician recommendation or referral being important contributors (Costanza, Stoddard, Gaw, & Zapka, 1992; Grady, Lemkau, McVay, & Reisine, 1992; King, Resch, et al., 1993; King, Rimer, et al., 1993; Lane, Zapka, Breen, Messina, & Fotheringham, 2000; Rimer, 1993). Absence of a physician recommendation for mammograms for older women is related to the fact that, as women age, their visits to obstetrician–gynecologists decrease, and other physicians are less likely to make the necessary referrals (Sutton & Doner, 1992). Studies also have found that, although physicians report conducting or supporting breast-health procedures for women over the age of 65 years, actual practice indicates otherwise (Herman, Speroff, & Cebul, 1995; Roetzheim, Fox, & Leake, 1995). Older women also may have more competing health problems or functional limitations for primary care physicians to address (Caplan & Haynes, 1996), decreasing the likelihood that mammography recommendations will be made (Nosek & Gill, 1998).
Conflicting screening guidelines for older women may also reduce physician recommendations (Mandelblatt & Yabroff, 2000). The U.S. Preventive Services Task Force recommends screening mammography, with or without a clinical breast exam, every 1 to 2 years for women 40 years of age and older (U.S. Preventive Services Task Force, 2002). The American College of Obstetricians & Gynecologists recommends annual clinical breast exams and mammography every 1 to 2 years for women from 40 to 49 years of age, and every year thereafter (American College of Obstetricians & Gynecologists, 2000). The American Cancer Society recommends annual clinical breast exams and mammograms for women after the age of 40, with no upper age limit specified (American Cancer Society, 2000). The American Academy of Family Physicians recommends mammography and clinical breast exams every 1 to 2 years for women between the ages of 50 and 69 (American Academy of Family Physicians, 2001). The U.S. Preventive Services Task Force notes that although the effectiveness of mammography screening is greater for women between the ages of 50 and 69, the evidence of mammography-screening effectiveness is applicable to women 70 years of age and older if they are not imperiled by comorbid conditions (U.S. Preventive Services Task Force, 2002). If women see multiple clinicians who provide them with inconsistent breast cancer early-detection messages, this may have a significant negative impact on rescreening behaviors.
Health Care System Factors That Influence Mammography Rescreening
A multifaceted patient-education intervention that targeted patient barriers, physician reminders, and physician-focused audit with feedback was successful in improving biennial mammography use among women 65 to 74 years of age (Preston, Scinto, Grady, Schulz, & Petrillo 2000). Rimer and colleagues (1992) conducted a study of 412 women 65 years of age and older who had not had a mammogram in the previous year. Women in the control group received cost subsidy alone, and women in the experimental group received cost subsidy, mammography van, and tailored health education. Women in the experimental group were nearly four times more likely to obtain a mammogram in the subsequent 3 months following baseline than women in the control group. These results suggest that cost coverage of procedures alone will not address the full range of older women's barriers to regular mammography screening.
The addition of access and health-education interventions may improve rescreening rates. For example, telephone counseling alone or in conjunction with a personalized letter has been shown to improve utilization of mammography screening (King, Ross, Seay, Balshem, & Rimer, 1995), as did a two-step intervention that included a personal letter from a physician combined with an incentive and a peer-counseling call (Janz et al., 1997). Targeted mailings informing individuals of having insurance coverage for mammograms have also been shown to improve screening outcomes (Fox, Stein, Sockloskie, & Ory, 2001). However, a study examining the benefits of a lay-health-worker-education model with African American women aged 65 and older found there were no significant effects on rescreening (Zhu et al., 2000).
A combination of individual, provider, and system factors contribute to the underutilization of mammography by older women. In this study, we examine both patient and health care system factors that influence mammography rescreening behavior among a group of older multiethnic, low-income women.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Clinical Sites
Clinical sites that performed screening and diagnostic services and where program participants received their clinical breast exam (CBE) were identified by use of claims data. First, sites were randomly selected and called to confirm eligibility. Second, eligible sites were then recruited from the randomized lists into the study. Over three fourths (78.3%) of the eligible sites consented to participate, for a total of 108 clinical sites. Third, 3 sites were subsequently eliminated because there was an insufficient number of clients served, and 3 more refused to participate during the site visit data-collection interview. This resulted in a final sample of 102 clinical sites.
Patients
Patients were eligible for the larger study if they were 50 years of age or older with mammography results that were negative or benign at the time of their 1996 index mammogram. Our sampling goal was to obtain completed interviews for 10 women per clinic. A stratified random-sampling procedure was used to ensure representation among the ethnic groups served by the programs. Because of our familiarity with the difficulties associated with surveying low-income multiethnic women, we randomly selected approximately 35 survey participants from each of the 108 clinical sites to ensure that we obtained our target of 10 women per clinic. Our sample included women who belonged to four ethnic groups as a proxy for the languages that they were likely to speak: Latina, Chinese, Filipina, and all others (e.g., European American, African American, Native American, and Asian–Pacific Islanders other than Chinese or Filipina).
By using claims lists, we randomly sampled potential participants from each ethnic group and mailed them letters in four languages (Spanish, English, Tagalog, and Chinese). We oversampled from the Chinese and Tagalog strata to ensure sufficient numbers of patients for analysis, as there were fewer women from these ethnic groups screened by the programs. We mailed letters in the four languages to the patients' last known addresses, informing them of the study. Interviewers were instructed to call women until 10 patient interviews had been completed for each clinical site. Each telephone number was called a minimum of nine times at varying times of the day and on different days of the week.
Following refusals (12.1%), nonworking phone numbers, the inability to reach clients who could not be reached after an attempted 10 phone contacts (29.2%), and incomplete interviews (4.6%), 1,199 telephone interviews were completed. Six out of 108 sites dropped out of the study, leaving a final sample of 1,050 women. Of these patients, 446 were over the age of 60, and, of this group, a total of 391 women had complete data. It is from this subset of women that the data for this study were utilized. When comparing the total subsample of 446 women to the analytical sample of 391 women, we found that the only significant difference was a higher proportion of Cantonese or Mandarin-speaking patients in the former compared with the latter group (9.9% vs. 5.9%, respectively).
Measures and Data-Collection Procedures
Clinical Site Interviews
An in-person interview was conducted with one administrative representative at each site utilizing a questionnaire covering the following domains: (a) organizational characteristics such as size, location, organization type (public vs. private), staffing patterns, and language(s) spoken by staff; (b) characteristics of clientele, including race, age, health care reimbursement sources; (c) availability of parking, public transportation, and safety of the clinic's neighborhood; (d) methods used for breast-health education and methods of outreach; (e) use of tickler systems, databases, or other "in-reach" methods used to track patients for rescreening; (f) interviewer's assessment of the site's physical environment, including the ambience of the waiting room and the examining rooms; and (g) relationship among sites, the state agency, and local breast cancer partnerships.
Patient Interviews
A telephone survey instrument was developed to retrospectively assess the experiences of women who went through the breast cancer screening process at the programs' clinical sites. Women were considered to have been rescreened if they returned for another mammogram within 18 months of their program baseline examination. The interview instrument was developed specifically for this study, and many of the items had been used in a prior study of early cancer detection among Latinas (Perez-Stable, Otero-Sabogal, Sabogal, & Napoles-Springer, 1996). Interviews took a median of 30 min and included the following: (a) demographic information, such as age, gender, marital status, race, ethnicity, language spoken and read, country of origin, years of education, income, health insurance status, employment, health history and comorbid conditions, and years in the United States; (b) health practices and beliefs, such as religiosity, social support, access to a source of regular health care, reasons for beginning the screening process, breast cancer knowledge and beliefs, decisional balance, fatalism, reported breast self-exams, and perception of personal risk for breast cancer; (c) access to health care, that is, barriers related to insurance or transportation, or the need to use multiple providers to meet one's health care needs; (d) patient provider communication, that is, the use of interpreters and receipt of communication about screening results; and (e) reasons for being screened and barriers and promoters of rescreening. (These are questions theorized to be specifically indicative of satisfaction, such as, whether looking back, women would have preferred to have their exams somewhere else; whether they were inappropriately charged for services; whether there were any problems with need for prior authorization of services; and whether or not they would recommend the program to friends.)
The general survey response categories were yes–no (dichotomous), and scales were composed of 3 points (good, neutral, or bad). Pretesting indicated that these simplified response categories offered better construct validity for each answer and reduced the number of "do not know" responses in the sample population surveyed. After pretesting, the English-language survey was translated into Spanish, pretested again, and then backtranslated into English. This backtranslation served as the basis for the final English-language version of the client questionnaire, which was then translated into Tagalog, Cantonese, and Mandarin. Each translated version was backtranslated into English to ensure consistency across languages.
Data Analysis
We merged the clinical site and patient data sets for analysis, linking them through a master list of identifiers. As noted earlier, because of missing values for either response or explanatory variables, we deleted 55 observations, leaving an effective sample size of 391 for the final analysis. The variables we tested for inclusion in the final model were those hypothesized to be related to rescreening in the literature or were based on results of bivariate analyses where there was a moderate statistical relationship between either an individual or site-level variable and the dependent variable (defined as whether or not the woman had had a second rescreening mammogram within 18 months of her baseline mammogram). The client-level variables originally tested for inclusion in the models predicting the likelihood of rescreening were numerous.
Of note is that among the variables that did not prove significant in the final model were those reflecting ethnicity. The instrument used in this study included many variables reflecting ethnicity and language of origin, but the variable with the least amount of missing data was a question regarding the language in which the interview was conducted. From this list a four-level categorical variable was created based on the four languages in which the interview was conducted. The first category was interviewed in Chinese, the second category was interviewed in Spanish, the third was interviewed in Tagalog, and the fourth was interviewed in English. The last category (i.e., English) was the dropout category against which all others were compared. None of the three were significant predictors of rescreening in this sample.
Patient income level was another variable that was not associated with the likelihood of rescreening because there was little variability in income among clients in the sample in that all must be under 200% of the federal poverty level and ineligible for Medicaid in order to receive services under the BCCCP and BCEDP programs. There were also significant problems with colinearity when we included both years of education and measures of income in the same model. As education consistently retained significance across models and income did not, we deleted income from the final model.
As there was a limited dependent variable (rescreened vs. not rescreened), we first considered logistic regression procedures for the analysis. However, as the sample contained both subject-level variables (such as years of education) and site-level variables (such as site procedures for tracking patients), and as the participants were nested within the sites, we used a type of general linear modeling procedure (SAS PROC GENMOD) with a logit link function to analyze the data (the GEE procedure). This modeling procedure takes into account the effect of each woman's having repeated measures and being nested in sites. It is an extension of generalized linear models in that it allows the mean of a population to depend on a linear predictor through a nonlinear link function and allows, among other options, the response probability distribution to be for binary data. The robustness of this method is due to the fact the GEE produces a marginal model in which the resulting parameter estimates are population averaged (Koch, Gillings, & Stokes, 1980; Stokes, Davis, & Koch, 2000).
We modeled clinical site as a class variable in the analysis with the repeated-subject option to allow the correlations between individual women within one site to be removed, an option which is not available when we use the logistic procedure. We used goodness-of-fit tests (the ratio of the scaled deviance to the degrees of freedom) to develop the model that both converged and was the best fit to the data. Finally, we generated odds ratio estimates, along with their confidence intervals, for each variable retaining significance in the analysis.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
Clinic-level variables included the following: (a) The pap smear test is offered at the site for program patients (the pap test would be free if the site participated in the national breast and cervical cancer screening program, so this variable tended to separate these sites from sites that participated only in the statewide breast cancer screening program); (b) the site uses a hands-on demonstration of how to detect breast cancer with models; (c) the site offers a telephone information-and-advice line; (d) the site uses program brochures to explain program benefits; (e) the site uses posters to advertise the program; and (f) the site uses patient-tracking procedures to remind patients that they are due for a mammogram.
Among patient-level variables, both years of education and an endorsement of the statement that the reason the index mammogram was received was because it was important to get an annual mammogram were highly significant predictors of rescreening (b = 0.072, Pr > |Z| =.002 and b = 1.57, Pr > |Z| = <.0001, respectively). Odds ratios indicate that, for each year of education, sample women are 1.08 times more likely to be rescreened. For women who endorsed the statement about the importance of annual mammograms, the odds of being rescreened were almost five times greater than for women who had not endorsed this statement.
Among site-level variables, both offering a pap smear on site and using hands-on demonstrations with breast models were most significant in predicting rescreening (b = 1.422, Pr > |Z| = 0.029 and b = 0.487, Pr > |Z| = 0.029, respectively). In clinical sites where a free pap test was offered, women were more than four times more likely to return for rescreening. Two site-level variables were also significantly inversely predictive of rescreening: The site offers a telephone information-and-advice line (b = -0.555, Pr > |Z| = 0.036), and the site uses program brochures to explain program benefits (b = -1.149, Pr > |Z| = 0.003). Of the outreach–educational techniques used by the sites, program posters describing the importance of regular mammograms were most successful in increasing the likelihood that women returned for mammography rescreening (odds ratio, 1.764).
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
We found that both individual and health-care-system-delivery factors play an important role in low-income women's obtaining regular mammograms. Women aged 60 and older who are more highly educated and are already convinced that mammography is important are most likely to be rescreened. Our finding that more highly educated women are more likely to complete rescreening mammography is consistent with much of the previous literature (Baker, 1982; Fink et al., 1972; Hitchcock, Steckevicz, & Thompson, 1995; Sobel, Curtin, & Fell, 1991; Zapka et al., 1991). Within this sample of low-income multiethnic women, it appears that educational level, which may be linked to preventive-health beliefs, is a substantial factor affecting rescreening behavior—even more so than income or race or ethnicity.
Given the low annual household income of women in this sample (approximately three fourths have annual household incomes below $20,000), this study provided the opportunity to examine within-low-income-group factors among older women that are associated with their rescreening behaviors. These significant individual factors (educational level, being convinced of the importance of mammography) highlight the importance of patient knowledge and awareness in improving adherence to recommended screening practices. The findings that women who are convinced that mammograms are important are supported by a Netherlands study that examined mammography compliance over a 17-year period and found that women who began screening at approximately 50 years of age, and made a habit of repeat screening, were more likely to continue screening even after reaching the age of 70 (Scaf-Klomp et al., 1995).
Among clinical-site variables, offering a low-cost pap smear at the same site where the mammogram is offered, using hands-on demonstrations with breast models, and using posters to advertise the importance of mammography substantially increase the likelihood of rescreening. These findings suggest that regular breast cancer screening behaviors may be facilitated by integrating mammography into a women's general health care. They further suggest that using concrete methods of breast-health education, such as demonstrations using breast models as well as visually interesting methods such as posters to reinforce the importance of regular mammography screening, can be effective in reaching older low-income women.
Our findings regarding the lack of effectiveness of program brochures and telephone help lines to increase mammography rescreening are also supported in the literature. An intervention study focusing on older women found that a multifaceted intervention such as physician reminders, audit with feedback in which physicians' performance was compared with peers, and a patient educational brochure for waiting-room reading that highlighted the importance of age as a breast cancer risk factor and stressed the benefits of regular mammograms was only modestly successful. Although overall screening rates increased, there was a relatively low rate of patient adherence to physician recommendation for mammography. In particular, educational brochures were generally not effective (Preston et al., 2000).
Although there are many strengths of this study (e.g., the relatively large sample size; the prospective study design; the focus on older, underserved women; and the examination of both individual and site factors), the study is not without limitations. One may be reliance on patient self-report, as some studies have found that women tend to overestimate the frequency of mammography screening (McGovern et al., 1998; McPhee et al., 2002). However, there are other studies that have found that self-report is a reasonably reliable method (Degnan et al., 1992; Etzi, Lane, & Grimson, 1994).
Ruchlin (1997) pointed out that researchers have used overly broad age categories when exploring national data and when referring to established guidelines. This is particularly pertinent in that one of the most critical influences on breast cancer rescreening in older women has been found to be physician or clinician recommendation (Grady et al., 1992; Rimer, 1993), and practitioners often rely on these recommendations when providing services. A limitation of this study is that we have no data to examine whether the often varying or nonspecific age-related screening recommendations and dearth of research on older women contribute to confusion among health care practitioners when they are trying to decide on the recommendation of preventative health care services to older women.
|
Footnotes |
|---|
1 Institute for Health & Aging, University of California, San Francisco.
2 School of Social Work, University of Hawaii, Manoa, HI.
3 Institute of Social Development and Public Policy, Beijing Normal University, People's Republic of China.
4 California Department of Health Services, Cancer Detection Section, Sacramento.
Decision Editor: Linda S. Noelker, PhD
Received for publication September 4, 2002. Accepted for publication May 22, 2003.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. Calvocoressi, M. Stolar, S. V. Kasl, E. B. Claus, and B. A. Jones Applying Recursive Partitioning to a Prospective Study of Factors Associated with Adherence to Mammography Screening Guidelines Am. J. Epidemiol., December 15, 2005; 162(12): 1215 - 1224. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
