The best magazine
Methods
Data
The present study used data from the Health Information National Trends Survey 2007 (HINTS 2007) (Health Communication and Informatics Research Branch, Division of Cancer Control and Population Sciences, National Cancer Institute), a study of a civilian, non-institutionalized population in the US aged 18 years and older. The HINTS 2007 was conducted between December 2007 and April 2008. The 2007 survey started just over a year after the quadrivalent HPV-6/11/16/18 vaccine (Gardasil/Silgard®) was licensed in the US (June 2006); the bivalent HPV-16/18 vaccine (Cervarix®) was not FDA approved until October 2009. The sample population was surveyed using either of these two instruments: a computer-assisted telephone interview or a mailed questionnaire, with response rates of 24.23% and 30.99%, respectively. The HINTS focused on the public's access to and use of cancer-related information, including a set of questions that was specifically related to cervical cancer and the HPV. In the present study, the sample was restricted to women who reported having any female children under the age of 18 in the household (N = 880). This particular group was chosen because the Advisory Committee on Immunization Practices (ACIP) recommends routine HPV vaccination for 11- and 12-year-olds with catch-up vaccination up to 26 years old. For females below 18 years old, parental consent was needed to authorize the HPV vaccination. Thus, parental awareness and acceptability of the HPV vaccine is a crucial factor in the utilization of HPV vaccination among these female adolescents. After excluding observations with missing values for the study variables, the final sample consisted of 742 women.
Variables
Acceptability of HPV vaccination was defined using the survey question, "A vaccine to prevent the human papillomavirus or HPV infection is recommended for girls aged 11–12 and is called the cervical cancer vaccine, HPV shot, or GARDASIL®. If you had a daughter that age, would you have her get it?" Respondents were given the following options to choose from: "Yes", "No", and "Not sure/It depends". For the statistical analysis, a binary indicator variable was created using "Yes" (= 1) versus "No" and "Not sure/It depends" (= 0), as defined by Fang and colleagues. A dichotomous variable for HPV vaccine awareness was defined using the survey question, "A vaccine to prevent HPV infection is available and is called the cervical cancer vaccine or HPV shot. Before today, have you ever heard of the cervical cancer vaccine or HPV shot?" In the HINTS survey, this question on HPV vaccine awareness (No. 61) preceded the question on acceptability (No. 74).
The regression analysis included the following covariates: age, race, household income, and education level; having any insurance, any regular health care provider, past medical history of cancer, and family history of cancer; and self-rated health status, and census division. The specific categories used are shown in Table 1 and four-category dummy variables were created for combined annual household income. Considering the observed frequencies and the categories used by Fang et al. (2010), income categories from the original questionnaire were combined as follows: "$20,000 -< $35,000" and "$35,000 -< $50,000" in one category, and "$75,000 -< $100,000" and "$100,000 or More" in another. A respondent was regarded as having a regular health care provider if she responded positively to the question, "Not including psychiatrists and other mental health professionals, is there a particular doctor, nurse, or other health professional that you see most often?" Similarly, having any insurance referred to an affirmative response to the question, "Do you have any kind of health care coverage, including health insurance, prepaid plans such as HMOs, or government plans such as Medicare?" Health status of respondents were based on their responses to the question, "In general, would you say your health is excellent/very good/good/fair/poor." There were nine census divisions, covering the New England Census Division throughout the Pacific Census Division.
Statistical Analysis
To address the issue of unmeasured differences between the aware and non-aware groups which may influence acceptability of HPV vaccination, the current study employed an instrumental variable (IV) bivariate probit model. This method has been applied in various public health and health economics research to address similar statistical issues. In this study, the IV bivariate probit model estimated the following joint model of acceptability of HPV vaccination (Eq. 1) and awareness of the HPV vaccine (Eq. 2).
where Accept* is a continuous latent variable for the indicator variable of HPV vaccine acceptability (Accept = 1 if Accept* > 0, otherwise 0) and Aware* is a continuous latent variable for the indicator variable of whether the respondent has heard of HPV vaccine (Aware = 1 if Aware* > 0, otherwise 0). X denotes a set of covariates common to Eq. 1 and Eq. 2, and Z denotes a set of IVs included only in Eq. 2. β, γ1, γ2, and δ are coefficients for corresponding variable(s) in the models, whereas ε1 and ε2 are error terms of Eq. 1 and Eq. 2, respectively.
It is possible that unexplained differences in HPV vaccine awareness in Eq. 2 (ε2) are not correlated with unmeasured factors in Eq. 1 (ε1) that influence HPV vaccine acceptability. In this case, estimating Eq. 1 alone would already provide a valid estimate of β, which will capture the extent to which being aware of the HPV vaccine leads to HPV vaccine acceptability. If ε1 and ε2 are correlated, however, consistent estimates can be obtained by jointly estimating the two equations in the bivariate probit model. A formal test can be conducted to examine the correlation coefficient (ρ) between ε1, and ε2. Estimating the IV bivariate probit model efficiently requires good IVs (Z in Eq. 2) that must satisfy two basic conditions, these IVs: (1) should predict the outcome of Aware and (2) should not directly affect the outcome of Accept. The HINTS 2007 contains such potentially good IVs because access to information related to the various types of cancers is available. For example, a respondent who gave a positive answer to the question, "Have you heard of any tests to find lung cancer before the cancer creates noticeable problems?", would be more likely to have also heard about the HPV vaccine. However, responses to this question would be unlikely to directly influence the extent to which the respondent has greater acceptability of HPV vaccination. In the same vein, a positive response to the question "Have you ever heard of a clinical trial?" would strongly predict a greater probability of HPV vaccine awareness but unlikely affect the level of HPV vaccine acceptability. The current study used these two IVs, after exploring other variables related to the access to different types of information on cancer. Tests for these underlying assumptions suggested that these IVs predict Aware statistically significantly but are not directly associated with Accept other than through Aware. Stata 11.0 (StataCorp LP, College Station, TX) was used in all statistical analyses, taking into account a complex survey design.
Source: ...