Investing in early detection versus intensive treatment for breast cancer: a study of the Israeli public priorities

Abstract

Objective Public opinion has become one of the main inputs in setting priorities, rationing and allocating health resources. The present study focuses on the priorities of the Israeli public in allocating marginal breast cancer funds between early detection in the healthy majority of the population and intensive treatment of the sick minority in need.

Design A sample of 2030 individuals representing the Israeli Jewish urban population aged 45–75 was interviewed in 1993/4. A full sit‐down interview collected information on several health related issues.

Results Sixty‐one per cent chose to direct the funds to early detection efforts and 35% chose to direct the funds to treating the sick. Four per cent of the population could not decide. Higher education and better health are related to choosing early detection; religious observance is related to choosing the treatment option; and men and older persons tend, more than women and younger respondents, to be undecided.

Conclusions Whilst the majority of the population tend to follow a cost‐effectiveness rationale in the marginal use of breast cancer funds, for more than a third of the population efficiency is not thus important, and they adhere to entitlement based upon a need‐equity principle in allocating health resources.

Introduction

Given the presence of market failures due to incomplete information, externalities, and health being a merit or social good, societies are in a continuing search for indications – other than market price signals – of how to achieve an efficient and fair allocation of scarce health resources. Economic evaluations, such as cost‐effectiveness analysis, are useful indicators of efficiency. However, efficiency is not the only criteria with which to evaluate the functioning of a system. In many cases, particularly in the health care system, equity and fairness are no less important. However, whilst ‘efficiency’ is clearly defined, ‘equity’ is a highly subjective criterion that reflects values, ethical considerations and social norms. From both an efficiency and an equity point of view, a natural indicator of the desired allocation of resources is the public, who is served by and pays for the system. Although ‘asking the public’ has its problems in terms of knowledge and consistency, there is a general consensus that it should have a major input into the allocation process. ^{1 , 2}

The present study examines the views of the Israeli public on the desired (marginal) allocation of resources towards breast cancer detection and treatment. Israelis aged 45–75 were asked whether the government should spend supplementary funds on early screening and detection efforts or on treatment and care for patients with advanced stages of the disease. Quite coincidentally and for another purpose, Eddy ³ has provided a simplified clinical‐economic description of the two options using US data. Breast cancer screening for 1000 50‐year‐old women in average health from age 50–65 would prevent about seven breast cancer deaths, add about 120 person‐years of life, and cost about $1.2 million. Alternatively, providing high‐dose chemotherapy with autologous bone marrow transplantation (HDC‐ABMT) for women who develop the disease would prevent about one death, add about 7 person‐years of life, and cost about $1.5 million.

It is clear (even without the detailed clinical‐economic data) that as long as certain screening standards have not been met, resources invested in early detection are used more efficiently than those invested in treatment. However, once equity and values are considered, the choice becomes much more complex.

The issues of equity, ethics, the ‘just’ allocation of resources, setting of priorities and rationing in health care have been the subject of considerable discussion by economists, physicians, philosophers and social policy scholars. ^{2 , 4 , 5} The present choice between screening of the healthy and treatment of the sick requires weighting the perceived social value of each. Screening for the healthy majority constitutes a more efficient use of resources. However, there is a considerable perceived social value of directing care to the minority in need (‘of rescue’ ⁶ ), even though the benefits in terms of survival and quality of life might be minimal.

Methods

The survey

Between October 1993 and February 1994, the Gertner Institute conducted a national survey on health issues. A sample of 2030 individuals was chosen to represent a population of about 800 000 urban Jewish Israelis aged 45–75. A proportionate stratified sample was used, in which the stratifying variable was settlement size (above 200 000; 100 000–200 000; 50 000–100 000; and 5000–50 000 inhabitants). Specific regions within big settlements and certain settlements within the small settlement strata were chosen to represent each stratum with respect to socio‐economic status, geographical distribution and year of establishment. Within the regions and the settlements chosen, the addresses (and the replacement addresses) were chosen according to a procedure developed and used by the institute to assure the representability of the sample. Data were collected in face‐to‐face full sit‐down interviews covering several areas relevant to health.

The variables

Priority choice. The question was worded as follows: ‘The Department of Health has received a donation from abroad to be used in relation to breast cancer. Two alternative uses are considered: (a) to designate the fund to expand screening and early detection efforts; (b) to use the funds to offer more expensive curative treatments and care for patients with advanced breast cancer. Which option should the DoH choose?’. A third category (which was not presented to the interviewees) was a ‘don’t know’ category.

Other characteristics. Age is defined by three age groups: 45–54, 55–64 and 65–75. The 55–64 age group serves as the base category in the analysis. Gender is indicated by a binary variable, with men designated by the value of 1. Three groups measure level of education: 0–8, 9–12 and 13+ years of schooling. The 9–12 category is the base category. Religious observance is indicated by a binary variable with 1 denoting observance. Ethnic origin (continent of birth; and for respondents who are Israeli born that of his or her father) is defined by a two‐category variable: Asia‐Africa and Europe‐America‐Israel. General health is measured using a Hebrew translation of the SF‐36. ⁷ The scale used here is the average of the eight scales (Alpha reliability = 0.89). This scale is recoded into a binary variable where 1 indicates an above‐average score. Finally, included in the analysis is a variable indicating whether the individual suffers from cancer (of any type).

The statistical strategy

Bivariate analysis was used to explore the relationships (Chi‐squared tests were used to test for independence) between public choice and several personal characteristics. For the multivariate analysis, since the preferences were indicated by a three‐category variable (i.e. early detection, intensive treatments, and ‘don’t know’), a Multinomial Logit Model ⁸ was used. This model allows the impact of explanatory variables on the probability of choice of how funds should be spent to be explored. Odds ratios (OR) were used to measure the impact of the explanatory variables on the choices.

Results

A profile of the survey population is presented in Table 1. This profile matched the population distribution in terms of gender, age, education and ethnic origin.

Table 1.

 A profile of the survey population (n = 2030)

Sixty‐one per cent of the survey population chose to use the funds for more intensive screening and early detection efforts. Thirty‐five per cent supported the alternative, namely, to devote the funds to curative treatments and care for patients with advanced breast cancer. Four per cent could not decide (answering ‘don’t know’). Over two thirds of those who could not decide specifically suggested splitting the funds equally.

Table 2 reports the bivariate analysis relating the public choice to several personal characteristics. Age, gender, level of education and religiosity were each significantly related to the choice between detection, treatment and ‘don’t know’. General health was only marginally related to that choice, and having cancer or ethnic origin were not related to the choice at all. Older persons tended to favour intensive treatments and care for the sick rather than early detection for the healthy. Older persons were more likely than those in younger age groups to report ‘don’t know’. Whilst about 61% of both men and women chose the early detection option, 36% of women and 33% of men chose the treatment option. Men were more likely than women to consider the issue irrelevant (‘don’t know’). Higher levels of education were clearly associated with a tendency to choose the detection option, and religious persons tended to choose the treatment option. Persons with above‐average general health tended to choose the detection option. There was no difference in choice patterns between persons who suffered from any type of cancer and those who did not.

Table 2.

 Bivariate associations of public choices

Table 3 presents the multinomial logit estimates of public choices. The results are reported for three log‐odds ratios: detection versus missing value, intensive treatment versus missing value, and detection versus treatment. The entries are the odds ratios (OR) and their 95% confidence intervals. The 55–64 age group is omitted and serves as the base category (OR = 1) for the effect of age on choice. Similarly, the 9–12 years of schooling category serves as the base category for education.

Table 3.

 A Multinomial Logit Model of public choices

The results reported in the last column show that the main determinants of the detection‐treatment choice were level of education, religiosity and general health. Respondents with 0–8 years of schooling were 28% more likely than those with 9–12 years of schooling (the base category) to choose the treatment option. There was no difference in the choices of people with 9–12 and 13+ years of schooling. Religious respondents were 31% more likely than non‐observants to choose the treatment option. Respondents with above‐average health were 11% less likely than people with below‐average health to choose the treatment option. The first two columns show that age and gender were the main determinants of the choice between the treatment and detection options on the one hand, and the ‘don’t know’ option on the other. With the 55–64 age group serving as the base category, persons in the 45–54 age group were half as likely to choose the ‘don’t know’ option whilst older people in the 65–75 age group were about twice as likely to be undecided. Men were more likely than women to choose the ‘don’t know’ option.

A separate analysis performed on women only (not reported for the sake of brevity), showed two main differences. Firstly, whilst the effect of education remains the same, the significant difference was between women with 0–12 and 13+ years of schooling (as opposed to the distinction between respondents with 0–8 and 9+ years of schooling in the general population). Secondly, women with cancer (2.4%) were markedly more likely to choose the ‘don’t know’ option over both the treatment and detection options.

Discussion

Social choices made by individuals are the result of a complex process shaped by knowledge, values and self‐interest. These factors and the way each affects the choice are not observable. However, inferences can be made using observable traits, which have some intuitive relation to the above factors.

Higher education is clearly related to choosing the detection option over treatment. This relationship seems to reflect two factors. The first is knowledge. Higher education renders the (existing) information on the benefits of preventive measures in general, and breast cancer screening in particular, more accessible. The second factor is related to differences in time preferences. As Fuchs ⁹ argues, people pursue higher education (as an investment in their human capital) partly because they attach a relatively high weight to the future. Such an investment‐minded attitude applies to health as well, leading people with higher education to take preventive measures as an investment in (future) health. People with a stronger present‐preference are less likely to invest either in their human capital or in their health. These people attach greater weight to the ‘present sick’ than to the (preventable) ‘future sick’, leading them to choose the treatment option over the detection one. The same two factors of knowledge and time‐preference may account for the finding that healthier people, like those who are more educated, tend to choose the detection option.

Religious people appear to hold a more deterministic view of the world and feel more obliged to offer help to the needy. This may explain their tendency to choose the treatment option over the detection one.

There has been some research trying to relate public opinion and choice to self‐ and group‐interest motives. ¹⁰ The conclusions of these studies are mixed. In the present case, the small number of women who suffered from cancer (of any type) did not favour the treatment option, but chose the ‘don’t know’ one. This is probably because those women cannot consider the issue as a hypothetical question. Women, who do not suffer from cancer, however, tend to choose the early detection option. The marginal benefit of directing the funds towards detection efforts, making screening more available and accessible (e.g. using personal invitations), is obviously greater for those who can actually use the service. These people are healthier and more educated. On the other hand, the choice between detection and treatment of breast cancer is irrelevant for men and less relevant for older women. The results show that men and those in the 65+ age group are more likely to choose the ‘don’t know’ option than women and younger people.

The choices made refer to the marginal allocation of funds. At the margin, the Israeli public would like to see about 2/3 of the funds directed to early detection efforts in the healthy population, and about 1/3 to intensive treatment of the sick. The choices are shaped by the actual situation in Israel with respect to breast cancer. In general, the situation in Israel is not different from that in North America and Western Europe with respect to both the epidemiology of the disease and the recommendations about early detection. ¹¹ , 12, ^–13 In 1985, the morbidity rate in Israel was 137 per 100 000 women aged 30–74, and in 1989, the mortality rate was 36 per 100 000 women aged 15–75. At the time of the survey, mammography tests were covered by the health insurance policies of all sick funds for women over 40. It is interesting to note that there is a dramatic difference in morbidity rate by ethnic origin. In 1992, the age‐adjusted morbidity rates were 91/100 000 amongst women of Euro‐American‐Israeli origin, and 53/100 000 amongst those of Asian‐African origin. In spite of this epidemiological difference, there was no difference in the choices of the two ethnic groups.

The restriction of the study to those aged 45–75 probably underestimates the importance to the wider population of early detection, since younger and more educated individuals tend to prefer the detection option. In the general population, that option may expected to be even more popular.

The present findings contribute to the growing body of evidence from around the world that the public can provide some input to the decision making process regarding the allocation of scarce national health resources. Routine retrieval of these views, through focus groups, public meetings and surveys, may provide policy makers with data needed for making allocation decisions more efficient and fair.

Acknowledgements

This paper was completed whilst I was visiting Yale University, Departments of Economics and Public Health. I thank both for their hospitality. The contribution of Nira Shamai, Noah Lewin‐Epstein, and Tami Sagiv‐Shifter to the formulation of the questionnaire and to the fieldwork is gratefully acknowledged.