Explaining variations in breast cancer screening across European countries

Moreover, critics argue that problems associated with overdiagnosis and false positives might not be communicated adequately by public health communicators and physicians. For example screening invitations are not balanced regarding harms and benefits (compare Gøtzsche and Nielsen [10] or Gummersbach et al. [11]) and physicians often do not adequately communicate possible harms of screening (compare Wegwarth and Gigerenzer [12]). Furthermore, awareness of women regarding potential overdetection of indolent breast cancer is minimal [13]. Compare for the definition of overdiagnosis and false positives [8].

Note that physicians might depart from perfect agency especially if the financial incentives they face deviate from the interests of the patient (see McGuire [19] for a critical discussion). Moreover, with regard to risk communication in clinical practice in the context of cancer, physicians might have limited expertise to help patients make informed medical decisions as discussed by Schwartz et al. [26] or Wegwarth and Gigerenzer [12].

In the EU-member states the characteristics of organised screening programmes are widely generalisable [37].

Opportunistic screening happens when someone asks their doctor or health professional for a mammogram.

The SHARE used computer-assisted personal interviewing methods among participants drawn from probability samples in all contributing countries [38]. The sampling procedure followed a complex probabilistic multistage design to produce estimates representative of the non-institutionalised population above the age of 50 years in each country [38]. The SHARE overall response rate in 2004 was 61.8 per cent, being lowest in Switzerland (38 per cent) and highest in France (74 per cent). Compare Börsch-Supan and Jürges [39] for a comprehensive description of the methodology used.

I assumed that a woman had a screening regularly in 2004 (wave 1) and in 2006 (wave 2) if she stated in the SHARELIFE (wave 3 of the SHARE) that she always had a mammogram regularly at least every 2 years before 2007. I decided not to use the third wave of the SHARE, because many explanatory variables that are included in wave 1 and wave 2 are not included in wave 3. To account for repeated observations I calculated robust standard errors that considered clustering at the individual level. As a robustness test, I calculated all regressions using the last wave of SHARE data. The results were substantially unchanged and are available upon request.

For example better education may increase the use of screening services, implying more efficiency in producing health (e.g. [17]). A better educated woman may be more likely to understand the benefits of mammography screening. In addition, these women may be more prone to recognise the early warning signs of breast cancer and be more apt to visit a physician when symptoms first occur. Wübker [6] provides detailed reasoning, based on economic theory (i.e. human capital models), for including the above-mentioned variables. Moreover he tests them empirically using the SHARE database.

Autier et al. [45], p. 1188, argued that “the total radiologists registered in a country could represent a reasonable approximation to those specialising in mammography”.

For example a fee for service payment system might be associated with higher supply of screening, since doctors have financial incentives to increase the volume of screening and theory and empirical evidence show that physicians respond to incentives [23]. In contrast gatekeeping is expected to lead to less use of mammogram screening since women might see a general physician before going to the specialist (who is often provider of mammograms). A higher demand for breast cancer screening as measured by incidence rates should be associated with more screening activity.

SHARELIFE is wave 3 of the SHARE and focusses on people's life histories and contains detailed information on historical mammography screening use. The question in SHARELIFE is “What are the reasons you [have never had/stopped having] mammograms regularly?” It provides the following possible answers: (1) not affordable, (2) not covered by health insurance, (3) did not have health insurance, (4) time constraints, (5) not enough information about this type of care, (6) not usual to get this type of care, (7) no place to receive this type of care close to home and (8) not considered to be necessary. Compare Table A1 in the “Appendix” for the detailed definition of variables.

By assumption both error terms have zero mean and constant variances (σ ², σ _ε ² ).

So far I assumed that the coefficients of the individual variables are identical for each country. However the impact of these variables might depend on the institutional framework within a country. For example, whether income has an impact on the individual screening decision depends on whether a woman has to pay for screening.

However, it might also reflect that invitation leaflets are not balanced with regards to benefits and harms of screening (compare [10, 11]). Gummersbach et al. [11] found that in different European counties (Germany, France, Spain and Italy) the information on side effects and risks provided by the brochures was generally of poor quality and none of them referred to the problem of overdiagnosis.

Considerable differences might exist in the scientific and public debate about screening benefits and harms across countries. For example, in Sweden leading professionals are positive about screening and two major RCTs were undertaken in there [5]. In contrast, influential Cochrane collaborators in Denmark [10] have heavily criticised the randomised trials, not least one of the Swedish trials. These country differences in the public debates and scientific disagreement on the harms and benefits of screening might influence both the differences in the availability of screening programmes as well as the differences in screening rates between Sweden and Denmark. However, I control for these unobservable impact factors within my sensitivity analysis using country-fixed effects. Moreover, I exploit regional differences in the availability of organised screening programmes (i.e. Copenhagen versus other Danish regions) to identify the impact of organised screening on screening uptake. Thus, only if public debates and scientific disagreement on the effectiveness of screening differ also within countries and are correlated with regional availability of organised screening programmes might the results be driven by these differences.