What is important in a fight for life? – preferences of women diagnosed with breast cancer for attributes of an oncological institute

Abstract

A Discrete Choice Experiment (DCE) is used to establish preference patterns for attributes of a medical institute that are exhibited by Israeli women who were diagnosed with breast cancer. It contributes and adds to the growing body of studies in Health Economics that use DCEs: (i) the domain of experimentation is extended to a country other than the UK or the US, where most experiments took place; (ii) a life-threatening disease has been chosen for the DCE, to test whether in this case too individuals relate to trade-offs between attributes, or alternatively focus only on quality of treatment; (iii) substitution or complementarities between attributes have been investigated; and (iv) differences in preference patterns within diverse socio-economic groups have been explored. The findings are followed by suggestions for possible extensions.

Notes

¹ In the public health-care sector revealed preferences market data are not readily available. Hospitalization data (e.g. mortality rates, number of staff per patient) are used in some studies as revealed preferences data. Hospitalization data do not include information on qualitative attributes (e.g. attitude of staff, information, hospitalization facilities) and are also subject to various methodological and statistical problems.

² Ben-Akiva and Lerman (1985) and Louviere et al. (2000) provide a comprehensive overview of the design and analysis of DCEs.

³ The very few women who failed the ‘internal consistency’ test were excluded from the sample.

⁴ ‘Consumerism’ is basically searching for information. In this study it is detected by having a second opinion and/or visiting several institutes before choosing the one patient currently treated in (Lloyd et al., 1991).

⁵ Israel has a public health-care system. Workers pay a Health Insurance Tax of 4.8% of their income (up to a ceiling of about US $500). Health services are basically free. Few private hospitals also operate, but the great majority of the population is treated in public hospitals.

⁶ However, it is more costly and therefore most DCEs that are reported in the literature used self-complete questionnaires that are posted or handed out and returned.

⁷ The average age of women in our sample is 52.3 years (range of 27–76 years). Over 50% have (at least partial) academic education. Women in the sample are more educated than the general Israeli population in the relevant age group. Figures of the other socio-economic variables are similar to those of the Israeli population in the same age group (not reported due to space constraints).

⁸ Women with dominant preferences were identified using three complimentary tests. The first is ‘unique’ and has not been employed in the literature before. A woman was defined as having a dominant preference for reputation of doctor if: (i) in a pair of scenarios where one scenario had the higher level of doctor's reputation but lower levels of ‘all’ other attributes – she still preferred the first one; (ii) she ‘always’ chose the scenario with the higher level of doctor's reputation. This test is weaker compared to the first one; and (iii) in an open question in which respondents were asked to rank the five characteristics (from first to fifth), ‘reputation of doctor’ was ranked first. A respondent who ‘passed’ these three tests has been declared with a dominant preference for ‘reputation of doctor’. Similarly, we have identified women with dominant preferences for other attributes. For those we did not have the first test described earlier and used the other two tests. Out of 44 women with dominant preferences (46% of the sample), the majority (25) had a dominant preference for ‘reputation of doctor’. Some (15) had a dominant preference for ‘professionalism of institute’ and few (4) for ‘attitude’. Dominant preferences weaken the theoretical basis of DCE that assumes that individuals employ compensatory decision-making, thus trading each attribute for the other.

⁹ Our findings are in line with other studies that indicated that the inclusion of respondents with dominant preferences in a DCE is not terribly disturbing the results (e.g. Mazzotta and Opaluch, 1995).

¹⁰ There is some discussion on this issue within the medical literature, employing techniques other than DCE (e.g. time-trade-offs). See Jansen et al. (2000) that also includes a review of the medical literature.