G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
Health Policy xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Health Policy journal homepage: www.elsevier.com/locate/healthpol
Female preventive practices: Breast and smear tests Ana Isabel Gil Lacruz a , Marta Gil Lacruz b,∗ , Sophie Gorgemans a a Department of Organization and Management Business, Industrial Engineering School, Edificio Agustín de Betancourt. C/María de Luna s/n, Zaragoza 50018, Spain b Psychology and Sociology Department, Health Sciences Faculty, C. Domingo Miral s/n, Zaragoza 50009, Spain
a r t i c l e
i n f o
Article history: Received 3 September 2013 Received in revised form 14 April 2014 Accepted 21 April 2014
Keywords: Gender and health Cancer and palliative care Primary care Statistical methods Risk and health
a b s t r a c t Breast cancer and cervical cancer are the most common female cancers in Spain and in many developed countries. The main goal of this paper is to identify the determinants of individual decisions on breast screening and smear testing, that is to say, the decision to take a test for the first time and the decision to test with suitable regularity. To that end, we have combined analyses of micro and macro data (the Spanish National Health Survey and Spanish Regional Social Indicators) and employed multilevel estimation models. Among the main results, we highlight the fact that regional public screening programmes improve individual decisions on screening (more women testing for the first time and more women testing regularly) and, furthermore, they generate positive synergies; for example, regional public programmes for smear testing improve individual decisions on both cervical and breast cancer screening. In addition, we conclude that it is not only important to know if the numbers of women undergoing breast screening and smear testing are increasing, it is also important to know if they are testing regularly. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Breast cancer and cervical cancer are the most common types of female cancer in Europe and North America [1,2]. In Spain, as in most developed countries, the annual incidence of breast cancer is 51 new cases per 100,000, whilst for cervical cancer the figure is 8 new cases per 100,000. Breast cancer most frequently occurs in women between the ages of 35 and 80 but the peak onset period is between 45 and 65 [3]. The average age of cervical cancer diagnosis is 48 years old although approximately 47% of women with invasive cervical carcinoma are diagnosed before the age of 35 [4].
∗ Corresponding author. Tel.: +34 976 76 44 43; fax: +34 976 76 20 03. E-mail addresses: [email protected] (A.I. Gil Lacruz), [email protected] (M. Gil Lacruz), [email protected] (S. Gorgemans).
Since the introduction of smear tests, deaths caused by carcinoma of the cervix have fallen by up to 99% in populations in which women are regularly screened. A routine smear testing programme with appropriate follow-up can reduce the incidence of cervical cancer by up to 80% [5]. The mortality rate for breast cancer has fallen by almost 30% and two-thirds of the decrease has been attributed to screening [6,7]. Regular screening from the age of 50 is not only cost-effective [8], but it saves about 2 lives over 15 years for every 1000 women screened [7]. Screening can lead to longer survival rates but it can also result in over-diagnosis, or have no effect at all. In fact, the use of mammography as a screening tool for the detection of early breast cancer in healthy women without symptoms continues to be debated. The figures have to be seen in the light of errors in diagnosis, overtreatment, and radiation exposure. 1 woman in 2000 will have her life prolonged by 10 years through screening but 10 healthy
http://dx.doi.org/10.1016/j.healthpol.2014.04.012 0168-8510/© 2014 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
2
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
women will undergo unnecessary breast cancer treatment and 200 women will suffer from significant psychological stress due to erroneous diagnosis [9]. Despite these negative aspects, health organisations continue to recommend regular breast screening – every 2 years for women between the ages of 50 and 74 [10] or every 2–3 years between 50 and 69 [11]. In Spain, cancer screening is in accordance with national and international recommendations. Despite the fact that screening is free, or the charge is symbolic, there are important socio-economic inequalities in the take-up of breast screening. There are a number of possible explanations for this: (i) the poor and less educated might feel less confident and be less inclined to request specialist consultation (instead of, for example, asking the family doctor); (ii) they are also more likely to suffer from fear and anxiety which are negatively correlated with screening; and (iii) the poor and less educated are more likely to turn to preventive care later than the economically advantaged, for example, they wait until after the appearance of health problems [12]. Health expenditure and finance has become a key issue in these times of economic crisis. Demographic and socioeconomic changes are affecting the financial sustainability of the welfare state and an economic perspective on health issues is essential. Mammographic screening programmes for women aged 50–70 have been shown to be effective in reducing mortality rates at reasonable cost [13] but they are not necessarily cost-effective in every situation and the inclusion of other age groups is far from accepted, particularly in terms of cost and the impact on quality of life. Economic evaluations of interventions for malignant neoplasms are not common despite their gradual increase in recent years in Spain [14]. In the case of cervical cancer, the Spanish government recommends human papilloma virus vaccination for teenage girls, despite its high cost and persistent doubts about its effectiveness and the need to continue regular cytology screening [13]. The aim of this paper is to study the determinants of breast and cervical cancer screening among Spanish women. The Spanish National Health Survey (2006/07 and 2011/12) [15] has been used to provide information on sociodemographic characteristics, risk behaviours, access to health services and the state of health of Spanish women. Special attention is paid to questions related to preventive practices such as screening controls and frequency. The research should make empirical evidence on vulnerable population groups and the efficiency of public health policies available to policy makers. The main research contribution of this paper is the use of both micro data and regional health public policies in the analyses. We have examined data at different aggregation levels (individuals and regions – the Spanish Autonomous Communities) in order to determine if geographical differences are consequences of population characteristics (e.g. gender, age or working status), contextual data (e.g. regional public health policies) or simply unobserved contextual data. We have also repeated estimations by age cohorts in order to understand if different population groups require specific policies. Not all population groups are targeted by public screening programmes, so we are able to check if women of different ages are particularly sensitive to specific factors.
2. Review of the published literature Breast and cervical cancer incidence rates in Spain are slowly increasing; this is probably due to the country’s ageing population and more frequent early diagnosis [4]. An ageing population is not unique to Spain, it affects most developed countries. The positive relationship between age and screening participation with breast and cervical cancer rates has been known for decades [16]. Screening is a key tool to prevent cancer deaths. For example, screening attendance of 70% might reduce breast cancer mortality by about 25% among women aged 50–69 [11]. In Europe, the availability of screening programmes explains about 40% of cross-country differences in screening rates and factors like age, education, health status, etc. are associated with screening uptake within countries [2]. Screening is not the only preventive action that can be taken; healthy habits also play an important role. Smoking is a leading factor in cervical cancer [16] and a high-fat diet [17] and excessive alcohol intake [18] are risk factors for breast cancer. Smoking is a negative health behaviour that serves as a proxy for poor health habits such as lack of exercise, bad diet and alcohol abuse [19]. Age, smoking and health care access are inversely related to screening, whilst education, income, insurance and perceived risk of cancer are directly related [19–21]. Women who are strongly influenced by the advice of their General Practitioners tend to take regular tests, women who do not receive this advice screen less frequently, or not at all [21]. In the USA, health insurance reduces the cost of pointof-service care and has a positive effect on the demand for screening [19]. In the UK, there is empirical evidence that income does not influence uptake as screening is free of charge [22]. The literature on Spain concludes that invitation to screening, visits to the gynaecologist and women’s attitudes are the main reasons for testing [23]. More effort is required for women over 65 with regards to breast cancer testing and women over 55 for cytology. Special attention needs to be paid to women in lower socioeconomic level groups [23–26]. Additional medical insurance coverage increases the probability of regular screening and this is especially true for mammography [24]. Unhealthy lifestyles have been associated with non-adherence [25]. The implications of lifestyles on health are also validated by international comparisons. The incidence of both types of cancer has increased in Spain in recent decades but they are still lower than rates in North America. Cervical and breast cancer are related to lifestyles as well as cultural and environmental factors [27]. Individual and contextual covariates (such as the number of health care centres or population characteristics) are associated with the uptake of breast and cervical cancer screening [28]. 3. The data: public health policies, individual observations and regional indicators The Spanish Constitution of 1978 gave its Autonomous Communities (also known as regions) the legal authority
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
to regulate health care. A National general strategy against cancer was defined in 2003 (finally adopted in 2006) with specific objectives in early detection [29,30]. All the regions have specific breast cancer screening programmes that include women between 50 and 69 years old, but in some regions the age range can be as low as 45. In general, the frequency of testing is every two years. For breast cancer screening, each autonomous community uses an information system based on an individual health card (in some regions this is combined with the data from the population census) and sends a written invitation to take part in the programme. The first population programme for early detection breast cancer began in Navarra in 1990. Navarra was also the first community that achieved 100% coverage of women in the target population in 1999. New programmes were gradually established and all regions achieved 100% coverage in 2009. Screening programmes for the detection of cervical cancer are less well established but since 1993, the list of public health services offered has included a Pap cytology test for women between the ages of 35 and 64. The frequency of testing is usually between 3 and 5 years [31]. All regions currently have cervical cancer screening programmes, though these are mostly of the opportunistic type, except in La Rioja and Castilla-Leon where the programme is population based. Opportunistic selection methods are based on access to primary health care and specialists [29]. 100% coverage is easier to achieve with selection methods based on population; opportunistic selection depends on women attending primary health care centres or specialists. In the NHS Cancer Strategy (2009) [32] the recommendation was to reach 80% of women between 40 and 50 years old with a smear test every 5 years [29,30]. This information has been used to study health policy variables in the regression analysis. The micro analysis is based on women in the target population groups for breast screening and smear tests. The macro analysis considers the length of time that public health screening programmes have been operating in the Spanish regions (Table 1). Individual data and regional information on social indicators and health issues have been used for the empirical study (Table 2). Individual observations have been taken from the Spanish National Health Surveys of 2006/07 and 2011/12 [15]. The National Health Survey (NHS) has been carried out by the Spanish Ministry of Public Health since 1987. The reason why waves before 2006 were not considered is because important questions for our research goals, such as response categories for individual screening frequency, are not comparable through time. The survey is cross-sectional and uses a stratified multistage sample. The first stage units are census sections. The second stage units are the main family residences. One adult (aged 16 years or over) is selected from each household to complete the Adult Questionnaire; if there any minors in the house (under 16 years of age), one is selected to complete the Minor Questionnaire. The method for collecting information is personal interview, in exceptional cases this can be done by telephone. At a national level, the survey sample is representative to 95%. A figure of over 90% is generally considered as sufficient to
3
represent target populations. The national response rate of 95% was achieved by an effective household response rate of 65% and a substitution rate of 30%. In all the autonomous communities the effective response rates are over 90% [15]. The survey provides information on: sociodemographic characteristics (age, gender, civil status, educational level and employment); lifestyles (smoking, drinking, physical exercise, Body Mass Index and nutrition); access to health services (private insurance, visits to the family doctor and prescriptions); and state of health (general health problems, stress, unhappiness, menopausal problems). 25,782 observations were made regarding women over 16 years old and living in Spain. To construct the dependent variables for breast cancer screening (BreastTest and BreastTestRight), we have used the following questions: “Have you ever had a mammography?” (1: Yes; 6: No; 8: Don’t know; 9: No answer) and “If you have had a mammography, please indicate the time period since your last screening” (1: In the last 12 months; 2: More than 1 year but less than two years; 3: More than two years but less than 3 years; 4: Over 3 years ago; 8: Don’t know; 9: No answer). To construct the dependent variables for cervical cancer screening frequency (SmearTest and SmearTestRight), the questions were: “Have you ever had a cervical smear test?” (1: Yes; 6: No; 8: Don’t know; 9: No answer) and “If you have had a cervical smear test, please indicate the time period since your last screening” (1: 3 years ago or less; 2: More than 3 years but no more than 5 years; 3: Over 5 years ago; 8: Don’t know; 9: No answer). The dependent variables BreastTest and SmearTest are categorical variables calculated for all the interviewed women. The variable BreastTest takes the value 1 if the woman has breast screened at least once, 0 if not. In the same way, SmearTest takes the value 1 if the woman has had a smear test at least once, 0 if not. The dependent variables BreastTestRight and SmearTestRight are categorical variables calculated for all the interviewed women that have breast screened – BreastTestRight – and had a smear test – SmearTestRight – at least once. The variable BreastTestRight takes the value 1 if the woman has breast screened every 2 years or more, 0 if not. SmearTestRight takes the value 1 if the woman has had a smear test every 5 years or more often. It is very important to know if women have ever been screened for breast cancer and cervical cancer as this provides information on whether women access and use screening services; moreover, subsequent appointments are automatically generated after the first screening. This is particularly relevant for women who are younger than the target population group for the breast test. For example, there might be women younger than 45 that want breast screening because they have a lump in the breast or because they have a family history of breast cancer. In the case of the smear test, most regions have opportunistic programmes so screening history begins when the family doctor makes the first appointment. In general, it appears that women need to be more proactive in demanding the first smear test than the first breast test. The data shows that around 56% of Spanish women have breast screened and 67% have smear tested at least once.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
4
Table 1 Breast and cervical screening programme in Spanish regions. Mammography
Papanicolau cytology Selection method: PB
Year 100% coveragea
Year
Population
Periodicity
Selection method
50–69 50–69a
CP CP & IHC
2005 2006
1986 1993
20–60b 35–64
OP OP
1991
50–69
CP & IHC
2000
1993
25–65
Balearic Islands Canary Islands
1997 1999
50–69 50–69
CP & IHC CP & IHC
2009 2005
1993 1995
25–64 18–65
Cantabria
1997
50–69
CP
1997
1997
25–65
Castilla Mancha Castilla León
1992 1992
45–69 45–69
CP & IHC CP & IHC
1997 1996
1993 1986
25–60 30–65
Catalonia Extremadura
1992 1998
50–69 50–69
CP & IHC CP & IHC
2004 2005
1993 1983
25–65 20–65b
Galicia Madrid
1992 1999
50–69 50–69
CP & IHC IHC
1998 2001
1993 1993
20–65 35–65
Murcia
1994
50–69
IHC
1999
1993
35–64
Navarra The Basque Country
1990 1995
45–69 50–69
CP CP
1992 2000
2000 1999
25–65 25–59
La Rioja Valencia
1993 1992
45–69 45–69
CP & IHC CP & IHC
1995 2001
1993 1995
25–65 20–65
3 years 3 years (annual first 2 years) 3 years (annual first 2 years) 3 years 3 years (annual first 2 years) 5 years (annual first 2 years) 3 years 3 years (annual first 2 years) 3 years 3 years (annual first 2 years) 3 years No periodicity defined 5 years (annual first 2 years) From 3 to 5 years 3–5 years (annual first 2 years) 3 years 3 years
Year
Population
Andalucía Aragon
1995 1997
Asturias
Periodicity
2 years
OP OP OP OP OP PB OP OP OP OP OP OP OP PB OP
PB, population based; CP, census population; IHC, individual health card; OP, opportunistic. Table: authors’ work [29–32]. a To 64 years old before 2007. b From 15 years old before 2007.
80% of women who have screened and tested at least once have done so with the appropriate frequency. An analysis of regional information on public health screening programmes reveals that 29% of the women interviewed are in the target population for breast screening programmes and 58% of them are in the target population for cervical screening programmes (TargetBreastTest and TargetSmearTest). In addition, in the year of the interview, 90% had lived in communities where the coverage of breast screening is 100% of the target population while 7% lived in areas where smear testing is based on a population method (TotalCoverageBreastTest and TotalCoverageSmearTest). Other relevant statistics were: 11% of Spanish women have private health insurance; 87% of the population visited the family doctor at least once in the 4 weeks prior to the interview; 44% of Spanish women are overweight, 20% are daily smokers and 43% do no regular physical activity; and 72% eat fresh fruit and vegetables every day. 4. The empirical framework This research focuses on two questions: (a) Do women screen for breast cancer and smear test for cervical cancer? (b) If the answer is ‘yes’, are they screening and testing with appropriate frequency? We considered a set of sociodemographic variables (age, gender, marital status, working conditions, etc.) that may help to explain individual decisions. The main contribution of this work is the utilisation
of policy variables, such as screening programmes, on screening uptake. As empirical strategy, we used multilevel models (STATA: xtmelogit). Multilevel regression models are indicated when there is a hierarchical structure in the levels of data, with a single dependent variable measured at the lowest level and a set of explanatory variables on the other levels. The advantage of these models is their capacity to define and explore variations at each level of the hierarchy after controlling for relevant explanatory variables [33]. In this case, the multilevel analysis helps us to understand individual and regional determinants of screening decisions and to determine if regional differences remain after controlling for individual variables and regional policies. The data is structured by j Spanish Autonomous Communities, in each of which, nj persons have been interviewed. The dependent variables, BreastTestij and SmearTestij , consider if the individual i of region j has tested for breast and cervical cancer at least once. Therefore, the demand functions1 for breast screening are characterised as: BreastTestij = ˇj + +εij
(1)
1 Analysis of the robustness of the estimated parameters involved repeated estimations including a different set of explanatory variables. Given that the coefficient parameters remain constant, these results have not been included (for reasons of space); they are available on request.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
5
Table 2 Spanish National Health Survey (2006 and 2011). Sample = 25,782 women from 16 years and older. Variables
Description of dummy variables (1: yes/0: no)
Mean/SD
BreastTest: has undergone breast screening BreastTestRight: screening for breast cancer every 2 years or more often (only women who have screened at least once) SmearTest: has taken a smear test SmearTestRight: smear test every 5 years or more often (only women who have smear tested at least once)
0.56 (0.40) 0.80 (0.39)
Age 16–29 Age 30–49 Age 50–69 Age 70–older
0.09 (0.29) 0.35 (0.47) 0.29 (0.45) 0.27 (0.26)
Marital status
Married Single Widow Divorced/separated
0.48 (0.50) 0.24 (0.48) 0.20 (0.40) 0.38 (0.49)
Educational level
Women with primary or no education Secondary education Tertiary education
0.47 (0.40) 0.38 (0.48) 0.15 (0.35)
Employment status
Employed Retired Unemployed Homemaker Other
0.35 (0.47) 0.27 (0.43) 0.06 (0.25) 0.28 (0.45) 0.04 (0.20)
Lifestyles
Overweight: Body Mass Index > 25 Smoker: daily smoker Alcohol: drinks alcohol every day Sedentarism: no physical exercise Daily fruit: consumes fresh fruit and vegetable every day Eco-goods: consumes ecological products
0.44 (0.49) 0.20 (0.40) 0.15 (0.36) 0.43 (0.49) 0.72 (0.44) 0.05 (0.23)
Health good and services access
Private insurance: holds a private health insurance policy Family doctor: visit to family doctor in last 4 weeks Hormone products: uses hormone products Pill: uses contraceptive pills
0.11 (0.32) 0.87 (0.33) 0.01 (0.11) 0.04 (0.19)
Health state
Health problems: had health problem Stress: feels stressed Unhappiness: feels unhappy Menopause problem: has problems with menopause
0.40 (0.49) 0.29 (0.46) 0.22 (0.41) 0.09 (0.28)
TargetBreastTest: target population for regional public breast screening TargetSmearTest: target population for regional public smear test TotalCoverageBreastTest: living in a region with 100% public coverage for breast screening for the target population PopulationBasedSmearTest: living in a region where the selection for smear test is population based
0.29 (0.45) 0.59 (0.49) 0.90 (0.29)
Screening
Age
a
b
Health policies
a b
0.67 (0.47) 0.80 (0.40)
0.07 (0.26)
All variables are dummy variables, except those related to the age of the individual which are numerical. Authors’ work, based on the data in Table 1. Contextual dummy variables (Year 2006/07 and Year 2011/12; 17 Spanish Autonomous Communities).
where ˇj represents the mean of the dependent variable of all the individuals of region j and εij represents the deviation of the value of this variable for the individual i from the mean of that variable in the other interviewees who live in the same j. ¯ Eq. (1) can be expressed Deviation of ˇj from its mean ˇ, as: ¯ + uj + εij where uj = ˇj − ˇ ¯ BreastTestij = ˇ
(2)
The model formulated in this equation coincides with the components of variances with fixed and random effects. ¯ represents the fixed effects. This model The parameter ˇ assumes that the random effects uj are distributed normally with mean 0 and variance u2 = ˇ2 , which means that differences in the variable BreastTest are attributable to the
country. It also assumes that the error component εij is distributed normally with mean 0 and variance 2 . Finally, it assumes that the random effects uj and the error component εij are independent and that the εij are all independent from each other. The second step introduces information about K−1 individual characteristics (socioeconomic factors: lifestyles, health care and state of health). These characteristics can be included in a vector Xij that when introduced into the model would involve the appearance of K new fixed effects (K−1 variables and a constant). Eq. (2) would then become: BreastTestij = Xij ˇj + uj + εij
(3)
in which X includes K regressors and εij ≈ N(0, 2 ).
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
6
The third, and final step, parameterises the coefficients ˇj of Eq. (3) by adding regional explanatory variables (regional health policies): ¯ j + uj + εij BreastTestij = Xij ˇ
(4)
These models represent general specifications of multilevel models in which the responses to a dependent variable are related in a linear function to a set of explanatory variables and to a simple hierarchical structure. Because our dependent variables are binary, taking values of 1 if the woman has screened and 0 if not, we applied multilevel analysis with a logistic function. Fixed effects are now dependent on individual and regional variables and a constant. The same process was repeated for SmearTest. Agepopulation subsamples (16–29, 30–49, 50–69, and 70+) were also considered for both dependent variables. Different population groups were used because not all of them are targeted by public screening programmes, so it is possible to determine if women of different ages are especially sensitive to different individual factors. Screening for breast and cervical cancer at least once is better than not at all, but it might be relevant to look at individual decisions on screening and testing frequency. Therefore, for the subsample of women who had screened at least once, we searched for the determinants for screening for breast cancer every 2 years or more (BreastTestRight) and testing for cervical cancer every 5 years or more (SmearTestRight). Once again, the dependent variables are binary, taking values of 1 if the woman tests with adequate frequency and 0 if not. Finally, we report the results as odds ratios (OR). The odd ratio is a measure of the association between an exposure and an outcome that represents the chances that an outcome will occur given a particular exposure, compared to the chances of the outcome occurring in the absence of that exposure. Odds ratios are most commonly used in case–control studies but they are also used in crosssectional studies [34]. 5. Results The results are presented in the same order as the previous section: Women aged 50–69 are the most likely to have participated in breast screening programmes; younger women (aged 16–29) are the least likely. Odds ratios provide information about association effects in terms of sense and intensity. For example, looking at the total sample in model 1, women aged from 30 to 49 are 3.75 times more likely to have undertaken breast screening than women aged from 16 to 19. In general, demand for breast screening is higher among more mature women and this is especially true for women aged 30–49. Widows are less likely to have demanded breast screening, and they are tested with less regularity than married women (a result may well be affected by age – widows tend to be older than married women). Single women are less likely to have demanded breast screening than married women – a result that was robust for all age groups. Employment status and educational levels appear to have no significance. The only two statistically
significant results in this section were that among women aged 50–69, housewives are less likely to have demanded breast screening than women who work and, among women aged 30–49, having a university degree is positively correlated with testing for breast cancer. Similar results can be inferred for healthy behaviours (a healthy diet and consuming ecological products) and the contrary results apply to unhealthy behaviours (smoking and sedentarism). These results are particularly consistent for women aged 30–69. Private health insurance reinforces individual decisions on breast screening, but this result was only robust for women aged 30–49, and in this group, visits to the family doctor have the opposite effect. Consuming hormone products, self-reporting poor health, suffering stress and having menopausal problems are positively correlated with being screened. Being in the target population for public breast screening programmes is effective, in that it improves individual screening decisions (the women are more likely to screen) and there is a further, indirect effect, as being in the target population for smear testing also improves individual decisions on breast screening. Among older women, breast cancer screening levels increased between 2006/07 and 2011/12. The regional level variance u2 is significant, even after controlling for regional public screening programmes. Although the addition of regional public health policies does not reduce the unexplained variability among regions, the estimation for breast screening decisions by age cohorts shows that regional variance is not significant for the population groups with the highest and the lowest rates for breast screening (women aged 50–69 and women younger than 29). Women aged 50–69 are in the target population groups for all regions; women younger than 29 years old are not screened under the public health system unless there are health symptoms or a family history of breast cancer. Nevertheless the coefficients of the regional variance for both age groups are not statistically significant. On the other hand, in the 30–49 age group, there are women who are in the target population and women who are not. Furthermore, women that are in the target population in one region might not be in the target population in other regions (women from 45 to 49 years old). For older women, the date when the public screening programmes were implemented may be more relevant to the question of whether they have been considered as being in the target population groups at least once in their lives (Table 3). With regards to smear tests, older women (over 70) and younger women (under 36) are less likely to participate in screening programmes. Women between the ages of 30 and 49 are more likely to have demanded smear tests. For women between 16 and 49, age is positively correlated with the demand for smear tests; for women older than 50, age is negatively correlated. Widows and single women are less likely to have demanded a smear test than married women. Homemakers are less likely to have tested for cervical cancer than women in paid employment, but this result is only robust for women in the youngest age cohort. Educational level is positively correlated with testing for cervical cancer for all age groups except for women younger than 30 years old. Unhealthy behaviours
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
7
Table 3 Determinants of breast screening decisions: logistic multilevel regression results and odds ratios. Dependent variable: BreastTest. Model 1
Age Age: 16–29a Age: 30–49 Age: 50–69 Age: 70–older Marrieda Single Widowed Divorced Primary/No formal educationa Secondary education Tertiary education Employeda Retired Unemployed Homemaker Other Overweight Daily smoker Daily drinker Sedentary lifestyle Daily fruit Eco-goods Private health insurance Family doctor Hormone products Contraceptive pill Health problems Stress Unhappiness Menopause problems TargetBreastTest TargetSmearTest TotalCoverageBreastTest PopulationBasedSmearTest Year 2006/07 Year 2011/12a Random effect: u2 LR test (Prob > 2 ): a * ** ***
Model 2
Total sample
Age: 16–29
Age: 30–49
Age: 50–69
Age: 70–older
Total sample
– – 3.75*** 5.48*** 7.21*** – 0.53*** 0.67*** 1.20 – 1.00 1.16 – 0.98 0.84 1.12 0.45*** 1.17** 0.97 1.17 0.82*** 1.51*** 1.75*** 1.50*** 0.85 2.24** 0.62*** 1.40*** 1.04 1.13 2.21*** – – – – 0.85 – 0.17 0.00
1.14** – – – – – 1.76 (omitted) 2.52 – 0.94 1.28 – (omitted) 1.96 0.34 0.64 0.90 1.85*** 0.27 1.70* 1.85** 2.01 0.28* 2.08 (omitted) 1.19 0.67 0.71 1.24 2.35 – – – – 0.76 – 0.00 1.00
1.23*** – – – – – 0.71** 1.03 0.94 – 1.12 1.41* – 1.33 0.87 1.20 0.32*** 0.80* 0.77** 0.78 0.97 1.20 1.72*** 2.12*** 0.78* 1.28 1.00 1.42*** 1.37** 0.98 1.22 – – – – 0.99 – 0.25 0.01
1.03 – – – – – 0.14*** 0.71 1.03 – 0.96 0.84 – 0.54 0.57 0.53* 0.98 2.00*** 1.34 1.18 0.65 1.92** 1.12 1.63 0.56 3.93 0.04** 1.81** 1.66 0.65 3.72** – – – – 2.17** – 0.00 1.00
0.88*** – – – – – 0.53** 0.62** 2.83 – 1.34 1.99 – 0.00 (omitted) 0.00 0.00 1.29 2.64 1.24 0.61*** 1.47 1.10 1.16 0.71 1.48 0.00 1.06 0.76 1.45* 1.00 – – – – 0.35*** – 0.27 0.00
– – 2.93*** 1.54*** 3.62*** – 0.56*** 0.78* 1.18 – 0.99 1.16 – 0.93 0.87 1.11 0.48*** 1.12 0.97 1.15 0.85** 1.49*** 1.73*** 1.54*** 0.87 2.31** 0.67** 1.43*** 1.06 1.06 2.09*** 2.97*** 1.83*** 1.17 0.81 0.94 – 0.24 0.00
Denotes reference variable. We have also included an intercept. Significance level of 10%. Significance level of 5%. Significance level of 1%.
(alcohol abuse and sedentarism) are negatively correlated with smear test decisions. Smoking is positively correlated, but this result is only robust for older women. Private health insurance reinforces individual decisions on smear testing, while regular visits to the family doctor reduce it. Women that suffer from stress, have menopausal problems or take the pill are more likely to have tested for cervical cancer. Being in the target population for public smear test programmes is effective in reducing the probability of never taking the test. There is empirical evidence that the demand for smear tests increased in 2011/12 compared with 2006/07, for all age groups except women aged 50–69. The regional level variance u2 is significant, even after controlling for regional public screening programmes. The addition of regional public health policies does not reduce the unexplained variability among regions. Regional variance is not significant for age population groups, except for women aged 50–69 (Table 4).
Finally, we selected those women who have screened for breast cancer at least once and analysed the determinants for breast screening every 2 years or more. Similarly, we selected those women who have tested for cervical cancer and we analysed the determinants for smear testing every 5 years or more. Determinants of screening with adequate frequency were similar to those obtained for the decision on whether to screen or not. Younger women are the most likely to undertake breast screening, older women are the most likely to take smear tests, reaching a peak for women aged 50–69. In comparison with married women, widows are less likely to screen or test and single women are less likely to take a smear test. Lack of physical exercise and self-reporting of health problems are negatively correlated with regular breast screening whilst consuming ecological products and having private health insurance are positively correlated with regular smear tests. Living in a region which guarantees 100% coverage for breast screening is effective in encouraging women to take
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
8
Table 4 Determinants of cervical screening decisions: logistic multilevel regression results and odds ratios. Dependent variable: SmearTest. Model 1
Age Age: 16–29a Age: 30–49 Age: 50–69 Age: 70–older Marrieda Single Widowed Divorced Primary/No formal educationa Secondary education Tertiary education Employeda Retired Unemployed Homemaker Other Overweight Daily smoker Daily drinker Sedentary lifestyle Daily fruit Eco-goods Private heath insurance Family doctor Hormone products Contraceptive pill Health problems Stress Unhappiness Menopause problems TargetBreastTest TargetSmearTest TotalCoverageBreastTest PopulationBasedSmearTest Year 2006/07 Year 2011/12a Random effect: u2 LR test (Prob > 2 ): a * ** ***
Model 2
Total sample
Age: 16–29
Age: 30–49
Age: 50–69
Age: 70–older
Total sample
– – 1.90*** 1.25 0.34*** – 0.24*** 0.77** 1.12 – 1.80*** 2.47*** – 0.50*** 0.78 0.65*** 0.34*** 0.92 1.32** 0.83* 0.74*** 1.50*** 1.82*** 1.69*** 0.78* 1.60 2.76*** 1.09 1.19* 1.06 2.03*** – – – – 0.52***
1.23*** – – – – – 0.46** (omitted) 0.20 – 1.05 1.49 – (omitted) 0.66 0.47* 0.36*** 0.77 1.39 1.29 0.93 1.89*** 3.48* 1.70 0.60 (omitted) 3.52*** 0.91 1.02 1.63 6.14 – – – – 0.20**
1.04*** – – – – – 0.25*** 0.29** 1.49 – 1.93*** 2.13*** – 1.23 0.74 0.71 0.70 0.93 1.13 0.63** 0.89 1.30 1.46 2.95*** 0.97 1.43 1.66 1.07 1.93*** 0.76 1.10 – – – – 0.45**
0.94*** – – – – – 0.29*** 0.99 0.66 – 2.11*** 2.64*** – 0.89 1.25 0.85 2.02 0.81 0.91 0.86 0.72** 1.51** 2.00** 1.48 0.79 2.24 8.72 0.96 1.13 1.15 2.16*** – – – – 1.22
0.93*** – – – – – 0.28*** 1.02 6.52*** – 1.69** 2.20** – 0.00 (omitted) 0.00 0.00 0.87 3.16** 0.86 0.62*** 1.48 1.53 1.15 0.69 0.44 3.52 1.30 0.92 1.02 2.25*** – – – – 0.30***
– – 1.50** 0.77 0.27*** – 0.25*** 0.83 1.12 – 1.81*** 2.45*** – 0.50*** 0.80 0.65*** 0.35*** 0.90 1.34** 0.82* 0.76*** 1.51*** 1.80*** 1.69*** 0.80* 1.62 2.82*** 1.09 1.19* 1.02 1.99*** 1.24 1.70*** 0.93 0.71 0.54***
0.37 0.00
0.18 0.35
0.36 0.83
0.49 0.00
0.35 0.68
0.39 0.00
Denotes reference variable. We have also included an intercept. Significance level of 10%. Significance level of 5%. Significance level of 1%.
regular checks. Living in a region where the public smear test programme is population based may be also be effective in encouraging women to smear test regularly, but this result was not statistically significant. Previous results suggested that in 2011/12 (compared to 2006/07) women were more likely to have undergone breast screening at least once, although an analysis of screening frequency, shows that this does not mean that they are screening with adequate frequency. The regional level variance, u2 , is meaningful and significant for both breast screening and smear testing (Table 5). 6. Conclusions and policy implications This, Spanish-based, research validates international evidence from other European countries and North America. Results demonstrate that educational levels have a positive effect on the decision to screen. Women who adopt healthy habits and avoid risk behaviours are also more
likely to test for breast cancer and cervical cancer. Having private health insurance makes women more likely to screen and access to a public health care professional has the same effect. There were significant socioeconomic health disparities and this makes public health screening strategies especially important. It should be noted that there are significant regional variances that remain unexplained even after controlling for regional public health policies. Results further underline the fact that public health screening programmes improve individual screening decisions (more women screen). Being in the target population for public smear test programmes improves screening decisions for both breast cancer and cervical cancer. It is clear that public health programmes are positive in direct and indirect ways. Finally, even if dummy time variables reveal that the rates of screening for breast and cervical cancer are increasing, there are doubts that women are attending these screenings more regularly. It is not only important to
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx Table 5 Determinants of screening frequency decisions for breast cancer and cervical cancer: logistic multilevel regression results and odds ratios.
a
Age: 16–29 Age: 30–49 Age: 50–69 Age: 70–older Marrieda Single Widowed Divorced Primary/No formal educationa Secondary education Tertiary education Employeda Retired Unemployed Homemaker Other Overweight Daily smoker Daily drinker Sedentary lifestyle Daily fruit Eco-goods Private health insurance Family doctor Hormone products Contraceptive pill Health problems Stress Unhappiness Menopause problems TargetBreastTest TargetSmearTest TotalCoverageBreastTest PopulationBasedSmearTest Year 2006/07 Year 2011/12a Random effect u2 LR test (Prob > 2 ): a * ** ***
BreastTestRight
SmearTestRight
– 0.28*** 0.36** 0.10*** – 0.98 0.70* 0.84 – 0.79 0.92 – 1.17 2.24** 0.96 3.53*** 1.04 1.00 0.98 0.76** 1.05 0.82 1.17 0.70 1.32 0.65 0.74** 1.15 0.79 0.97 1.05 0.77 1.70** 0.64 6.47*** – 0.23 0.06
– 1.88*** 2.31*** 1.78** – 0.80* 0.51*** 0.85 – 1.16 1.18 – 1.14 0.98 0.82 1.25 0.94 1.01 1.00 1.05 0.98 0.95 1.87*** 1.05 1.65 0.99 0.89 1.10 0.83* 0.89 1.03 1.08 0.59** 1.20 1.00 – 0.30 0.00
Denotes reference variable. We have also included an intercept. Significance level of 10%. Significance level of 5%. Significance level of 1%.
9
individual might be able to change and improve (for example, lifestyles) [36]. Future research could examine differences in age groups and the effect of regular screening over time. Age is the most common risk factor for breast cancer - approximately one-third of all breast cancers occur in women of 70 or older. Although women over 70 years old are not a target population group for breast cancer, screening for breast cancer among older women has increased if we compare 2011/12 with 2006/07. There is empirical evidence that public screening programmes and treatment protocols have helped to reduce cancer mortality rates in developed countries [9]. Further national studies are important in order to promote the implementation of screening programmes in developing countries but also to rank and select the most cost-effective screening strategies in those countries with a long history of public screening. The present work is not free of technical problems. Among the main limitations is the use of two cross-sections for the analysis of micro-data. As previously stated, for reasons of data comparability, it was not possible to use previous Spanish National Health Surveys, which would have, for example, allowed a micro–macro analysis. In addition, the use of a panel data would have reduced biases derived from individual heterogeneity. Although we feel comfortable that our results provide the right correlation effects, causality should be analysed with caution. Acknowledgements This research was completed as part of the Spanish Ministry of Innovation and Science CSO2011-30089 Health Policies research project: ‘Sexual and Reproductive Behaviour in Spain: Risk Factors for Health’. The authors are grateful to the reviewers and editor whose comments have contributed to improving the quality of the work. References
test for breast and cervical cancer, it is necessary to do it regularly. The results are consistent with other professional recommendations and findings on healthy lifestyles (regular exercise, a balanced diet, low alcohol and tobacco consumption and not using hormone replacement therapy during the menopause) and participation in screening programmes [4,31]. Once the correct level of attention is paid to individual behaviour, the question of socio-economic inequalities should be addressed. Even in countries with widespread testing programmes, socioeconomic inequalities in health care are important. Economically advantaged women are more likely to seek treatment by a specialist and this is particularly true if they have private insurance. In Spain, long waiting lists are the main reason why women take out private health insurance [32]. Health policies should target women with low socioeconomic levels [20,33–35]. Inequity in health and health care access should not be restricted to the analysis of individual characteristics; it should also consider factors that the
[1] Howlader N, Noone AM, Krapcho M, Garshell J, Neyman N, Altekruse SF, et al. SEER cancer statistics review 1975–2010. Bethesda, MD: National Cancer Institute; 2013. [2] Wübker A. Explaining variations in breast cancer screening across European countries. European Journal of Health Economics 2013, http://dx.doi.org/10.1007/s10198-013-0490-3. [3] Spanish Society of Medical Oncology. Info cancer; 2013. Available from: http://www.seom.org [accessed 10.07.13]. [4] Spanish Association against Cancer. Sobre el cancer; 2012. Available from: https://www.aecc.es/SobreElCancer [accessed 10.07.13]. [5] DeMay M. Practical principles of cytopathology, revised ed. Chicago, IL: American Society for Clinical Pathology Press; 2007. [6] Kopans DB. Why the critics of screening mammography are wrong. Diagnostic Imaging 2009;31(12):18–24. [7] Mulcahy N. Screening mammography benefits and harms in spotlight again. Medscape 2009;94(6):411–3. [8] de Koning HJ, van Ineveld BM, van Oortmarssen GJ, de Haes JC, Collette HJA, Hendriks J, et al. Breast cancer screening and costeffectiveness; policy alternatives, quality of life considerations and the possible impact of uncertain factors. International Journal of Cancer 1991;49(4):531–7. [9] Gøtzsche PC, Nielsen M. Screening for breast cancer with mammography. Cochrane Database of Systematic Reviews 2011;(1). Available from: CD001877. [10] U.S. Preventive Services Task Force. Breast cancer: screening. Rockville, MD: United States Preventive Services Task Force; 2009.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
[11] The European Cancer Observatory; 2011. Available from: http://eco.iarc.fr [accessed 10.07.13]. [12] Carrieri V, Wübker A. Assessing inequalities in preventive care use in Europe: a special case of health-care inequalities? Health Policy 2013;113:247–57. [13] de Geldera Bulliard J, de Wolf C, Fracheboud J, Draisma G, Schopperd D, de Koning J. Cost-effectiveness of opportunistic versus organised mammography screening in Switzerland. European Journal of Cancer 2009;45:127–38. [14] Catalá-López F, García-Altés A, Álvarez-Martín E, Gènova-Maleras R, Morant-Ginestar C, Arana E. Evaluación económica de interven˜ revisión sistemática ciones en enfermedades oncológicas en Espana: y análisis comparativo. Farmacia Hospitalaria 2012;36(3):141–7. [15] Spanish National Survey on Health (2006/07 and 2011/12). Data available from: http://www.msc.es/estadEstudios/estadisticas/ encuestaNacional/ [accessed 14.10.13]. [16] Mandelblatt J, Andrews H, Kemer J, Zauber A, Bumetu W. Determinants of late stage diagnosis of breast and cervical cancer: the impact of age, race, social class and hospital type. American Journal of Public Health 1991;5:646–9. [17] Blackburn GL, Wang KA. Dietary fat reduction and breast cancer outcome: results from the Women’s Intervention Nutrition Study (WINS). American Journal of Clinical Nutrition 2007;86(3):878–81. [18] Boffetta P, Hashibe M, La Vecchia C, Zatonski W, Rehm JA. The burden of cancer attributable to alcohol drinking. International Journal of Cancer 2006;119(4):884–7. [19] Lairson D, Chan W, Newmark GR. Determinants of the demand for breast cancer screening among women veterans in the United States. Social Science and Medicine 2005;61:1608–17. [20] Sutton S, Rutherford C. Sociodemographic and attitudinal correlates of cervical screening uptake in a national sample of women in Britain. Social Science and Medicine 2005;61:2460–5. [21] Whynes D, Philips Z, Avis M. Why do women participate in the English cervical cancer screening programme? Journal of Health Economics 2007;26:306–25. [22] Carney P, O’Neill S, O’Neill C. Determinants of breast cancer screening uptake in women, evidence from the British household panel study. Social Science and Medicine 2013;82:108–14. [23] Luengo-Matos S, Polos-Santos M, Saz-Parkinson Z. Mammography use and factors associated with its use after the introduction of breast cancer screening programmes in Spain. European Journal of Cancer Prevention 2006;15(3):242–8. [24] Cabeza E, Esteva M, Pujol A, Thomas V, Sánchez-Contador C. Social disparities in breast and cervical cancer preventive practices. European Journal of Cancer Prevention 2007;16(4):372–9.
[25] Martín-López R, Hernández-Barrera V, López-Andrés A, GarridoCarrasco P, Gil-Miguel A, Jiménez-García R. Breast and cervical cancer screening in Spain and predictors of adherence. European Journal of Cancer Prevention 2010;19(3):239–45. [26] Puig-Tintoré L, Castellsagué X, Torné A, Sanjosé S, Cortés J, Roura E, et al. Coverage and factors associated with cervical cancer screening: results from the AFRODITA study: a population-based survey in Spain. Journal of Lower Genital Tract Disease 2008;12(2): 82–9. [27] Johnson-Thompson MC, Guthrie J. Ongoing research to identify environmental risk factors in breast carcinoma. Cancer 2000;88(5):1224–9. [28] Coughlin SS, Leadbetter S, Richards T, Sabatino SA. Contextual analysis of breast and cervical cancer screening and factors associated with health care access among United States women, 2002. Social Science and Medicine 2008;66:260–75. ˜ [29] Ascunce N, Salas D, Zubizarreta R, Almazán R, Ibánez J, Ederra M. Cancer screening in Spain. Annals of Oncology 2010;21: 43–51. [30] Ascunce N, Delfrade J, Salas D, Zubizarreta R, Ederra M. Programas ˜ características y de detección precoz de cáncer de mama en Espana: principales resultados. Medicina Clínica 2013;141:13–23. [31] Sala M, Queiro T, Castells X, Salas D. Situación del cribado de cáncer ˜ In: Castells X, Sala M, Ascunce N, Salas D, de cérvix en Espana. Zubizarreta R, Casamitjana M, editors. Descripción del cribado del ˜ cáncer en Espana Proyecto DESCRIC. Madrid: Ministerio de Sanidad y Consumo; 2007. [32] Spanish Ministry of Health, Social Policy. Estrategia en Cáncer del Sistema Nacional de Salud. Madrid: Spanish Ministry of Health and Social Policy; 2010. [33] Pinilla J, González B, Barber P, Santana Y. Smoking in young adolescents: an approach with multilevel discrete choice models. Journal of Epidemiological Community Health 2002;56: 227–32. [34] Szumilas M. Explaining odds ratios. Journal of the Canadian Academy of Child and Adolescent Psychiatry 2010;19:227–9. [35] Paléncia L, Espelt A, Rodríguez-Sanz M, Puigignós R, Pons-Vigués M, Pasarín MI, et al. Socio-economic inequalities in breast and cervical cancer screening practices in Europe: influence of the type of screening program. International Journal of Epidemiology 2010;39: 757–63. [36] Doescher M, Jackson JE. Trends in cervical and breast cancer screening practices among women in rural and urban areas of the United States. Journal of Public Health Management and Practice 2009;15(3):200–9.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
Health Policy xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Health Policy journal homepage: www.elsevier.com/locate/healthpol
Female preventive practices: Breast and smear tests Ana Isabel Gil Lacruz a , Marta Gil Lacruz b,∗ , Sophie Gorgemans a a Department of Organization and Management Business, Industrial Engineering School, Edificio Agustín de Betancourt. C/María de Luna s/n, Zaragoza 50018, Spain b Psychology and Sociology Department, Health Sciences Faculty, C. Domingo Miral s/n, Zaragoza 50009, Spain
a r t i c l e
i n f o
Article history: Received 3 September 2013 Received in revised form 14 April 2014 Accepted 21 April 2014
Keywords: Gender and health Cancer and palliative care Primary care Statistical methods Risk and health
a b s t r a c t Breast cancer and cervical cancer are the most common female cancers in Spain and in many developed countries. The main goal of this paper is to identify the determinants of individual decisions on breast screening and smear testing, that is to say, the decision to take a test for the first time and the decision to test with suitable regularity. To that end, we have combined analyses of micro and macro data (the Spanish National Health Survey and Spanish Regional Social Indicators) and employed multilevel estimation models. Among the main results, we highlight the fact that regional public screening programmes improve individual decisions on screening (more women testing for the first time and more women testing regularly) and, furthermore, they generate positive synergies; for example, regional public programmes for smear testing improve individual decisions on both cervical and breast cancer screening. In addition, we conclude that it is not only important to know if the numbers of women undergoing breast screening and smear testing are increasing, it is also important to know if they are testing regularly. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Breast cancer and cervical cancer are the most common types of female cancer in Europe and North America [1,2]. In Spain, as in most developed countries, the annual incidence of breast cancer is 51 new cases per 100,000, whilst for cervical cancer the figure is 8 new cases per 100,000. Breast cancer most frequently occurs in women between the ages of 35 and 80 but the peak onset period is between 45 and 65 [3]. The average age of cervical cancer diagnosis is 48 years old although approximately 47% of women with invasive cervical carcinoma are diagnosed before the age of 35 [4].
∗ Corresponding author. Tel.: +34 976 76 44 43; fax: +34 976 76 20 03. E-mail addresses: [email protected] (A.I. Gil Lacruz), [email protected] (M. Gil Lacruz), [email protected] (S. Gorgemans).
Since the introduction of smear tests, deaths caused by carcinoma of the cervix have fallen by up to 99% in populations in which women are regularly screened. A routine smear testing programme with appropriate follow-up can reduce the incidence of cervical cancer by up to 80% [5]. The mortality rate for breast cancer has fallen by almost 30% and two-thirds of the decrease has been attributed to screening [6,7]. Regular screening from the age of 50 is not only cost-effective [8], but it saves about 2 lives over 15 years for every 1000 women screened [7]. Screening can lead to longer survival rates but it can also result in over-diagnosis, or have no effect at all. In fact, the use of mammography as a screening tool for the detection of early breast cancer in healthy women without symptoms continues to be debated. The figures have to be seen in the light of errors in diagnosis, overtreatment, and radiation exposure. 1 woman in 2000 will have her life prolonged by 10 years through screening but 10 healthy
http://dx.doi.org/10.1016/j.healthpol.2014.04.012 0168-8510/© 2014 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
2
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
women will undergo unnecessary breast cancer treatment and 200 women will suffer from significant psychological stress due to erroneous diagnosis [9]. Despite these negative aspects, health organisations continue to recommend regular breast screening – every 2 years for women between the ages of 50 and 74 [10] or every 2–3 years between 50 and 69 [11]. In Spain, cancer screening is in accordance with national and international recommendations. Despite the fact that screening is free, or the charge is symbolic, there are important socio-economic inequalities in the take-up of breast screening. There are a number of possible explanations for this: (i) the poor and less educated might feel less confident and be less inclined to request specialist consultation (instead of, for example, asking the family doctor); (ii) they are also more likely to suffer from fear and anxiety which are negatively correlated with screening; and (iii) the poor and less educated are more likely to turn to preventive care later than the economically advantaged, for example, they wait until after the appearance of health problems [12]. Health expenditure and finance has become a key issue in these times of economic crisis. Demographic and socioeconomic changes are affecting the financial sustainability of the welfare state and an economic perspective on health issues is essential. Mammographic screening programmes for women aged 50–70 have been shown to be effective in reducing mortality rates at reasonable cost [13] but they are not necessarily cost-effective in every situation and the inclusion of other age groups is far from accepted, particularly in terms of cost and the impact on quality of life. Economic evaluations of interventions for malignant neoplasms are not common despite their gradual increase in recent years in Spain [14]. In the case of cervical cancer, the Spanish government recommends human papilloma virus vaccination for teenage girls, despite its high cost and persistent doubts about its effectiveness and the need to continue regular cytology screening [13]. The aim of this paper is to study the determinants of breast and cervical cancer screening among Spanish women. The Spanish National Health Survey (2006/07 and 2011/12) [15] has been used to provide information on sociodemographic characteristics, risk behaviours, access to health services and the state of health of Spanish women. Special attention is paid to questions related to preventive practices such as screening controls and frequency. The research should make empirical evidence on vulnerable population groups and the efficiency of public health policies available to policy makers. The main research contribution of this paper is the use of both micro data and regional health public policies in the analyses. We have examined data at different aggregation levels (individuals and regions – the Spanish Autonomous Communities) in order to determine if geographical differences are consequences of population characteristics (e.g. gender, age or working status), contextual data (e.g. regional public health policies) or simply unobserved contextual data. We have also repeated estimations by age cohorts in order to understand if different population groups require specific policies. Not all population groups are targeted by public screening programmes, so we are able to check if women of different ages are particularly sensitive to specific factors.
2. Review of the published literature Breast and cervical cancer incidence rates in Spain are slowly increasing; this is probably due to the country’s ageing population and more frequent early diagnosis [4]. An ageing population is not unique to Spain, it affects most developed countries. The positive relationship between age and screening participation with breast and cervical cancer rates has been known for decades [16]. Screening is a key tool to prevent cancer deaths. For example, screening attendance of 70% might reduce breast cancer mortality by about 25% among women aged 50–69 [11]. In Europe, the availability of screening programmes explains about 40% of cross-country differences in screening rates and factors like age, education, health status, etc. are associated with screening uptake within countries [2]. Screening is not the only preventive action that can be taken; healthy habits also play an important role. Smoking is a leading factor in cervical cancer [16] and a high-fat diet [17] and excessive alcohol intake [18] are risk factors for breast cancer. Smoking is a negative health behaviour that serves as a proxy for poor health habits such as lack of exercise, bad diet and alcohol abuse [19]. Age, smoking and health care access are inversely related to screening, whilst education, income, insurance and perceived risk of cancer are directly related [19–21]. Women who are strongly influenced by the advice of their General Practitioners tend to take regular tests, women who do not receive this advice screen less frequently, or not at all [21]. In the USA, health insurance reduces the cost of pointof-service care and has a positive effect on the demand for screening [19]. In the UK, there is empirical evidence that income does not influence uptake as screening is free of charge [22]. The literature on Spain concludes that invitation to screening, visits to the gynaecologist and women’s attitudes are the main reasons for testing [23]. More effort is required for women over 65 with regards to breast cancer testing and women over 55 for cytology. Special attention needs to be paid to women in lower socioeconomic level groups [23–26]. Additional medical insurance coverage increases the probability of regular screening and this is especially true for mammography [24]. Unhealthy lifestyles have been associated with non-adherence [25]. The implications of lifestyles on health are also validated by international comparisons. The incidence of both types of cancer has increased in Spain in recent decades but they are still lower than rates in North America. Cervical and breast cancer are related to lifestyles as well as cultural and environmental factors [27]. Individual and contextual covariates (such as the number of health care centres or population characteristics) are associated with the uptake of breast and cervical cancer screening [28]. 3. The data: public health policies, individual observations and regional indicators The Spanish Constitution of 1978 gave its Autonomous Communities (also known as regions) the legal authority
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
to regulate health care. A National general strategy against cancer was defined in 2003 (finally adopted in 2006) with specific objectives in early detection [29,30]. All the regions have specific breast cancer screening programmes that include women between 50 and 69 years old, but in some regions the age range can be as low as 45. In general, the frequency of testing is every two years. For breast cancer screening, each autonomous community uses an information system based on an individual health card (in some regions this is combined with the data from the population census) and sends a written invitation to take part in the programme. The first population programme for early detection breast cancer began in Navarra in 1990. Navarra was also the first community that achieved 100% coverage of women in the target population in 1999. New programmes were gradually established and all regions achieved 100% coverage in 2009. Screening programmes for the detection of cervical cancer are less well established but since 1993, the list of public health services offered has included a Pap cytology test for women between the ages of 35 and 64. The frequency of testing is usually between 3 and 5 years [31]. All regions currently have cervical cancer screening programmes, though these are mostly of the opportunistic type, except in La Rioja and Castilla-Leon where the programme is population based. Opportunistic selection methods are based on access to primary health care and specialists [29]. 100% coverage is easier to achieve with selection methods based on population; opportunistic selection depends on women attending primary health care centres or specialists. In the NHS Cancer Strategy (2009) [32] the recommendation was to reach 80% of women between 40 and 50 years old with a smear test every 5 years [29,30]. This information has been used to study health policy variables in the regression analysis. The micro analysis is based on women in the target population groups for breast screening and smear tests. The macro analysis considers the length of time that public health screening programmes have been operating in the Spanish regions (Table 1). Individual data and regional information on social indicators and health issues have been used for the empirical study (Table 2). Individual observations have been taken from the Spanish National Health Surveys of 2006/07 and 2011/12 [15]. The National Health Survey (NHS) has been carried out by the Spanish Ministry of Public Health since 1987. The reason why waves before 2006 were not considered is because important questions for our research goals, such as response categories for individual screening frequency, are not comparable through time. The survey is cross-sectional and uses a stratified multistage sample. The first stage units are census sections. The second stage units are the main family residences. One adult (aged 16 years or over) is selected from each household to complete the Adult Questionnaire; if there any minors in the house (under 16 years of age), one is selected to complete the Minor Questionnaire. The method for collecting information is personal interview, in exceptional cases this can be done by telephone. At a national level, the survey sample is representative to 95%. A figure of over 90% is generally considered as sufficient to
3
represent target populations. The national response rate of 95% was achieved by an effective household response rate of 65% and a substitution rate of 30%. In all the autonomous communities the effective response rates are over 90% [15]. The survey provides information on: sociodemographic characteristics (age, gender, civil status, educational level and employment); lifestyles (smoking, drinking, physical exercise, Body Mass Index and nutrition); access to health services (private insurance, visits to the family doctor and prescriptions); and state of health (general health problems, stress, unhappiness, menopausal problems). 25,782 observations were made regarding women over 16 years old and living in Spain. To construct the dependent variables for breast cancer screening (BreastTest and BreastTestRight), we have used the following questions: “Have you ever had a mammography?” (1: Yes; 6: No; 8: Don’t know; 9: No answer) and “If you have had a mammography, please indicate the time period since your last screening” (1: In the last 12 months; 2: More than 1 year but less than two years; 3: More than two years but less than 3 years; 4: Over 3 years ago; 8: Don’t know; 9: No answer). To construct the dependent variables for cervical cancer screening frequency (SmearTest and SmearTestRight), the questions were: “Have you ever had a cervical smear test?” (1: Yes; 6: No; 8: Don’t know; 9: No answer) and “If you have had a cervical smear test, please indicate the time period since your last screening” (1: 3 years ago or less; 2: More than 3 years but no more than 5 years; 3: Over 5 years ago; 8: Don’t know; 9: No answer). The dependent variables BreastTest and SmearTest are categorical variables calculated for all the interviewed women. The variable BreastTest takes the value 1 if the woman has breast screened at least once, 0 if not. In the same way, SmearTest takes the value 1 if the woman has had a smear test at least once, 0 if not. The dependent variables BreastTestRight and SmearTestRight are categorical variables calculated for all the interviewed women that have breast screened – BreastTestRight – and had a smear test – SmearTestRight – at least once. The variable BreastTestRight takes the value 1 if the woman has breast screened every 2 years or more, 0 if not. SmearTestRight takes the value 1 if the woman has had a smear test every 5 years or more often. It is very important to know if women have ever been screened for breast cancer and cervical cancer as this provides information on whether women access and use screening services; moreover, subsequent appointments are automatically generated after the first screening. This is particularly relevant for women who are younger than the target population group for the breast test. For example, there might be women younger than 45 that want breast screening because they have a lump in the breast or because they have a family history of breast cancer. In the case of the smear test, most regions have opportunistic programmes so screening history begins when the family doctor makes the first appointment. In general, it appears that women need to be more proactive in demanding the first smear test than the first breast test. The data shows that around 56% of Spanish women have breast screened and 67% have smear tested at least once.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
4
Table 1 Breast and cervical screening programme in Spanish regions. Mammography
Papanicolau cytology Selection method: PB
Year 100% coveragea
Year
Population
Periodicity
Selection method
50–69 50–69a
CP CP & IHC
2005 2006
1986 1993
20–60b 35–64
OP OP
1991
50–69
CP & IHC
2000
1993
25–65
Balearic Islands Canary Islands
1997 1999
50–69 50–69
CP & IHC CP & IHC
2009 2005
1993 1995
25–64 18–65
Cantabria
1997
50–69
CP
1997
1997
25–65
Castilla Mancha Castilla León
1992 1992
45–69 45–69
CP & IHC CP & IHC
1997 1996
1993 1986
25–60 30–65
Catalonia Extremadura
1992 1998
50–69 50–69
CP & IHC CP & IHC
2004 2005
1993 1983
25–65 20–65b
Galicia Madrid
1992 1999
50–69 50–69
CP & IHC IHC
1998 2001
1993 1993
20–65 35–65
Murcia
1994
50–69
IHC
1999
1993
35–64
Navarra The Basque Country
1990 1995
45–69 50–69
CP CP
1992 2000
2000 1999
25–65 25–59
La Rioja Valencia
1993 1992
45–69 45–69
CP & IHC CP & IHC
1995 2001
1993 1995
25–65 20–65
3 years 3 years (annual first 2 years) 3 years (annual first 2 years) 3 years 3 years (annual first 2 years) 5 years (annual first 2 years) 3 years 3 years (annual first 2 years) 3 years 3 years (annual first 2 years) 3 years No periodicity defined 5 years (annual first 2 years) From 3 to 5 years 3–5 years (annual first 2 years) 3 years 3 years
Year
Population
Andalucía Aragon
1995 1997
Asturias
Periodicity
2 years
OP OP OP OP OP PB OP OP OP OP OP OP OP PB OP
PB, population based; CP, census population; IHC, individual health card; OP, opportunistic. Table: authors’ work [29–32]. a To 64 years old before 2007. b From 15 years old before 2007.
80% of women who have screened and tested at least once have done so with the appropriate frequency. An analysis of regional information on public health screening programmes reveals that 29% of the women interviewed are in the target population for breast screening programmes and 58% of them are in the target population for cervical screening programmes (TargetBreastTest and TargetSmearTest). In addition, in the year of the interview, 90% had lived in communities where the coverage of breast screening is 100% of the target population while 7% lived in areas where smear testing is based on a population method (TotalCoverageBreastTest and TotalCoverageSmearTest). Other relevant statistics were: 11% of Spanish women have private health insurance; 87% of the population visited the family doctor at least once in the 4 weeks prior to the interview; 44% of Spanish women are overweight, 20% are daily smokers and 43% do no regular physical activity; and 72% eat fresh fruit and vegetables every day. 4. The empirical framework This research focuses on two questions: (a) Do women screen for breast cancer and smear test for cervical cancer? (b) If the answer is ‘yes’, are they screening and testing with appropriate frequency? We considered a set of sociodemographic variables (age, gender, marital status, working conditions, etc.) that may help to explain individual decisions. The main contribution of this work is the utilisation
of policy variables, such as screening programmes, on screening uptake. As empirical strategy, we used multilevel models (STATA: xtmelogit). Multilevel regression models are indicated when there is a hierarchical structure in the levels of data, with a single dependent variable measured at the lowest level and a set of explanatory variables on the other levels. The advantage of these models is their capacity to define and explore variations at each level of the hierarchy after controlling for relevant explanatory variables [33]. In this case, the multilevel analysis helps us to understand individual and regional determinants of screening decisions and to determine if regional differences remain after controlling for individual variables and regional policies. The data is structured by j Spanish Autonomous Communities, in each of which, nj persons have been interviewed. The dependent variables, BreastTestij and SmearTestij , consider if the individual i of region j has tested for breast and cervical cancer at least once. Therefore, the demand functions1 for breast screening are characterised as: BreastTestij = ˇj + +εij
(1)
1 Analysis of the robustness of the estimated parameters involved repeated estimations including a different set of explanatory variables. Given that the coefficient parameters remain constant, these results have not been included (for reasons of space); they are available on request.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
5
Table 2 Spanish National Health Survey (2006 and 2011). Sample = 25,782 women from 16 years and older. Variables
Description of dummy variables (1: yes/0: no)
Mean/SD
BreastTest: has undergone breast screening BreastTestRight: screening for breast cancer every 2 years or more often (only women who have screened at least once) SmearTest: has taken a smear test SmearTestRight: smear test every 5 years or more often (only women who have smear tested at least once)
0.56 (0.40) 0.80 (0.39)
Age 16–29 Age 30–49 Age 50–69 Age 70–older
0.09 (0.29) 0.35 (0.47) 0.29 (0.45) 0.27 (0.26)
Marital status
Married Single Widow Divorced/separated
0.48 (0.50) 0.24 (0.48) 0.20 (0.40) 0.38 (0.49)
Educational level
Women with primary or no education Secondary education Tertiary education
0.47 (0.40) 0.38 (0.48) 0.15 (0.35)
Employment status
Employed Retired Unemployed Homemaker Other
0.35 (0.47) 0.27 (0.43) 0.06 (0.25) 0.28 (0.45) 0.04 (0.20)
Lifestyles
Overweight: Body Mass Index > 25 Smoker: daily smoker Alcohol: drinks alcohol every day Sedentarism: no physical exercise Daily fruit: consumes fresh fruit and vegetable every day Eco-goods: consumes ecological products
0.44 (0.49) 0.20 (0.40) 0.15 (0.36) 0.43 (0.49) 0.72 (0.44) 0.05 (0.23)
Health good and services access
Private insurance: holds a private health insurance policy Family doctor: visit to family doctor in last 4 weeks Hormone products: uses hormone products Pill: uses contraceptive pills
0.11 (0.32) 0.87 (0.33) 0.01 (0.11) 0.04 (0.19)
Health state
Health problems: had health problem Stress: feels stressed Unhappiness: feels unhappy Menopause problem: has problems with menopause
0.40 (0.49) 0.29 (0.46) 0.22 (0.41) 0.09 (0.28)
TargetBreastTest: target population for regional public breast screening TargetSmearTest: target population for regional public smear test TotalCoverageBreastTest: living in a region with 100% public coverage for breast screening for the target population PopulationBasedSmearTest: living in a region where the selection for smear test is population based
0.29 (0.45) 0.59 (0.49) 0.90 (0.29)
Screening
Age
a
b
Health policies
a b
0.67 (0.47) 0.80 (0.40)
0.07 (0.26)
All variables are dummy variables, except those related to the age of the individual which are numerical. Authors’ work, based on the data in Table 1. Contextual dummy variables (Year 2006/07 and Year 2011/12; 17 Spanish Autonomous Communities).
where ˇj represents the mean of the dependent variable of all the individuals of region j and εij represents the deviation of the value of this variable for the individual i from the mean of that variable in the other interviewees who live in the same j. ¯ Eq. (1) can be expressed Deviation of ˇj from its mean ˇ, as: ¯ + uj + εij where uj = ˇj − ˇ ¯ BreastTestij = ˇ
(2)
The model formulated in this equation coincides with the components of variances with fixed and random effects. ¯ represents the fixed effects. This model The parameter ˇ assumes that the random effects uj are distributed normally with mean 0 and variance u2 = ˇ2 , which means that differences in the variable BreastTest are attributable to the
country. It also assumes that the error component εij is distributed normally with mean 0 and variance 2 . Finally, it assumes that the random effects uj and the error component εij are independent and that the εij are all independent from each other. The second step introduces information about K−1 individual characteristics (socioeconomic factors: lifestyles, health care and state of health). These characteristics can be included in a vector Xij that when introduced into the model would involve the appearance of K new fixed effects (K−1 variables and a constant). Eq. (2) would then become: BreastTestij = Xij ˇj + uj + εij
(3)
in which X includes K regressors and εij ≈ N(0, 2 ).
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
6
The third, and final step, parameterises the coefficients ˇj of Eq. (3) by adding regional explanatory variables (regional health policies): ¯ j + uj + εij BreastTestij = Xij ˇ
(4)
These models represent general specifications of multilevel models in which the responses to a dependent variable are related in a linear function to a set of explanatory variables and to a simple hierarchical structure. Because our dependent variables are binary, taking values of 1 if the woman has screened and 0 if not, we applied multilevel analysis with a logistic function. Fixed effects are now dependent on individual and regional variables and a constant. The same process was repeated for SmearTest. Agepopulation subsamples (16–29, 30–49, 50–69, and 70+) were also considered for both dependent variables. Different population groups were used because not all of them are targeted by public screening programmes, so it is possible to determine if women of different ages are especially sensitive to different individual factors. Screening for breast and cervical cancer at least once is better than not at all, but it might be relevant to look at individual decisions on screening and testing frequency. Therefore, for the subsample of women who had screened at least once, we searched for the determinants for screening for breast cancer every 2 years or more (BreastTestRight) and testing for cervical cancer every 5 years or more (SmearTestRight). Once again, the dependent variables are binary, taking values of 1 if the woman tests with adequate frequency and 0 if not. Finally, we report the results as odds ratios (OR). The odd ratio is a measure of the association between an exposure and an outcome that represents the chances that an outcome will occur given a particular exposure, compared to the chances of the outcome occurring in the absence of that exposure. Odds ratios are most commonly used in case–control studies but they are also used in crosssectional studies [34]. 5. Results The results are presented in the same order as the previous section: Women aged 50–69 are the most likely to have participated in breast screening programmes; younger women (aged 16–29) are the least likely. Odds ratios provide information about association effects in terms of sense and intensity. For example, looking at the total sample in model 1, women aged from 30 to 49 are 3.75 times more likely to have undertaken breast screening than women aged from 16 to 19. In general, demand for breast screening is higher among more mature women and this is especially true for women aged 30–49. Widows are less likely to have demanded breast screening, and they are tested with less regularity than married women (a result may well be affected by age – widows tend to be older than married women). Single women are less likely to have demanded breast screening than married women – a result that was robust for all age groups. Employment status and educational levels appear to have no significance. The only two statistically
significant results in this section were that among women aged 50–69, housewives are less likely to have demanded breast screening than women who work and, among women aged 30–49, having a university degree is positively correlated with testing for breast cancer. Similar results can be inferred for healthy behaviours (a healthy diet and consuming ecological products) and the contrary results apply to unhealthy behaviours (smoking and sedentarism). These results are particularly consistent for women aged 30–69. Private health insurance reinforces individual decisions on breast screening, but this result was only robust for women aged 30–49, and in this group, visits to the family doctor have the opposite effect. Consuming hormone products, self-reporting poor health, suffering stress and having menopausal problems are positively correlated with being screened. Being in the target population for public breast screening programmes is effective, in that it improves individual screening decisions (the women are more likely to screen) and there is a further, indirect effect, as being in the target population for smear testing also improves individual decisions on breast screening. Among older women, breast cancer screening levels increased between 2006/07 and 2011/12. The regional level variance u2 is significant, even after controlling for regional public screening programmes. Although the addition of regional public health policies does not reduce the unexplained variability among regions, the estimation for breast screening decisions by age cohorts shows that regional variance is not significant for the population groups with the highest and the lowest rates for breast screening (women aged 50–69 and women younger than 29). Women aged 50–69 are in the target population groups for all regions; women younger than 29 years old are not screened under the public health system unless there are health symptoms or a family history of breast cancer. Nevertheless the coefficients of the regional variance for both age groups are not statistically significant. On the other hand, in the 30–49 age group, there are women who are in the target population and women who are not. Furthermore, women that are in the target population in one region might not be in the target population in other regions (women from 45 to 49 years old). For older women, the date when the public screening programmes were implemented may be more relevant to the question of whether they have been considered as being in the target population groups at least once in their lives (Table 3). With regards to smear tests, older women (over 70) and younger women (under 36) are less likely to participate in screening programmes. Women between the ages of 30 and 49 are more likely to have demanded smear tests. For women between 16 and 49, age is positively correlated with the demand for smear tests; for women older than 50, age is negatively correlated. Widows and single women are less likely to have demanded a smear test than married women. Homemakers are less likely to have tested for cervical cancer than women in paid employment, but this result is only robust for women in the youngest age cohort. Educational level is positively correlated with testing for cervical cancer for all age groups except for women younger than 30 years old. Unhealthy behaviours
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
7
Table 3 Determinants of breast screening decisions: logistic multilevel regression results and odds ratios. Dependent variable: BreastTest. Model 1
Age Age: 16–29a Age: 30–49 Age: 50–69 Age: 70–older Marrieda Single Widowed Divorced Primary/No formal educationa Secondary education Tertiary education Employeda Retired Unemployed Homemaker Other Overweight Daily smoker Daily drinker Sedentary lifestyle Daily fruit Eco-goods Private health insurance Family doctor Hormone products Contraceptive pill Health problems Stress Unhappiness Menopause problems TargetBreastTest TargetSmearTest TotalCoverageBreastTest PopulationBasedSmearTest Year 2006/07 Year 2011/12a Random effect: u2 LR test (Prob > 2 ): a * ** ***
Model 2
Total sample
Age: 16–29
Age: 30–49
Age: 50–69
Age: 70–older
Total sample
– – 3.75*** 5.48*** 7.21*** – 0.53*** 0.67*** 1.20 – 1.00 1.16 – 0.98 0.84 1.12 0.45*** 1.17** 0.97 1.17 0.82*** 1.51*** 1.75*** 1.50*** 0.85 2.24** 0.62*** 1.40*** 1.04 1.13 2.21*** – – – – 0.85 – 0.17 0.00
1.14** – – – – – 1.76 (omitted) 2.52 – 0.94 1.28 – (omitted) 1.96 0.34 0.64 0.90 1.85*** 0.27 1.70* 1.85** 2.01 0.28* 2.08 (omitted) 1.19 0.67 0.71 1.24 2.35 – – – – 0.76 – 0.00 1.00
1.23*** – – – – – 0.71** 1.03 0.94 – 1.12 1.41* – 1.33 0.87 1.20 0.32*** 0.80* 0.77** 0.78 0.97 1.20 1.72*** 2.12*** 0.78* 1.28 1.00 1.42*** 1.37** 0.98 1.22 – – – – 0.99 – 0.25 0.01
1.03 – – – – – 0.14*** 0.71 1.03 – 0.96 0.84 – 0.54 0.57 0.53* 0.98 2.00*** 1.34 1.18 0.65 1.92** 1.12 1.63 0.56 3.93 0.04** 1.81** 1.66 0.65 3.72** – – – – 2.17** – 0.00 1.00
0.88*** – – – – – 0.53** 0.62** 2.83 – 1.34 1.99 – 0.00 (omitted) 0.00 0.00 1.29 2.64 1.24 0.61*** 1.47 1.10 1.16 0.71 1.48 0.00 1.06 0.76 1.45* 1.00 – – – – 0.35*** – 0.27 0.00
– – 2.93*** 1.54*** 3.62*** – 0.56*** 0.78* 1.18 – 0.99 1.16 – 0.93 0.87 1.11 0.48*** 1.12 0.97 1.15 0.85** 1.49*** 1.73*** 1.54*** 0.87 2.31** 0.67** 1.43*** 1.06 1.06 2.09*** 2.97*** 1.83*** 1.17 0.81 0.94 – 0.24 0.00
Denotes reference variable. We have also included an intercept. Significance level of 10%. Significance level of 5%. Significance level of 1%.
(alcohol abuse and sedentarism) are negatively correlated with smear test decisions. Smoking is positively correlated, but this result is only robust for older women. Private health insurance reinforces individual decisions on smear testing, while regular visits to the family doctor reduce it. Women that suffer from stress, have menopausal problems or take the pill are more likely to have tested for cervical cancer. Being in the target population for public smear test programmes is effective in reducing the probability of never taking the test. There is empirical evidence that the demand for smear tests increased in 2011/12 compared with 2006/07, for all age groups except women aged 50–69. The regional level variance u2 is significant, even after controlling for regional public screening programmes. The addition of regional public health policies does not reduce the unexplained variability among regions. Regional variance is not significant for age population groups, except for women aged 50–69 (Table 4).
Finally, we selected those women who have screened for breast cancer at least once and analysed the determinants for breast screening every 2 years or more. Similarly, we selected those women who have tested for cervical cancer and we analysed the determinants for smear testing every 5 years or more. Determinants of screening with adequate frequency were similar to those obtained for the decision on whether to screen or not. Younger women are the most likely to undertake breast screening, older women are the most likely to take smear tests, reaching a peak for women aged 50–69. In comparison with married women, widows are less likely to screen or test and single women are less likely to take a smear test. Lack of physical exercise and self-reporting of health problems are negatively correlated with regular breast screening whilst consuming ecological products and having private health insurance are positively correlated with regular smear tests. Living in a region which guarantees 100% coverage for breast screening is effective in encouraging women to take
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
8
Table 4 Determinants of cervical screening decisions: logistic multilevel regression results and odds ratios. Dependent variable: SmearTest. Model 1
Age Age: 16–29a Age: 30–49 Age: 50–69 Age: 70–older Marrieda Single Widowed Divorced Primary/No formal educationa Secondary education Tertiary education Employeda Retired Unemployed Homemaker Other Overweight Daily smoker Daily drinker Sedentary lifestyle Daily fruit Eco-goods Private heath insurance Family doctor Hormone products Contraceptive pill Health problems Stress Unhappiness Menopause problems TargetBreastTest TargetSmearTest TotalCoverageBreastTest PopulationBasedSmearTest Year 2006/07 Year 2011/12a Random effect: u2 LR test (Prob > 2 ): a * ** ***
Model 2
Total sample
Age: 16–29
Age: 30–49
Age: 50–69
Age: 70–older
Total sample
– – 1.90*** 1.25 0.34*** – 0.24*** 0.77** 1.12 – 1.80*** 2.47*** – 0.50*** 0.78 0.65*** 0.34*** 0.92 1.32** 0.83* 0.74*** 1.50*** 1.82*** 1.69*** 0.78* 1.60 2.76*** 1.09 1.19* 1.06 2.03*** – – – – 0.52***
1.23*** – – – – – 0.46** (omitted) 0.20 – 1.05 1.49 – (omitted) 0.66 0.47* 0.36*** 0.77 1.39 1.29 0.93 1.89*** 3.48* 1.70 0.60 (omitted) 3.52*** 0.91 1.02 1.63 6.14 – – – – 0.20**
1.04*** – – – – – 0.25*** 0.29** 1.49 – 1.93*** 2.13*** – 1.23 0.74 0.71 0.70 0.93 1.13 0.63** 0.89 1.30 1.46 2.95*** 0.97 1.43 1.66 1.07 1.93*** 0.76 1.10 – – – – 0.45**
0.94*** – – – – – 0.29*** 0.99 0.66 – 2.11*** 2.64*** – 0.89 1.25 0.85 2.02 0.81 0.91 0.86 0.72** 1.51** 2.00** 1.48 0.79 2.24 8.72 0.96 1.13 1.15 2.16*** – – – – 1.22
0.93*** – – – – – 0.28*** 1.02 6.52*** – 1.69** 2.20** – 0.00 (omitted) 0.00 0.00 0.87 3.16** 0.86 0.62*** 1.48 1.53 1.15 0.69 0.44 3.52 1.30 0.92 1.02 2.25*** – – – – 0.30***
– – 1.50** 0.77 0.27*** – 0.25*** 0.83 1.12 – 1.81*** 2.45*** – 0.50*** 0.80 0.65*** 0.35*** 0.90 1.34** 0.82* 0.76*** 1.51*** 1.80*** 1.69*** 0.80* 1.62 2.82*** 1.09 1.19* 1.02 1.99*** 1.24 1.70*** 0.93 0.71 0.54***
0.37 0.00
0.18 0.35
0.36 0.83
0.49 0.00
0.35 0.68
0.39 0.00
Denotes reference variable. We have also included an intercept. Significance level of 10%. Significance level of 5%. Significance level of 1%.
regular checks. Living in a region where the public smear test programme is population based may be also be effective in encouraging women to smear test regularly, but this result was not statistically significant. Previous results suggested that in 2011/12 (compared to 2006/07) women were more likely to have undergone breast screening at least once, although an analysis of screening frequency, shows that this does not mean that they are screening with adequate frequency. The regional level variance, u2 , is meaningful and significant for both breast screening and smear testing (Table 5). 6. Conclusions and policy implications This, Spanish-based, research validates international evidence from other European countries and North America. Results demonstrate that educational levels have a positive effect on the decision to screen. Women who adopt healthy habits and avoid risk behaviours are also more
likely to test for breast cancer and cervical cancer. Having private health insurance makes women more likely to screen and access to a public health care professional has the same effect. There were significant socioeconomic health disparities and this makes public health screening strategies especially important. It should be noted that there are significant regional variances that remain unexplained even after controlling for regional public health policies. Results further underline the fact that public health screening programmes improve individual screening decisions (more women screen). Being in the target population for public smear test programmes improves screening decisions for both breast cancer and cervical cancer. It is clear that public health programmes are positive in direct and indirect ways. Finally, even if dummy time variables reveal that the rates of screening for breast and cervical cancer are increasing, there are doubts that women are attending these screenings more regularly. It is not only important to
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model
ARTICLE IN PRESS
HEAP-3218; No. of Pages 10
A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx Table 5 Determinants of screening frequency decisions for breast cancer and cervical cancer: logistic multilevel regression results and odds ratios.
a
Age: 16–29 Age: 30–49 Age: 50–69 Age: 70–older Marrieda Single Widowed Divorced Primary/No formal educationa Secondary education Tertiary education Employeda Retired Unemployed Homemaker Other Overweight Daily smoker Daily drinker Sedentary lifestyle Daily fruit Eco-goods Private health insurance Family doctor Hormone products Contraceptive pill Health problems Stress Unhappiness Menopause problems TargetBreastTest TargetSmearTest TotalCoverageBreastTest PopulationBasedSmearTest Year 2006/07 Year 2011/12a Random effect u2 LR test (Prob > 2 ): a * ** ***
BreastTestRight
SmearTestRight
– 0.28*** 0.36** 0.10*** – 0.98 0.70* 0.84 – 0.79 0.92 – 1.17 2.24** 0.96 3.53*** 1.04 1.00 0.98 0.76** 1.05 0.82 1.17 0.70 1.32 0.65 0.74** 1.15 0.79 0.97 1.05 0.77 1.70** 0.64 6.47*** – 0.23 0.06
– 1.88*** 2.31*** 1.78** – 0.80* 0.51*** 0.85 – 1.16 1.18 – 1.14 0.98 0.82 1.25 0.94 1.01 1.00 1.05 0.98 0.95 1.87*** 1.05 1.65 0.99 0.89 1.10 0.83* 0.89 1.03 1.08 0.59** 1.20 1.00 – 0.30 0.00
Denotes reference variable. We have also included an intercept. Significance level of 10%. Significance level of 5%. Significance level of 1%.
9
individual might be able to change and improve (for example, lifestyles) [36]. Future research could examine differences in age groups and the effect of regular screening over time. Age is the most common risk factor for breast cancer - approximately one-third of all breast cancers occur in women of 70 or older. Although women over 70 years old are not a target population group for breast cancer, screening for breast cancer among older women has increased if we compare 2011/12 with 2006/07. There is empirical evidence that public screening programmes and treatment protocols have helped to reduce cancer mortality rates in developed countries [9]. Further national studies are important in order to promote the implementation of screening programmes in developing countries but also to rank and select the most cost-effective screening strategies in those countries with a long history of public screening. The present work is not free of technical problems. Among the main limitations is the use of two cross-sections for the analysis of micro-data. As previously stated, for reasons of data comparability, it was not possible to use previous Spanish National Health Surveys, which would have, for example, allowed a micro–macro analysis. In addition, the use of a panel data would have reduced biases derived from individual heterogeneity. Although we feel comfortable that our results provide the right correlation effects, causality should be analysed with caution. Acknowledgements This research was completed as part of the Spanish Ministry of Innovation and Science CSO2011-30089 Health Policies research project: ‘Sexual and Reproductive Behaviour in Spain: Risk Factors for Health’. The authors are grateful to the reviewers and editor whose comments have contributed to improving the quality of the work. References
test for breast and cervical cancer, it is necessary to do it regularly. The results are consistent with other professional recommendations and findings on healthy lifestyles (regular exercise, a balanced diet, low alcohol and tobacco consumption and not using hormone replacement therapy during the menopause) and participation in screening programmes [4,31]. Once the correct level of attention is paid to individual behaviour, the question of socio-economic inequalities should be addressed. Even in countries with widespread testing programmes, socioeconomic inequalities in health care are important. Economically advantaged women are more likely to seek treatment by a specialist and this is particularly true if they have private insurance. In Spain, long waiting lists are the main reason why women take out private health insurance [32]. Health policies should target women with low socioeconomic levels [20,33–35]. Inequity in health and health care access should not be restricted to the analysis of individual characteristics; it should also consider factors that the
[1] Howlader N, Noone AM, Krapcho M, Garshell J, Neyman N, Altekruse SF, et al. SEER cancer statistics review 1975–2010. Bethesda, MD: National Cancer Institute; 2013. [2] Wübker A. Explaining variations in breast cancer screening across European countries. European Journal of Health Economics 2013, http://dx.doi.org/10.1007/s10198-013-0490-3. [3] Spanish Society of Medical Oncology. Info cancer; 2013. Available from: http://www.seom.org [accessed 10.07.13]. [4] Spanish Association against Cancer. Sobre el cancer; 2012. Available from: https://www.aecc.es/SobreElCancer [accessed 10.07.13]. [5] DeMay M. Practical principles of cytopathology, revised ed. Chicago, IL: American Society for Clinical Pathology Press; 2007. [6] Kopans DB. Why the critics of screening mammography are wrong. Diagnostic Imaging 2009;31(12):18–24. [7] Mulcahy N. Screening mammography benefits and harms in spotlight again. Medscape 2009;94(6):411–3. [8] de Koning HJ, van Ineveld BM, van Oortmarssen GJ, de Haes JC, Collette HJA, Hendriks J, et al. Breast cancer screening and costeffectiveness; policy alternatives, quality of life considerations and the possible impact of uncertain factors. International Journal of Cancer 1991;49(4):531–7. [9] Gøtzsche PC, Nielsen M. Screening for breast cancer with mammography. Cochrane Database of Systematic Reviews 2011;(1). Available from: CD001877. [10] U.S. Preventive Services Task Force. Breast cancer: screening. Rockville, MD: United States Preventive Services Task Force; 2009.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
G Model HEAP-3218; No. of Pages 10
10
ARTICLE IN PRESS A.I. Gil Lacruz et al. / Health Policy xxx (2014) xxx–xxx
[11] The European Cancer Observatory; 2011. Available from: http://eco.iarc.fr [accessed 10.07.13]. [12] Carrieri V, Wübker A. Assessing inequalities in preventive care use in Europe: a special case of health-care inequalities? Health Policy 2013;113:247–57. [13] de Geldera Bulliard J, de Wolf C, Fracheboud J, Draisma G, Schopperd D, de Koning J. Cost-effectiveness of opportunistic versus organised mammography screening in Switzerland. European Journal of Cancer 2009;45:127–38. [14] Catalá-López F, García-Altés A, Álvarez-Martín E, Gènova-Maleras R, Morant-Ginestar C, Arana E. Evaluación económica de interven˜ revisión sistemática ciones en enfermedades oncológicas en Espana: y análisis comparativo. Farmacia Hospitalaria 2012;36(3):141–7. [15] Spanish National Survey on Health (2006/07 and 2011/12). Data available from: http://www.msc.es/estadEstudios/estadisticas/ encuestaNacional/ [accessed 14.10.13]. [16] Mandelblatt J, Andrews H, Kemer J, Zauber A, Bumetu W. Determinants of late stage diagnosis of breast and cervical cancer: the impact of age, race, social class and hospital type. American Journal of Public Health 1991;5:646–9. [17] Blackburn GL, Wang KA. Dietary fat reduction and breast cancer outcome: results from the Women’s Intervention Nutrition Study (WINS). American Journal of Clinical Nutrition 2007;86(3):878–81. [18] Boffetta P, Hashibe M, La Vecchia C, Zatonski W, Rehm JA. The burden of cancer attributable to alcohol drinking. International Journal of Cancer 2006;119(4):884–7. [19] Lairson D, Chan W, Newmark GR. Determinants of the demand for breast cancer screening among women veterans in the United States. Social Science and Medicine 2005;61:1608–17. [20] Sutton S, Rutherford C. Sociodemographic and attitudinal correlates of cervical screening uptake in a national sample of women in Britain. Social Science and Medicine 2005;61:2460–5. [21] Whynes D, Philips Z, Avis M. Why do women participate in the English cervical cancer screening programme? Journal of Health Economics 2007;26:306–25. [22] Carney P, O’Neill S, O’Neill C. Determinants of breast cancer screening uptake in women, evidence from the British household panel study. Social Science and Medicine 2013;82:108–14. [23] Luengo-Matos S, Polos-Santos M, Saz-Parkinson Z. Mammography use and factors associated with its use after the introduction of breast cancer screening programmes in Spain. European Journal of Cancer Prevention 2006;15(3):242–8. [24] Cabeza E, Esteva M, Pujol A, Thomas V, Sánchez-Contador C. Social disparities in breast and cervical cancer preventive practices. European Journal of Cancer Prevention 2007;16(4):372–9.
[25] Martín-López R, Hernández-Barrera V, López-Andrés A, GarridoCarrasco P, Gil-Miguel A, Jiménez-García R. Breast and cervical cancer screening in Spain and predictors of adherence. European Journal of Cancer Prevention 2010;19(3):239–45. [26] Puig-Tintoré L, Castellsagué X, Torné A, Sanjosé S, Cortés J, Roura E, et al. Coverage and factors associated with cervical cancer screening: results from the AFRODITA study: a population-based survey in Spain. Journal of Lower Genital Tract Disease 2008;12(2): 82–9. [27] Johnson-Thompson MC, Guthrie J. Ongoing research to identify environmental risk factors in breast carcinoma. Cancer 2000;88(5):1224–9. [28] Coughlin SS, Leadbetter S, Richards T, Sabatino SA. Contextual analysis of breast and cervical cancer screening and factors associated with health care access among United States women, 2002. Social Science and Medicine 2008;66:260–75. ˜ [29] Ascunce N, Salas D, Zubizarreta R, Almazán R, Ibánez J, Ederra M. Cancer screening in Spain. Annals of Oncology 2010;21: 43–51. [30] Ascunce N, Delfrade J, Salas D, Zubizarreta R, Ederra M. Programas ˜ características y de detección precoz de cáncer de mama en Espana: principales resultados. Medicina Clínica 2013;141:13–23. [31] Sala M, Queiro T, Castells X, Salas D. Situación del cribado de cáncer ˜ In: Castells X, Sala M, Ascunce N, Salas D, de cérvix en Espana. Zubizarreta R, Casamitjana M, editors. Descripción del cribado del ˜ cáncer en Espana Proyecto DESCRIC. Madrid: Ministerio de Sanidad y Consumo; 2007. [32] Spanish Ministry of Health, Social Policy. Estrategia en Cáncer del Sistema Nacional de Salud. Madrid: Spanish Ministry of Health and Social Policy; 2010. [33] Pinilla J, González B, Barber P, Santana Y. Smoking in young adolescents: an approach with multilevel discrete choice models. Journal of Epidemiological Community Health 2002;56: 227–32. [34] Szumilas M. Explaining odds ratios. Journal of the Canadian Academy of Child and Adolescent Psychiatry 2010;19:227–9. [35] Paléncia L, Espelt A, Rodríguez-Sanz M, Puigignós R, Pons-Vigués M, Pasarín MI, et al. Socio-economic inequalities in breast and cervical cancer screening practices in Europe: influence of the type of screening program. International Journal of Epidemiology 2010;39: 757–63. [36] Doescher M, Jackson JE. Trends in cervical and breast cancer screening practices among women in rural and urban areas of the United States. Journal of Public Health Management and Practice 2009;15(3):200–9.
Please cite this article in press as: Gil Lacruz AI, et al. Female preventive practices: Breast and smear tests. Health Policy (2014), http://dx.doi.org/10.1016/j.healthpol.2014.04.012
