Human papillomavirus and cervical cancer behavioral surveillance in the US§


  • During the completion of this article, Dr. Tiro was supported by the National Cancer Institute's Cancer Prevention Fellowship Program and the Division of Cancer Control and Population Sciences.

    We thank Tim McNeel and William Waldron from Information Management Systems and David Woodwell from the National Center for Health Statistics for their assistance in preparing this article.

  • The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the National Cancer Institute or the Centers for Disease Control and Prevention.

  • §

    This article is a US Government work and, as such, is in the public domain in the United States of America.


In the US, federal and state behavioral surveillance systems routinely monitor self-reported sexual behavior and Papanicolaou (Pap) test use to identify high-risk populations, trends, and disparities and to guide and evaluate interventions for cervical cancer prevention and control. Clinical uptake of human papillomavirus (HPV) vaccination and testing necessitates the expansion of behavioral surveillance systems. Cervical disease is the main focus of HPV-related behavioral surveillance because of greater cancer incidence and mortality relative to other susceptible organs, and the availability of effective technologies for prevention and control. In the current study, a framework is presented for the types of behaviors to monitor, their conceptual and operational definitions, target populations, and evidence supporting the reliability and validity of self-report measures. An overview is also provided of 8 population-based and 2 provider-based data systems that are nationally representative and accessible for behavioral surveillance research. Ongoing surveillance at the national, state, and local level is critical for monitoring the dissemination of HPV technologies and their impact on reducing disparities in the detection of precursor lesions, incidence of invasive cancer, and mortality. Cancer 2008;113:(10 suppl):3013–30. Published 2008 by the American Cancer Society.

Invasive cervical cancer is 1 of the few cancer sites that could be nearly eliminated in the US if prevention, screening, and treatment methods for precursor lesions were universally available and appropriately used. Yet in 2004 there were 3850 deaths and more than 11,892 incident cases reported.1 The recent release of clinical guidelines recommending the human papillomavirus (HPV) test and quadrivalent vaccine have created new prevention and early detection opportunities.2–8 Integration of HPV-related technologies into clinical care and cervical control programs necessitates the expansion of national behavioral surveillance systems and requires researchers from different areas (ie, cancer, sexually transmitted infections [STIs], and immunization) to work together. Monitoring of population and provider use of prophylactic immunization, the Papanicolaou (Pap) test and HPV test is essential to understand their combined effect on reducing morbidity and mortality and to evaluate public health interventions.

This article describes federally funded, publicly available behavioral surveillance systems in the US for tracking population and provider knowledge, attitudes, and practices toward cervical HPV infection and cancer. Cervical disease is the main focus of HPV-related behavioral surveillance because of greater cancer incidence and mortality relative to other susceptible organs, and also the availability of prevention and control technologies. First, we define behavioral surveillance and describe aspects of survey methods to monitor trends. Second, we describe key HPV-related behaviors, conceptual and operational definitions, and target populations. Third, we review the characteristics, content, and recent estimates of key behaviors from national population and provider surveys. Finally, we present future surveillance needs.

Definition and Methods of Behavioral Surveillance


For the purpose of public health practice, we define behavioral surveillance as the systematic, ongoing collection and reporting of data regarding health promotion and risk reduction behaviors to:

  • Monitor trends in individual and provider behaviors associated with changes in disease over time;

  • Identify populations at high risk;

  • Track changes in individual and environmental correlates of behavior; and

  • Guide the development, implementation, and evaluation of programs that encourage protective behaviors and discourage risky behaviors.9–11

Surveillance data can come from surveys (self-reports from a sample of the target population) as well as registry systems. Historically, the systematic collection of data regarding Pap test use in national population surveys has been helpful for identifying underserved populations with poor access to care. Use of the Pap test is now considered a key marker for US health disparities,12, 13 because national surveys have shown that racial and ethnic minorities, women of low socioeconomic status, and those living in specific geographic regions have a higher cervical cancer incidence and mortality.14, 15 Policymakers and public health practitioners have responded to identified disparities by creating and implementing effective programs such as the National Breast and Cervical Cancer Early Detection program.16, 17 By expanding cervical cancer behavioral surveillance to now encompass the newer technologies of the HPV vaccine and test, researchers can track disparities in uptake for all technologies, and policymakers and practitioners can more efficiently allocate resources for underserved populations.

Because the HPV vaccine and test have only recently been introduced, measures are being developed for national surveillance systems. A key goal is to ensure that self-reported measures for all technologies are reliable and valid and are implemented consistently across surveillance systems. In this way, patterns and trends can be compared and more thoroughly examined throughout the US population.

Population-based surveys

Large national probability surveys of the US population are an important source of health data. The 8 surveys described in Table 1 can be used to assess trends for items that are consistently measured over time. Multiple surveillance systems allow for differences in survey objectives, target populations (children, adolescents, young adults, parents), sampling strategies (eg, oversampling of minorities, high‒risk groups, or local areas), frequency of implementation (eg, annual, biennial, periodic), mode of administration (ie, face-to-face, telephone, paper, audio computer-assisted self-interviewing [ACASI]), and biologic specimen availability.

Table 1. Description of Population Surveys Assessing Cervical Cancer Prevention and Control Behaviors
Survey AttributesBehavioral Risk Factor Surveillance System (BRFSS)(Available at: Information National Trends Survey (HINTS)(Available at Health and Nutrition Examination Survey (NHANES)(Available at Health Interview Survey (NHIS)(Available at Immunization Survey (NIS)(Available at Survey of Family Growth (NSFG)(Available at Survey of Children's Health (NSCH) (Available at Risk Behavior Surveillance System (YRBS) (Available at
  1. Pap indicates Papanicolaou test; RDD, random digit dialing; CATI, computer-assisted telephone interviewing software; CAPI, computer-assisted personal interview; ACASI, audio computer-assisted self-interviewing; CASRO, Council of American Survey Research Organizations; AAPOR, American Association for Public Opinion Research; STDs, sexually transmitted diseases; HPV, human papillomavirus.

General purposeTracks health conditions and risk behaviorsTracks adults' knowledge and use of cancer-related informationTracks the health and nutrition of adults and children through interviews and physical examinationsMonitors trends and patterns in illness and disability and tracks progress toward achieving national health objectivesMonitors infant, teen, and adult immunization coverage; infant and teen vaccinations are provider-verifiedTracks fertility, infertility, family planning, childbearing, contraceptive practices, and other aspects of maternal and child health to gauge the effects of these processes on population growthMonitors the physical and emotional health and access to health care of children; a member of the State and Local Area Integrated Telephone Survey (SLAITS) ProgramMonitors health risk behaviors among high school students and tracks progress toward achieving teen-specific national health objectives
Geographic sampleNational and state levels; plus selected metropolitan and micropolitan statistical areas; oversampling by statesNational level; oversamples Hispanics and non-Hispanic blacksNational level; oversamples some groups depending on the NHANES cycle; 2003-2004 cycle oversampled adolescents (ages 15-19 y), non-Hispanic blacks, and MexicansNational level; oversamples Hispanics and non-Hispanic blacks (1995-) as well as ensures participants from each stateNational level; oversampling may occur in selected local areas chosen by state grantees; in 2008, the NIS-Teen survey will be state level with selected local areasNational level; oversamples of adolescents (ages 15-24 y), Hispanics, and non-Hispanic blacksNational and state levelsNational and state levels, plus selected local areas; national sample based on a 3-stage cluster design of counties, schools including grades 9-12, and classes; oversamples Hispanic and non-Hispanic black students
EligibilityNoninstitutionalized, civilian household population, aged ≥18 yNoninstitutionalized, civilian household population, aged ≥18 yNoninstitutionalized, civilian household population ages 14-59 yNoninstitutionalized, civilian household population aged ≥18 yNoninstitutionalized, civilian household population, teens (ages 13-17 y); adults (ages 18-65 y)Noninstitutionalized, civilian household population ages 15-44 yNoninstitutionalized, civilian household population ages birth-17 y9-12th grade students attending public and private schools
Frequency of implementationAnnual, continuous cycleBiennial; odd y2-y, continuous cycleAnnual, continuous cycle (cancer module: 1987, 1992, 2000, 2003, and 2005)Teen: annual; adult: periodic2-y continuous cyclePeriodicBiennial, odd y
Format of surveyCore module: Pap items; optional: immunization itemsTopic-specific moduleTopic-specific module and self-collected vaginal swab samplePeriodicTopic-specific moduleCoreCoreCore
Mode of administrationRDD using CATI softwareRDD/CATI (2003 and 2005); dual frame: RDD/CATI and self-administered paper questionnaire (2007)Face-to-face using CAPI software and ACASI for sensitive questionsFace-to-face/CAPI since 1996RDD/CATIFace-to-face/CAPI and ACASI for sensitive questionsRDD/CATISelf-administered paper questionnaire
LanguagesEnglish and SpanishEnglish and SpanishEnglish and SpanishEnglish and SpanishEnglish and SpanishEnglish onlyEnglish and SpanishEnglish only
Most recent sample size356,112 (2006)5394 (2005)2215 surveys; 2026 vaginal swabs (2003-2004)98,649 (2005)5468 teens (2006); 7055 adults (2007)2002: 12,571 (7643 females and 4928 males)102,353 (2003-2004)13,953 completed questionnaires (2005)
Most recent response rateCASRO* rate: 51.1% (state range: 34.6%-67.4%)AAPOR*: 20.8%; Screener: 34%; Extended Interview: 61.3%Unweighted: 80%; Adequate swabs: 94.8%Household: 86.5%; Final adult: 69.0% (2005)CASRO: Teen - 56.2% (2006); Adult - 32% (2007)AAPOR: 79%CASRO: 55.3% weightedOverall: 67%;
School: 78%;
Student: 86%
IncentiveNoneRandomized to $0, $5, or $15 incentive (2005)NoneNoneNone. Selected cases $15$40NoneNone
Measures Included        
Years1992-20082003, 2005, and 20072003-20041987, 1992, 2000, 2003, 2005, and 2008 (cancer module)2006 and 2007Women only: 1973, 1976, 1982, 1988, and 1995; men and women: 2002 and 200720071991-2007 (odd years)
Sexual behaviorNoneNoneEver had sex (vaginal, anal, oral sex); age of first sex if at least 1 lifetime sex partner. NoneAge at first sex; partner characteristics, reasons for not having sex (among virgins), number of partners (lifetime and past 12 mo); oral sex, anal sex, contraceptive use (lifetime and past mo)NoneEver had sexual intercourse; had first sexual intercourse before age 13 y; had sexual intercourse with > 4 people during lifetime; currently sexually active (had sexual intercourse with > 1 person during 3 mo preceding survey); condom use (respondent or partner used a condom during last sexual intercourse)
For those ages 14-17 y who were sexually active, additional questions included: age at first sex; number of lifetime partners; ever use of a condom.
For those age >18 y who were sexually active, questions also included: number and gender of sex partners in the last 12 months; lifetime sex partners; past history of STDs.
HPV vaccineOptional state module—women ages 18-49 y and parents of female child ages 9-17 y (2008): Ever had; no. of doses receivedWomen ages 18-64 y (2005; men added in 2007): Heard of HPV; HPV knowledge (causes cervical cancer, STD, goes away on its own without treatment, provider said had HPV infectionEver had (2008)Women and Men 18-64 (2008): Heard of HPV; HPV knowledge (causes cervical cancer, STD); heard of vaccineParents of teens ages 13-17 y (2006): HPV and HPV vaccine knowledge; daughter ever had; no. of doses; reasons for nonreceipt; future likelihood of getting itWomen ages 15-24 y and mothers of female child ages 9-18 y (2007): daughter ever had; intent to vaccinate in next 12 mo; reasons for nonreceiptParents of teens ages 12-17 y (2007): daughter ever had; provider recommendationNone
Women and Men ages 18-64 y (2007): source of HPV knowledge; heard of vaccine; provider communicationWomen ages 18-64 y (2008): ever had vaccine; number of doses; interested in getting vaccine; reasons for nonreceipt; get if had to pay $360; get if free or low costAdults ages 18-49 y (2006, 2007): HPV and HPV vaccine knowledge; ever had; no. of doses; reasons for nonreceipt; provider recommendation
Parents of female child ages 9-12 y (2007): vaccine acceptability; reasons for non-receiptParents of female child ages 8-17 y (2008): ever heard of vaccine; daughter ever had; no. of doses; give if doctor recommended; reasons for nonreceipt
Pap testCore: Ever; when most recent (included every even y; ever had hysterectomyEver; when most recent; ever had hysterectomyNoneEver; when most recent; ever had hysterectomyNonePelvic examination within past 12 mo; Pap test past 12 moNoneNot relevant for target population
Additional measures of interest: gets test on a regular schedule; reason for most recent test; when next test; ever treated for genital wartsAdditional measures of interest: number in past 6 y; reason for nonreceipt of a recent test; provider recommendation in the past (12 mo asked in 2005; 3 y asked in 2008); ever had abnormal Pap
HPV testNoneProvider communication about HPV test (2005)Self-collected vaginal swab specimenNoneNoneNoneNoneNot relevant for target population

Each surveillance system has unique strengths, such as the physical examination and biospecimen components of the National Health and Nutrition Examination Survey (NHANES; available at: accessed March 31, 2008), or the National Immunization Survey's (NIS; available at: accessed March 31, 2008) validation of parents' reports of pediatric vaccination by providers. The unique sampling frames are an additional strength. For example, the Behavioral Risk Factor Surveillance System (BRFSS; available at: accessed March 31, 2008) provides estimates for all 50 states and the District of Columbia (DC), and some metropolitan and micropolitan areas that have at least 500 respondents. The National Survey of Family Growth (NSFG; available at: accessed March 31, 2008) focuses on reproductive-age populations, whereas the Youth Risk Behavior Surveillance System (YRBS; available at: accessed March 31, 2008) is a collection of surveys that use a school-based sampling frame to identify adolescents in Grades 9 through 12.

Overall, survey response rates have been declining for several decades.18 Recent response rates range from 86.5% for the National Health Interview Survey (NHIS; available at: to 20.8% for the Health Information National Trends Survey (HINTS; available at: Variations in results from these 2 surveys are difficult to understand given the different modes of administration, sampling frames, and methods for calculating response rates (eg, Council of American Survey Research Organizations [CASRO]19 vs American Association for Public Opinion Research [AAPOR]20). Generally, face-to-face surveys, such as NHIS, yield higher response rates compared with telephone surveys such as HINTS. Declines in telephone surveys are most likely because of technologies that allow individuals to screen and block calls21 and the increasing number of adults who only have cellphones (reportedly 6.7% of households in early 200522). The impact on survey estimates is conflicting, with some arguing that lower rates may not reflect the most important source of bias,18, 23 whereas others express concern about sampling undercoverage because of the growth of cellphone-only households.24

Probability sampling methods in the 8 surveys yield results that are representative of the US population, which is diverse with respect to sociodemographic and geographic covariates. Findings from large-scale, national data, however, often cannot be disaggregated to subpopulations or local communities. Small area data collection and analysis are increasingly requested by local governments, researchers, policymakers, and providers to understand disease burden and engage in strategic planning to improve the health of their residents. Depending on the disease and research question, local areas could be defined as state, county, census tract, health district, and hospital service area.25, 26 For example, the BRFSS created the Selected Metropolitan/Micropolitan Area Risk Trends (SMART) program to provide stable estimates of particular geographic subdivisions that have a substantial population nucleus with at least 500 survey respondents. Data from the SMART program have been used to investigate local-area factors hypothesized to be associated with cervical cancer screening.27, 28 To meet the data needs of researchers and policymakers at the county or subcounty level, surveillance efforts that encompass smaller geographic units must be supported.29 For example, the California Health Interview Survey (CHIS; available at: is to our knowledge the only state-based health survey that provides robust estimates on the most prevalent racial and ethnic groups living in California and for the majority of the state's 58 counties.

Provider surveys

Routine surveillance of provider practices occurs through the review of administrative claims and medical record databases, as well as provider self-reports.29 Knowledge and attitudes toward clinical practices are not routinely ascertained as part of a national surveillance system; however, provider surveys are periodically conducted to obtain important information regarding cross-sectional and temporal patterns of healthcare. The 2 surveys described in Table 2 can be used to assess provider trends. These surveys are particularly useful during the initial dissemination of new technologies, and can help identify research questions needing further inquiry.29 In addition, adherence to clinical guidelines, provider adoption rates, and the off-label use of emerging technologies can be estimated, yielding important information not available from population surveys because awareness of specific procedures or treatment details is limited. When parallel items are simultaneously included in population and provider surveys, it is possible to compare perceptions of health services utilization, individual and system-level barriers to care, and other domains.30 Findings can help provide a framework for implementing organizational changes and other interventions to improve quality of care.

Table 2. Description of Provider Surveys Assessing Cervical Cancer Prevention and Control Behaviors
Survey AttributesNational Ambulatory Medical Care Survey (NAMCS) (Available at Hospital Ambulatory Medical Care Survey-Outpatient Department (NHAMCS) (Available at
  1. Pap indicates Papanicolaou test; HPV, human papillomavirus.

General purposeAssess the no. of visits at which Pap and HPV tests are ordered or providedAssess the no. of visits at which Pap and HPV tests are ordered or provided
Sampling frameVisits to private office-based physicians and providers at community health centersVisits to outpatient departments of nonfederal, short-stay hospitals (average stay <30 d)
EligibilityPrivate offices of physicians and providers at community health centersClinics of general medicine, surgery, pediatrics, obstetrics and gynecology, and other fields (eg, neurology and psychiatry). Ambulatory surgery centers and ancillary services are excluded
Frequency and y of implementationAnnually from 1973 to 1981, in 1985, and since 1989Annually since 1992
Format of surveyPaper-basedPaper-based
Mode of administrationPersonal interviews and chart abstractionPersonal interviews and chart abstraction
Most recent sample size3712 physicians (2007)1102 clinics in 464 hospitals (2007)
Most recent response ratePending (2005 response rate was 61.5% unweighted)Pending (2005 response rate was 74.1% unweighted)
Measures included:  
 HPV vaccineNoNo
 Sexual behaviorNoNo
 Pap testSingle checkbox (prior to 2006); conventional, liquid-based, or unspecified checkboxes (2006+)Single checkbox (prior to 2006); conventional, liquid-based, or unspecified checkboxes (2006+)
 HPV testYes, if used (2006+)Yes, if used (2006+)
 Diagnosis and proceduresYesYes (2006+)

Professional association databases, such as those of the American Medical Association (AMA) alone or in combination with the American Osteopathic Association (AOA), are typically used as sampling frames for provider surveys. The advantages of using the AMA and AOA master files for random sampling are that all licensed US physicians are included regardless of membership and a range of standard physician and practice characteristics are recorded. This enables researchers to stratify by specialty, geographic area, or other characteristics of interest. Limitations of provider surveys include incomplete or outdated information in professional association databases, poor response rates, and inaccurate reporting of clinical practices. Methods have been developed to improve provider response rates31 and accuracy.32

Conceptual Framework of the Key Behaviors and Target Populations

There are 6 key health outcomes in the progression from HPV infection to cervical cancer that correspond with key prevention and control behaviors (Fig. 1). The 6 outcomes are prevention of HPV infection, exposure to HPV, detection of an HPV infection, detection of a persistent infection, diagnosis of precursor lesions or invasive cancer, and treatment. The parallel behaviors are, respectively, vaccine utilization, sexual behavior, cervical cancer screening with either the Pap or HPV test (depending on patient age), follow-up to an abnormal test, triage to colposcopy and biopsy, and management with ablative or excisional treatment modalities. Access to care is a critical factor influencing whether women engage in all of these behaviors, and poor access is strongly associated with various health disparities. National surveillance of follow-up, biopsy, and management through population surveys is difficult for many reasons, including the ability of women to accurately report follow-up to an abnormal Pap test, a low prevalence of biopsy and management procedures, and the cross-sectional nature of the surveys.33–35 Given these limitations, monitoring of the latter 3 behaviors is best conducted using national registry systems or clinical surveillance systems at the local and regional levels (for a detailed description of relevant data systems, see Wingo et al29). Therefore, we limited our discussion to the 3 behaviors related to risk and screening. Evidence supporting the reliability and validity of these self-reported behaviors is also reviewed.

Figure 1.

Key health behaviors and outcomes related to the prevention and control of cervical cancer incidence, morbidity, and mortality. HPV indicates human papillomavirus; Pap, Papanicolaou test.

Vaccine utilization

Epidemiologic studies have shown that rates of HPV infection are high after first sexual intercourse and rise with the addition of new partners.36 Efficacy trials of the quadrivalent HPV vaccine in females have shown that it functions as a preventive, not therapeutic agent.37, 38 Therefore, the vaccine is most effective if administered before the onset of sexual activity. In 2005, 46.8% of US adolescents in Grades 9-12 reported being sexually active.39 The Advisory Committee on Immunization Practices recommends routine vaccination for girls ages 11 through 12 years and has approved the vaccine for females ages 9 through 26 years.7

Given the age recommendation for routine vaccination and issues regarding parental consent, the 2 key target groups for surveillance efforts are adolescent girls and their parents. To monitor uptake, it is important to know whether a girl and her parents have heard of the vaccine and how many doses she has received (ie, 0-3). To evaluate access, it is useful to know about insurance coverage, whether she has a usual source of care, whether her provider recommended the vaccine, whether she and/or her parents refused and, if so, why. To track catch-up and off-label vaccine use, it would also be of interest to expand the target population to females aged >26 years and males who report having sex with men, a subgroup at higher risk for anogenital cancer.40

To our knowledge, studies evaluating the test-retest reliability of self-reported HPV vaccine use have not been conducted to date, although limited data regarding other vaccines suggest that responses do not change over time.41 Nor are there validity studies of self-reported HPV vaccine use; however, the NIS is comparing parental report of adolescent vaccinations with provider records (data are expected in late 2008). Findings from studies of other age groups and vaccines may provide some insight. Validity studies have shown that self-report measures of children and adult vaccine use suffer from low sensitivity and specificity.42 For childhood vaccines, the sensitivity and specificity of parents reporting their children's vaccination are reported to be 65% and 47%, respectively, for pneumococcal vaccine.43 Adults' self-reports are slightly better, with sensitivity and specificity estimates of: 98% and 80%, respectively, for influenza vaccine; 80% and 77%, respectively, for pneumococcal vaccine; and 57% and 38%, respectively, for tetanus for those aged ≥18 years.44

Accurate self-report measures may be especially important because recent studies suggest that medical records may not be a ‘gold standard’ for vaccine use. Zimmerman et al45 found poor documentation by healthcare personnel in adult medical records. Adolescents may have incomplete medical records because of record scatter (ie, immunizations from >1 provider),46 a lack of a usual source of care in which all medical records are kept, and the use of nontraditional providers.47 Furthermore, parents may be unaware of all adolescent vaccinations because some, such as hepatitis B, may be administered without parental consent.48 Collectively, these findings suggest that there is no gold standard against which to measure self-reported vaccine use and that self-reports concerning the HPV vaccine may suffer from measurement error, especially from parents reporting about their adolescents. Therefore, validity studies evaluating adolescents' self-reports of HPV vaccine use in healthcare settings with strong medical record and auxiliary data systems are needed to understand the magnitude and direction of biases attributable to under-reporting or over‒reporting and to adjust national prevalence estimates.

Sexual behavior

Monitoring sexual behavior is a critical component of HPV and cervical cancer prevention activities because it has implications for identifying exposure opportunities and the optimum age range for vaccination. Adolescents and young adults (ages 15 years-24 years) are the population subgroups most in need of surveillance regarding sexual behavior because of the high prevalence of multiple sexual partners and HPV infection. Self-reports of sexual behavior are widely used in epidemiologic, health services, and behavioral research and commonly include information concerning sexual activity (vaginal, anal, and/or oral), age of initiation, number of partners (lifetime and recent), and frequency of protected and unprotected sex.

Unlike behaviors that can be observed or are less socially stigmatized, sexual behavior is inherently hidden from direct assessment. For this reason, the reliability and validity of self-reported sexual behaviors are difficult to assess and the accuracy of self-reported sexual behavior has been questioned since Kinsey's surveys of sexuality in the 1940s.49, 50 Nevertheless, reliability studies evaluating test-retest and internal consistency have found relatively high agreement when surveys are administered under the same conditions using meaningful reference time intervals.51, 52

The validity of self-reported sexual behavior has been called into question by studies comparing self-report with biologic outcomes such as STIs.53, 54 Dunne et al53 found HPV DNA in 5% of women who reported never having had sex. Peterman et al55 argue that related measures such as condom use may be problematic because utilization is often complex and situationally variable. Nevertheless, self-reports of sexual behavior represent the most feasible and ethical data collection mode.56 After reviewing the literature, Weinhardt et al56 recommended the following strategies to improve accuracy: compare patterns of results among surveys with similar samples; match paired partner responses when appropriate; provide anchor dates for reporting periods; encourage recall via the use of appointment calendars; appropriately sequence items within questionnaires; and use open‒response formats.

Cervical cancer screening

Theoretically, cervical cancer should be almost completely preventable through Pap testing and the treatment of precursor lesions; however, disparities in mortality still exist, especially among minority, immigrant, and low socioeconomic subgroups.14, 15, 57 Retrospective review of the screening histories of women diagnosed with cervical cancer, and data from national and state surveys, indicate that these subgroups are less likely to have been recently screened.58, 59

Regulatory approval of the HPV test in 2003 has provided more screening options. Guidelines for the efficient and cost-effective use of this test are complex; thus, it is critical to know which test or combination of tests is being used in practice. Several professional organizations have published guidelines on cervical cancer screening.2–5 As a whole, they agree that Pap testing should begin by age 21 years or within 3 years of a female becoming sexually active. At a minimum, the screening interval should be every 3 years according to the US Preventive Services Task Force (USPSTF).4 The American Cancer Society (ACS) and the American College of Obstetricians and Gynecologists state that the HPV test can be used as an adjunct to Pap testing for women aged >30 years, and if the test is negative, the screening interval can be extended to 3 years.2, 3 There is less consensus regarding when to stop screening, with the USPSTF recommending age 65 years for women with average risk and the ACS recommending age 70 years. Given what we know about sexual activity, the target age range for surveillance is between 18 and 65 years; however, it is anticipated that screening guidelines will change for vaccinated populations (eg, older age at initiation and longer screening intervals).60, 61

With respect to Pap tests, several operational definitions have been used to assess prevalence, correlates, and intervention effects. These include ever, recency of use (eg, within the past 3 years), and compliance with guidelines (eg, repeat or regular).62 The range of outcomes measured makes comparisons among studies challenging. Two studies comparing different operational definitions for mammography found that although prevalence estimates varied greatly, the different definitions had little effect on the patterns of association with correlates.63, 64 The pervasive adoption of the Pap test (95% of US women reported having undergone a Pap test in 200059) has implications for which operational definitions will be most informative for tracking disparities and evaluating interventions. For example, increasing screening among the proportion of women who never had, or have not recently had a recent Pap test, is likely to result in larger population health benefits than increasing the frequency of screening among those who have been regularly screened but are slightly off-schedule. In addition, other measures of interest include: age at initiation, reasons for not getting a recent test, provider recommendation, had a recent abnormal result, and combined use of Pap and HPV tests. Several operational definitions have been used across surveys and over time (Table 1) and items using open‒response formats such as on the NHIS have proven most useful and flexible as guidelines change. At a minimum, questionnaires need to provide clear definitions and measures to facilitate interpretation and enable comparisons with other studies.62

To our knowledge only 2 studies to date have examined the test-retest reliability of self-reported Pap tests.65, 66 Massachusetts BRFSS data showed that nonwhite women reported Pap testing less consistently over time than white women. A review of validity studies comparing Pap self-reports with medical records found low specificity estimates, leading to a consistent pattern of over‒reporting.62 The highest rates of over‒reporting were found among studies conducted in county health departments and public health clinics, and with ethnic populations. Possible reasons for over‒reporting include: respondents' use of schemas (eg, associating behavior with a pattern of events such as a regular check-up)67 and cultural differences.68 Recent studies have called into question the completeness of medical records in certain settings.69, 70 Additional research is needed to understand the influence of ethnicity, setting, data source, and data quality on validity estimates.

With respect to the HPV test, its characteristics are very similar to the Pap (eg, both require taking a sample of cervical cells); thus, many women may not be able to accurately report which test they received unless their provider clearly communicated test results and they recall results during the survey interview. Similar to an effort described in the colorectal cancer screening literature,71 qualitative research in the form of cognitive interviews with patients, consultation with medical experts, and observations of physician-patient encounters are needed to help survey researchers understand how to best describe the characteristics of the HPV test and design a utilization measure that promotes accurate self-reports.

Clinical practices of providers

The introduction of the HPV vaccine has occurred alongside a paradigm shift for cervical cancer screening. New screening technologies such as liquid-based cytology and HPV testing have been incorporated into many clinical guidelines. With the advent of new knowledge regarding cervical disease progression and technologies, recommendations have changed, such as raising the age at which to begin screening, increasing the screening interval, and less aggressive management for adolescents with abnormal cytology.2, 3, 5 Although the HPV vaccine has not immediately changed screening guidelines,7 models demonstrate that widespread vaccine uptake will reduce the predictive values of cytology, HPV tests, and colposcopy.72 In addition, due to the reimbursement of these emerging technologies at rates higher than Pap tests, and patient demand created by direct-to‒consumer marketing, providers may use tests other than as recommended, thereby increasing costs related to diagnosing and treating cervical disease.73, 74 A provider surveillance system that can capture uptake, changes in use, and off-label uses as well as knowledge, attitudes, system factors, and other determinants of clinical practice will be vital in the postmarketing surveillance era.

Characteristics, Content, and Trends of Population and Provider Surveys

With respect to the US population, Table 1 describes some salient characteristics of the 8 population surveys that systematically and regularly collect data regarding ≥1 of the behaviors outlined in the conceptual framework in Figure 1. Table 2 shows the characteristics of 2 national provider surveys and associated questionnaire supplements.

Vaccine utilization

Of the 8 surveys described in Table 1, all except BRFSS, HINTS, and YRBS plan to routinely assess vaccine coverage from 2007 onward. Data are expected to be publicly available in late 2008. BRFSS included vaccine coverage items as part of its optional module; states decide individually whether to provide funding to include the module. HINTS, along with the NHIS, NIS, and National Survey of Children's Health (NSCH; available at:, are measuring HPV vaccine uptake and patient-reported provider recommendation. NIS-Teen will collect data from states, 6 urban areas, and other areas selected by states from 2008 onward. HINTS, NHIS, and NIS are also measuring general HPV knowledge. With respect to small area estimates, in 2007 the CHIS added HPV questions similar to those in the NHIS. To our knowledge, CHIS is the only state-wide effort collecting data on HPV-related behaviors. Because CHIS has been culturally adapted for and translated into Spanish, Mandarin, Cantonese, Korean, and Vietnamese, CHIS data will complement national estimates and provide key information regarding California subgroups that suffer from excess cervical cancer mortality.

Sexual behavior

Three surveys in Table 1 (NHANES, NSFG, and YRBS) can monitor trends in sexual behavior. NHANES began collecting detailed data regarding adolescent and young adult sexual behavior in 2003; before that, data were limited and not comparable to the adult NHANES population and therefore were underused. In the 2003 through 2004 NHANES cycle, self-collected vaginal swab specimens were collected to complement self-reports.53 NSFG and YRBS have been tracking the sexual behavior of men and women since 1988 and 1991, respectively.

Both NSFG and YRBS reported decreases in sexual activity and risk of STIs; however, the year in which the decline began is difficult to determine because of differences in the data collection years, sampling plans (household vs school-based), and operational definitions. The only standard measure across surveys is “ever had sex.” Other potentially comparable measures, such as age at initiation, current sexual activity, number of lifetime partners, and condom use, require truncation on the NSFG because of item wording on the YRBS (eg, re-code “age at initiation” to “before age 13 or not”).

A review of sexual behavior trends in the 1988 and 1995 NSFG data demonstrated relatively little change.75 However, NSFG data from 1995 to 2002 show a decrease in sexual activity, defined as ever had sex, especially among males ages 15 years to 19 years (55.2% to 45.7%) and females ages 15 years to 17 years (38.0% to 30.3%).76 The pattern for older females (ages 18-19 years) remained relatively stable except for Hispanic females, who demonstrated a decline in sexual activity across ages 15 through 19 years (52.7% to 37.4%). From 1991 to 2005, YRBS estimates demonstrated a decline in the percentage of students in Grades 9 to 12 who ever had sexual intercourse (54.1% to 46.8%) or were currently sexually active (37.5% to 33.9%).38 With regard to the lifetime number of partners, 2002 NSFG data indicated that 28% of females and 31% of males had had ≥2 partners, and YRBS data demonstrated no change in the percentage of students reporting sexual intercourse with ≥4 people during their lifetime (14.2% in 2001 and 14.3% in 2005).

Differences in sampling, data collection methods, and question wording make comparison of sexual behavior estimates across national datasets difficult. For instance, a key sampling bias in the YRBS is a lack of data regarding nonstudents (eg, those who are out of school because of dropping out or expulsion). Some information concerning this special population may be captured through the NSFG and special monitoring efforts such as the 1998 Alternative High School YRBS77; however, neither survey provides representative estimates for nonstudents. Research suggests that the ACASI program used by the NSFG and NHANES is the best method for encouraging accurate responses to sensitive questions such as sexual behavior, although there may be some sampling biases.24 ACASI provides an added measure of privacy because respondents use headphones to listen to digitally recorded questions and then key responses directly into the computer. Measures and data collection methods need to be standardized between NHANES, NFSG, and YRBS to ensure comparability.

Cervical cancer screening

Four surveys (BRFSS, HINTS, NHIS, and NSFG) routinely collect data regarding Pap testing; however, the operational definition for recent Pap use on the NSFG differs from the others, thereby precluding comparisons across surveys. In addition, HINTS has only 1 year of comparable data. BRFSS has included Pap items as part of its core since 1992, as has NHIS periodically since 1987 through its cancer module. Both the NHIS and BRFSS measure recent Pap use through open-ended questions with precoded responses to streamline data collection and maximize flexibility. Other surveys should use this question format to enable surveillance systems to modify their outcome measure if guidelines change (eg, shift from annual to 3-year screening interval for Pap tests in 2002).

Figure 2 shows national estimates of the weighted percentage of women aged ≥18 years who reported having undergone a Pap test within the past 3 years. Data are from the BRFSS and NHIS, using only those years in which women who had undergone hysterectomies could be identified and excluded. Trends from both surveys were similar and demonstrated little change over time, with approximately 84% of women in the BRFSS reporting a Pap test over a 14-year period (1992-2006) and 82% of women in the NHIS reporting a Pap test over a 12-year period (1993-2005).

Figure 2.

Percent and 95% confidence intervals (as indicated by vertical bars at the data points) of nonhysterectomized US women aged ≥18 years who underwent Papanicolaou tests within the previous 3 years; Behavioral Risk Factor Surveillance System (BRFSS) and National Health Interview Survey (NHIS) data. BRFSS data from each state are weighted to the total adult population of the state and added across all states to obtain the total percentages shown herein. BRFSS estimates exclude data from Arkansas in 1992; Wyoming in 1992 through 1993; Rhode Island in 1994; and Washington, District of Columbia in 1995. NHIS data are weighted to the adult civilian noninstitutionalized population of the US.

None of the national surveys measure HPV test use, but HINTS asks whether a woman has been told by her provider that she had an HPV infection. An affirmative response would suggest that she has been tested. Qualitative studies that investigate how HPV test results are communicated to patients would help inform the best way to assess this behavior in population surveys.78

Clinical practices of providers

In recent years, the Centers for Disease Control and Prevention (CDC) and National Cancer Institute (NCI) have conducted several surveys of primary care providers and specialists containing items relevant to HPV behavioral surveillance. Surveys have assessed provider awareness, preferences, and practices in relation to emerging technologies, including HPV tests, vaccines, and liquid-based cytology. The surveys also covered screening guidelines on when to start and stop, what population, how often, and when and how to use HPV tests.73, 74, 79 Other topics include vaccination, integration of vaccination and screening, cervical disease management and treatment, and reimbursement as well as systems factors.

The National Ambulatory Medical Care Survey (NAMCS; available at: and the National Hospital Ambulatory Medical Survey (NHAMCS; available at: are annual panel surveys of physician outpatient offices and hospital outpatient departments, respectively. Patient visits are the unit of analysis, but NAMCS samples providers and NHAMCS samples hospitals (Table 2). These surveys have been used to monitor Pap test use since 1973 and 1992, respectively. From 2006 through 2010, questionnaire supplements are being administered annually by mail and Internet to a sample of providers and/or hospital clinics that identified themselves as Pap providers on the initial intake forms. Thus, both supplements use providers as the sampling unit and unit of analysis. The supplements will examine HPV and Pap test practices for both screening and management in addition to HPV vaccine practices. Self-reports from the supplements will be compared with data from the main surveys, which are obtained by personal interviews and chart abstraction. Comparisons can be made regarding HPV vaccine adoption, type of Pap test, and HPV test use, thus allowing for some cross-validation.

In addition to the 2 panel surveys and their supplements, 3 1-time mixed-mode surveys have collected provider self-reports through mail and Internet questionnaires. These surveys, which may be repeated in the future as needed, include: 1) the HPV Provider Survey (available at:, which to our knowledge is the only survey to include physician assistants, nurse-practitioners, and nurse-midwives in addition to physicians; 2) the National Survey of Primary Care Physicians Recommendations and Practices for Breast, Cervical, Colorectal, and Lung Cancer Screening (available at: healthservices.cancer. gov/surveys/screening_rp), which uniquely examines some of the system-level factors that may influence screening and triage (eg, the use of patient reminders, guideline implementation, and use of electronic medical records); and 3) the National Pediatrician Physician Survey, which to our knowledge is the only survey to assess HPV vaccination diffusion by pediatricians.

To date, results of HPV-related behaviors are available only from the HPV Provider Survey.73, 74, 79 Approximately 87% were aware of HPV tests, and of those who were aware, 62% reported using HPV tests at their principal practice site. Among those who used HPV tests, 50% reported using it for reasons not approved by the US Food and Drug Administration (FDA), including as a primary screening test for HPV infection; to test the infection status of patients diagnosed with anogenital warts or another STI; or to test the partners of persons diagnosed with abnormal Pap results, anogenital warts, or other STIs.79 Among clinicians who provided screening, 91% were aware of HPV tests, and 63% had used them for the management of borderline or abnormal cytology results as approved by the FDA. Only 21% used these tests as an adjunct to cytologic screening of women aged ≥30 years, another FDA-approved indication.73 Because adjunct testing was only recommended a year before the survey, this finding is not surprising. Finally, approximately 58% used HPV tests in response to patient demand, which may be consistent with FDA-approved indications if requested by women for either the management of abnormal results or adjunct screening.

The HPV Provider Survey allowed comparisons of screening and management practices between providers who reported participating (vs not participating) in the National Breast and Cervical Cancer Early Detection program.80 Program participants were more likely to see women who were on Medicaid or had abnormal Pap test results, and they were also more likely to provide colposcopy on site; however, overall, the practice patterns of program participants were very similar to those of nonparticipants.

Data for the other 2 provider surveys are forthcoming and the value of routinely adding them to the national HPV and cervical cancer surveillance system has yet to be determined. Pending results will determine whether they are fielded again.

Limitations of Population and Provider Surveys

Current population surveys do not provide us with data from institutionalized populations such as incarcerated females, who are at higher risk for HPV infection and cervical cancer.81 In addition, research on adolescents and young adults may prove challenging because of their geographic and healthcare system mobility. Minimal data are collected on the population's understanding of persistence and the role it plays in the progression from HPV infection to precancerous lesions. At the current time, it is unclear what key pieces of information are needed by women to make an informed decision regarding the HPV vaccine and test.

For provider surveys that collect data from different specialties, the effect of various response rates by specialty is unclear. Furthermore, to our knowledge, little is known concerning the tools that providers use to help them explain guidelines for the HPV vaccine and test to patients.

Future Surveillance Research in HPV and Cervical Cancer

Future surveillance research on HPV and cervical cancer should focus on 3 major goals: 1) monitoring patterns of vaccine and screening use by providers and the population; 2) expanding surveillance when new technologies are approved; and 3) analyzing trends to evaluate the impact of guideline changes and public health interventions on disparities.

The first goal is to continue to routinely assess the national population's vaccine, sexual, and screening behaviors and the patterns of care among providers in relevant specialties. These studies are needed to identify population subgroups at high risk of cervical cancer as well as to assess quality of care and barriers to access. The array of new technologies available is likely to increase disparities at least initially if they follow the same diffusion trend observed for other cancer screening modalities.82 Screening with Pap and HPV tests will remain a critical outcome to monitor because current vaccines do not protect against all oncogenic types of HPV and are not approved for all age groups. Of key interest will be whether HPV vaccine uptake adversely affects sexual behavior and screening use, and whether direct-to-consumer marketing and growing public awareness of the link between HPV and cervical cancer results in increased patient demand for HPV tests and/or vaccination. Also of interest is the effectiveness of ‘catch-up’ programs that aim to immunize females ages 13 to 26 years. As time passes, surveys will have to measure the age that the HPV vaccine was received to differentiate recommended versus catch-up vaccinations. In addition, it will be important to monitor whether providers are following current guidelines that indicate HPV tests as adjuncts to Pap tests for primary screening only among women aged ≥30 years. Such studies will also identify off-label uses and populations who are overscreening. These populations may benefit from interventions raising awareness of the risks of screening.

A second goal is to ensure that population and provider surveillance systems are adapted to track diffusion of emerging technologies, such as next-generation vaccines that protect against oncogenic types other than HPV‒16 and HPV‒18 and self-sampling for Pap tests. Surveillance systems must also have the flexibility to rapidly incorporate modifications to screening and vaccination guidelines (eg, extensions in screening intervals or expansion of age recommendations), especially if ongoing vaccine trials demonstrate efficacy in mid-adult women. As stated earlier, surveys that measure behaviors through open-ended questions with precoded responses are ideal.

A third goal is to conduct time trend studies to evaluate the impact of changes in policy and guidelines on population and provider knowledge, attitudes, and behaviors. A recent study on the likelihood of parental acceptance of the HPV vaccine identified lower acceptance rates among African Americans and Asian Americans.83 Research evaluating whether this translates into lower uptake in some subgroups is needed because it might exacerbate current disparities. Tracking whether new technologies are being used by underserved, low socioeconomic, minority, and immigrant populations who already experience disparities is an important priority. Patient-provider communication regarding cervical cancer prevention and control is likely to become even more complicated with the release of second-generation vaccines and changes in screening guidelines and must be monitored. To evaluate the impact on downstream outcomes, data regarding precancerous lesions will need to be collected at the national level.84


Ongoing surveillance of HPV-related and cervical cancer-related knowledge, attitudes, and behaviors at the national, state, and local levels are critical. Multiple surveillance systems are needed to capture the wide range of behaviors and target populations as well as to balance the strengths and weaknesses of different survey methodologies and content. Analyses and reports need to determine whether the new technologies in combination with Pap tests are reducing disparities in morbidity and mortality, especially in high-risk populations. Because excess risk and deaths are concentrated in particular geographic, socioeconomic, and racial and ethnic populations, regional and local cancer control researchers who match their questionnaire content with national surveys can enhance local efforts to evaluate progress in reducing disparities. Given the diversity of the US population, surveillance efforts focused on local communities are needed. Local area survey data that are geocoded or linked to cancer registries would greatly contribute to our ability to evaluate cervical prevention and control activities.