|Home | About | Journals | Submit | Contact Us | Français|
This paper examines the prevalence of cancer screening use as reported in 2005 among U.S. adults, focusing on differences among historically underserved subgroups. We also examine trends from 1992 through 2005 to determine if differences in screening use are increasing, staying the same, or decreasing.
Data from the National Health Interview Surveys between 1992 and 2005 were analyzed to describe patterns and trends in cancer screening practices, including Pap test, mammography, prostate-specific antigen (PSA), and colorectal screening. Logistic regression was used to report 2005 data for population subgroups defined by a number of demographic and socioeconomic characteristics.
Rates of use for cancer tests are rising only for colorectal cancer, due largely to the increase in colorectal endoscopy screening. Use of all the modalities was strongly influenced by contact with a physician and by having health insurance coverage.
There remain large gaps in use for all screening modalities by education, income, usual source of care, health insurance, and recent physician contact. These specific populations would benefit from interventions to overcome these barriers to screening.
The National Health Interview Survey (NHIS) has monitored the nation's progress toward Healthy People objectives in cancer screening for nearly two decades.1 A major objective of the Healthy People initiative is to increase use of recommended cancer screening tests and to help reduce disparities in the use of cancer screening services. Repeated monitoring of differences among groups can inform the planning of public health strategies.
When examining cancer screening from the 2005 NHIS, large gaps in use remain among population groups with differing levels of education, income, usual source of care, health insurance, and recent physician contact. Results identify specific populations that may benefit from efforts to overcome barriers to screening.
This report examines the prevalence of cancer screening practices reported by adults in the 2005 NHIS, and trends in screening usage from 1992 through 2005. The U.S. Preventive Services Task Force (USPSTF) reviews scientific evidence on preventive services and is the basis for Healthy People cancer screening objectives. Currently, the USPSTF recommends Pap tests for cervical cancer, mammography for breast cancer, and fecal occult blood test (FOBT), sigmoidoscopy, or colonoscopy for colorectal cancer.2 Although consensus has not been reached on the effectiveness of PSA testing in reducing cancer mortality, its use is widespread and increasing.3-6 Due to its widespread use, the NHIS began monitoring this screening modality in 2000.
Our previous report on trends in screening from 1987 to 2000 demonstrated progress toward Healthy People objectives for all these screening modalities.7 In that study, variables found to influence screening practices included age, gender, race/ethnicity, recency of immigration, income (poverty status), education, health insurance status, and having a usual source of health care. In this study, we test for associations between cancer screening and those variables again, and whether progress continued to 2005.
The NHIS is a large-scale household interview survey of a statistically representative sample of the U.S. civilian non-institutionalized population (http://www.cdc.gov/nchs/nhis.htm). Each year a random, stratified, multistage cluster sample of U.S. households is used to obtain the data. In-person interviews yield a basic set of demographic and health items for each participating household. For each survey year included in this study, respondents were asked to report the time since their most recent cancer screening tests. In 2005, the universe of respondents who were asked to choose a category was expanded from that used in 2000 and 2003. As a result, the rates for 2005 and subsequent years can be calculated in two ways. In Tables 1 and and2,2, we report 2005 results using the expanded group of respondents. In Figure 1, for the sake of consistency with earlier trend analyses, we report the 2005 results limited to the type of data collected and the estimation methods used with the earlier data in 2000 and 2003. Descriptions of the 2005 NHIS and these methods are available online.8
NHIS respondents who report having had a recent test are asked why they received the test. We included tests given that were not part of a routine exam; in short, we assumed that a person who had the test within the recommended interval had been screened. This is a reasonable assumption for mammography, Pap, PSA, and home FOBT, where 10% or less of tests are performed due to a symptom. A much larger percent of colorectal endoscopies result from symptoms,9 so we included only screening tests for this modality. We measured two types of colorectal cancer screening: home FOBT and endoscopy. Each modality was measured independently, so that respondents reporting both tests are included in rates for both. We excluded office-based FOBT from our analysis because it does not have a proven benefit. Recently published American Cancer Society (ACS) guidelines endorse the full range of colorectal screening technologies already in use, emphasizing that tests be offered in the context of shared decision making,10 but do not state a preference for any single test.
We used the baseline and progress measures for the Healthy People 2010 national cancer screening objectives to determine intervals for Pap tests and mammography, 3 years and 2 years, respectively. At the time of data collection, the USPSTF had no recommendation regarding interval for sigmoidoscopy/colonoscopy/proctoscopy (hereafter collectively referred to as “endoscopy”) screening. ACS recommends annual FOBT, flexible sigmoidoscopy every 5 years, or colonoscopy every 10 years. More recently, USPSTF released new recommendations similar to those of ACS except they add an interval FOBT between flexible sigmoidoscopies.11,12 Historical changes in both recommendations and the way data are collected create measurement challenges related to endoscopy usage. The categories used for the time interval question in the 1992 NHIS were within the past year, 1 to 3 years ago, and over 3 years ago, so we chose 3 years as the endoscopy cut-point for all survey years.13-15 Therefore, to make our trend comparisons consistent with these earlier data, we used 3 years as our maximum. For an examination of 2005 colorectal screening that is consistent with the Healthy People 2010 guidelines, see Shapiro et al.16 Finally, for the PSA test, we used the ACS screening recommendation that the test be offered annually.17 Similar to other organizations, the ACS stresses that physicians should be prepared to discuss PSA testing so men can make an informed decision.
Age groups were selected with consideration of screening recommendations and other age-related influences on screening, such as Medicare benefits for those age 65 and older. Although the survey questions on Pap tests were asked of women age 18 and older, educational attainment is often not completed until age 25. Therefore, we limited our analysis to age 25 and older. Mammography analysis focused on women age 40 and older. Recommendations for mammography for women age 40-49 have varied over the years, making this a group of particular interest. PSA and colorectal screening were studied for individuals age 50 and older.
Earlier studies have found that usual source of care (USOC), insurance, and immigration recency, are the strongest predictors of cancer screening.7,18-20 Other covariates included in this study are education, income (as a multiple of poverty ratio), race/ethnicity, and recent physician contact.
We present two independent analyses. For the tables, we present multiple logistic regression analysis showing bivariate associations of selected demographic and health utilization covariates with screening. For ease of interpretation, adjusted percentages (called predicted margins) were derived from the logistic regression analyses. Since predicted margins are adjusted proportions or percentages instead of odds ratios, this makes their interpretation easier for the reader, particularly when the rates of screening are not small and the odds ratios do not approximate relative risks, as is the case for our screening rates. In addition, the values of predicted margins are independent of the reference level used in the logistic regression model, which is not the case for odds ratios.21 For the figure, we present predicted margins derived from logistic regression results that are adjusted for the covariates over time. These covariates, dependent variables, and samples analyzed for each screening modality are described in the Test and Test Interval section above.
For trends, as shown in Figure 1, several years of the NHIS were analyzed using logistic regression. The year of the survey was included as a covariate. Interactions were tested between the access to care variables (USOC and insurance) and year of survey. For each screening modality, four logistic regression analyses were conducted: (1) all the covariates and survey year were included as main effects only; (2) interactions between survey year and health insurance were added to the main effects model; (3) interactions between survey year and usual source of care were added to the main effects model; and (4) both interactions between survey year and health insurance, and survey year and usual source of care were added to the main effects model. The first model was determined to be the best model as it was the most parsimonious. Even though some of the interactions were statistically significant, including the interactions did not change the predicted margins when compared to the first model without interactions. Since the patterns across groups were consistent over time, Figure 1 presents results only from the main effects model. Vertical lines represent 95% confidence intervals.
Unknown responses were excluded from the analyses except for income where, because of the large number missing, a separate missing income category was used. Some categories of variables resulted in wide confidence intervals when the sample sizes in those categories were small, such as for the smaller racial/ethnic groups and for the age-specific estimates among recent immigrants.
The analyses were conducted using SAS version 8.2 and SUDAAN Release 9.0.2 which take into account the complex multistage sample design of the NHIS.8 Wald F-tests were used to obtain p-values for significance tests of the regression coefficients in the logistic regression models including interactions.21 All significance tests were two-tailed with level of significance set at p<0.05.
Tables 1 and and22 present cancer screening test use among women and men, respectively, estimated from the 2005 NHIS, using all available data including the additional responses described above. We describe the statistically significant variables below. Of all the screening modalities measured in 2005, use of Pap test and mammography were highest at 78% and 67%, respectively, while use of home FOBT and endoscopy were low at 12% (men and women) and 20-24% (women-men), respectively. PSA use was intermediate at 41%.
In 2005, 78.4% (CI 77.6-79.2) of women age 25 and older reported having a Pap test within the last 3 years. Among the age groups studied, a gradient of use was found, with women under 40 most likely (86.8%, CI 85.5-88.1) and those age 65 and older least likely to get a Pap test (64.7%, CI 62.8-66.5). Also less likely to have had a test were women with a family income less than 200% of the federal poverty ratio (76.0%, CI 74.5-77.5), or those who lacked a usual source of care (71.7%, CI 69.4-74.1). Screening increased as education level increased, with those with less than a high school education least likely (74.8%, CI 72.9-76.6) and college graduates most likely (82.4%, CI 80.0-84.0). Women having private or military insurance coverage were the most likely to have reported a Pap test (79.9%, CI 78.9-80.9), followed by those with public insurance only (76.7%, CI 74.9-78.5), and the uninsured (74.0%, CI 71.9-76.2). Immigrants were less likely than women born in the United States (78.9%, CI 78.0-79.8) to have had a Pap test, with those in the country less than 10 years least likely (70.8%, CI 66.2-75.4). Women who did not see or talk to a doctor (69.9%, CI 67.6-72.2) or an OB/GYN (70.3%, CI 69.0-71.5) in the past 12 months were less likely to report a Pap test than those who had (80.0%, CI79.1-80.9 for doctor and 94.0%, CI 93.2-94.8 for OB/GYN). Except for the “non-Hispanic other” racial/ethnic category, which is a compilation of very disparate populations, there were no significant differences among Hispanics, non-Hispanic whites, and non-Hispanic blacks after adjustment for the other factors.
In 2005, 67.0% (CI 65.8-68.2) of women age 40 and older reported a mammogram within the past 2 years. Women 50-64 were more likely (71.7%, CI 70.1-73.3) than other age groups to report a recent mammogram, and there were positive associations in the prevalence of mammograms with education and income. Women without a usual source of care were less likely to report a mammogram (54.0%, CI 49.6-58.3) than those with a usual source of care (67.9%, CI 66.7-69.2). Those with private or military insurance were more likely (69.4%, CI 68.1-70.7) to report a recent mammogram than women without insurance (55.5%, CI 51.7-59.3), and women with public insurance only (63.8%, CI 60.8-66.8) were between the two. Women who did not see or talk to a doctor in the past year were far less likely to get a mammogram than those who did (44.4%, CI 40.1-48.8 vs 69.5%, CI 68.2-70.7), as were women who did not talk to or see an obstetrician or gynecologist (58.3%, CI 56.7-59.9 vs 82.5%, CI 80.8-84.2). Race/ethnicity and immigration status did not yield significant differences in mammography use.
In 2005, 11.8% (CI 10.8-12.8) of women age 50 and older reported having a home FOBT within the last year. Among women without a usual source of care, 5.7% (CI 2.0-9.4) reported a home FOBT compared to 12.0% (CI 11.0-13.0) of women with a usual source of care.
Uninsured women were less likely to report a home FOBT (6.9%, CI 4.1-9.8) compared to those with private or military insurance (12.3%, CI 11.1-13.6). Women who did not see a physician were less likely than women who did (2.5%, CI 0.6-4.4 vs 12.4%, CI 11.3-13.4). In 2005, 12.3% (CI 11.3-13.3) of men age 50 and older reported a home FOBT within the last year. Men 65 and older were more likely to have the test (15.6%, CI 13.7-17.5). There was an increasing gradient with increasing education, and men who were uninsured (6.3%, CI 2.7-9.9) or did not see or talk to a doctor in the past year (3.2%, CI 1.2-5.2) were very unlikely to be screened. For both women and men, the adjusted analysis resulted in no significant differences for race/ethnicity or immigration status.
In 2005, 19.8% (CI 18.5-21.0) of women age 50 and older reported having colorectal endoscopy for screening purposes within the last 3 years. Rates were positively associated with education and income. There was a large difference in screening between women who had seen or talked to a physician (20.6%, CI 19.3-21.8) and those who had not (7.3%, CI 3.9-10.6). A smaller difference was seen for those who had or had not seen or talked to an obstetrician or gynecologist in the past year (23.1%, CI 21.0-25.3 vs 18.2%, CI 16.7-19.6, respectively). In 2005, 23.7% (CI 22.3-25.0) of men age 50 and older reported a colorectal endoscopy for screening purposes within the last 3 years. More likely to be screened were men 65 and older (28.2%, CI 25.8-30.5), those with a usual source of care (24.1%, CI 22.7-25.6), and those who saw or talked to a doctor in the past year (25.5%, CI 24.0-26.9). Again there was a positive gradient with increasing education and age, and men with private or military coverage were much more likely to have the test (24.7%, CI 23.1-26.3) than those with no insurance (14.1%, CI 8.4-19.8). There were no significant differences among racial/ethnic or immigration groupings for men. There were racial-ethnic differences in colorectal endoscopy screening for women because rates were lower for non-Hispanic other women.
In 2005, 41.4% (CI 39.7-43.1) of men age 50 and older reported a PSA test within the last year. Men age 65 and older were more likely to get the test (52.3%, CI 49.8-54.8), as were those with a usual source of care (42.3%, CI 40.5-44.1). Men who saw or talked to a physician in the past 12 months (45.0%, CI 43.2-46.9) were more likely to report screening than their counterparts (9.8%, CI 6.6-12.9). Those without insurance were less likely (32.8%, CI 25.7-39.8) to have the test, compared with men having private or military insurance (43.3%, CI 41.4-45.2). No significant differences were found among racial/ethnic or immigrant groupings after adjustment for other factors.
Figures 1A and 1B present trends in cancer screening among women and men, respectively, from 1992 to 2005, adjusted for all characteristics shown in the tables. The vertical lines represent 95% confidence intervals. For the sake of consistency in computing the trends, we used the 2000 and 2003 method to compute the 2005 results.
Pap test use among women remained close to 80% over the entire period, although there is a suggestion of a downward trend from 2000 to 2005. Mammography rates rose steadily from 63% in 1992 to 70% in 2000, and then began to drop to 68% in 2005. Trends among women show adjusted rates for home FOBT rose from 1992 (12%) to 2000 (18%) and then fell. In 2005, it returned to the 1992 rate (12%). Among men, the adjusted rate for home FOBT rose from 1992 (10%) to 2000 (17%), was nearly unchanged until 2003 (16%), and then dropped in 2005 (13%). Colorectal endoscopy among women increased slightly between 1992 (8%) and 1998 (10%) and then rose rapidly through 2005 (20%). Adjusted rates for endoscopy rose overall for men from 14% in 1992 to 24% in 2005. Lastly, the NHIS monitored PSA for the first time in 2000. Looking at the adjusted trends, PSA use rose from 2000 (42%) to 2003 (45%), then dropped in 2005 (42%).
We found that screening rates have changed over time. Between 2003 and 2005, colorectal endoscopy screening rose, Pap testing was stable, PSA testing dropped, and as previously reported, mammography dropped.22 The lowest screening rates were found for persons without a usual source of care, those who had no physician contact in the past year, and the uninsured. The patterns of disparities we found in 2005 were consistent with those found in previous years of the NHIS.18 A review of other studies shows that factors associated with disparities have remained similar over recent years.7,18,23-25 However, it is important to note that, after adjustment for the other variables studied, race/ethnicity and immigration status did not yield significant differences in test usage for distinct racial-ethnic groups. It was the factors more directly related to the individual's interaction with the health care system that resulted in significant disparities.
The only cancer site for which screening increased in 2005 was colorectal, and this was due to an increase in use of endoscopy. Even when test rates increased, people without insurance or physician contact were not screened. These patterns of disparities are most clearly illustrated by colorectal endoscopy use in 2005, after the period of rapid uptake (Figure 1). The tables show that, among the uninsured, 2005 endoscopy rates were 6.5% for women and 14.1% for men. Rates were similarly low for those who did not see or talk to a physician within the past year. Even for women and men who saw a doctor within the past year, screening rates were conspicuously low, 7.3% and 8.0%, respectively.
Though use of all the modalities was strongly influenced by contact with a physician, its effect on PSA use was particularly striking. Forty-five percent of men who saw or talked to a doctor in the past 12 months reported a PSA test; in contrast, among men who did not report physician contact, only 9.8% reported the test.
As the nation considers changes to health care, it is critical to rely on scientific evidence of the effectiveness of clinical guideline dissemination. There are three essential components to the successful delivery of cancer screening: coverage, guidelines, and tracking systems.26 Few health plans in the United States had all three in place in 2000. Group and staff model HMO plans were far more likely than other types of plans to have any system at all for screening delivery.26 Successful cancer screening systems require the alignment of plans and clinician efforts at the practice level.27 By accomplishing that, and by adhering to cancer screening guidelines and a tracking system, Veterans Affairs medical centers had higher screening rates than surrounding communities in 2004.28 The greatest benefit from screening would come from extending such a delivery system to people who have never been screened for cancer according to the guidelines. Now, more than ever, we need to bring proven delivery strategies to those individuals.29 Two large public health efforts are doing that. CDC's Breast and Cervical Cancer Program has increased mammography and Pap testing rates, but it appears to be underfunded.30,31 NCI's Cancer Collaborative Project with the Health Resources Services Administration (HRSA) is enhancing primary care services in underserved areas to include cancer screening referrals, follow-up, and treatment.32 Even so, evidence-based cancer screening coverage of underserved populations remains incomplete.
Study limitations include the use of self-reported data and exclusion of institutionalized and non-civilian populations. In addition, American Indian, Alaska Native, Asian American and Pacific Islander samples were too small to analyze separately and were included in a “non-Hispanic other” group. We report on receipt of endoscopy within 3 years, but screening guidelines indicate an interval of 5 years for sigmoidoscopy and 10 years for screening colonoscopy. Therefore, we underestimate the percentage of the population that is “up to date” for colorectal screening using current guidelines. Other studies have suggested that about 50% of the eligible population is currently in compliance with colorectal screening guidelines.12,16,33-35 Survey questions are not always identical from year to year, a limitation that was especially problematic for our analysis of endoscopy. Another change in data collection in 2005 reduced the number of missing responses and slightly altered recent screening rates.
Reducing late-stage cancers is a top public health priority. High rates of screening throughout the population are necessary to accomplish this goal because routine screening is important for identifying cancer at an early stage, when it is most treatable. Studies in breast29 and cervical36 cancer screening in organized screening programs have found that the majority of late-stage disease occurred among women never screened. So, there is a critical need to identify effective approaches to increase use among those never screened to reduce the incidence of late-stage cancers. It is equally critical that suspicious tests receive timely follow up. Ensuring health insurance coverage and a usual source of care are the factors most likely to increase use among those never screened. Because trends for most screening modalities are stable or falling and disparities have remained large, we need to continue to carefully monitor rates.
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or the U.S. Census Bureau.
There are no financial disclosures from any of the authors.