Epistemic and non-epistemic values in economic evaluations of public health

ABSTRACT

We review methods for economic evaluation recently developed in health economics by focusing on the epistemic and non-epistemic values they embody. The emphasis is on insights into valuing health, health states or resource management, against conventional approaches using QALY and DALY. The main hypothesis is that economic evaluation and related allocations of economic resources for health and healthcare could be improved by jointly adopting robust methods and broader normative theories within contemporary liberal-egalitarianism. Here, Amartya Sen’s Capability Approach (CA), which shifts attention away from individuals’ preferences for different health, healthcare options and utility maximisation, thereby provide more accurate evaluations of people’s states of health and their objective needs. This combination establishes a beneficial interplay between crucial epistemic values (objectivity-impartiality, consistency, evidence) and non-epistemic values (fairness-equity, redistributive justice). Thus, it is expected to amend methods and normative concepts for economic evaluations of public health in a genuine ‘extra-welfarist’ perspective.

KEYWORDS:

• Economic evaluation
• robust methods
• capability approach
• health equity
• public health
• extra-welfarism

Acknowledgements

We are grateful for the efforts of editors John Davis and Wade Hands throughout the entire review process. Likewise, we thank two anonymous referees for their constructive comments, useful inputs and suggestions on the earlier version of the paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Alessandra Cenci is a Research Fellow in the department of Philosophy of the University of Southern Denmark. Her main interests are in Philosophy of Economics, Philosophy of the Social Sciences, Social epistemology, particularly, critical epistemological-practical operational issues raised by the value-free ideal in economics and in social research. She is currently carried out an own-designed project titled: Values and Knowledge in Public Health Analysis and Policy.

M. Azhar Hussain is an Associate Professor at the University of Sharjah (UAE) and Roskilde University (Denmark). His main research interests are in empirical analyses of welfare distribution, income inequality, poverty, deprivation, polarisation and robust multidimensional rankings. He primarily focuses on Denmark, the EU and developing countries. His national and international publications appear in journals, books and newspaper articles.

Notes

1 We adopt the definition of the World Health Organization. See http://www.who.int/healthsystems/topics/equity/en/

2 The terms objective and impartial are used interchangeably since refer to valuable epistemic and normative ideals and a similar attitude regarding desirable properties of reasoning, inferences, acceptable policy claims. While the first term is mostly used by philosophers of science and social scientists; the second one is most commonly used in political and ethical theory.

3 For an overview of the philosophical debate concerning values in scientific research and science-related policy making, see: (Longino, 1990; Kitcher, 2001; Douglas, 2009).

4 This approach, known as ‘liberal welfarism’, has strongly been criticised by Hausman (2015) since evaluations of health, health states are grounded on what he called the ‘private value of health’. That is, a reductive view of health that disregards its social importance. He alternatively proposes to value health, health states (and building public policy) according to the ‘public value of health’ (Ch. 12–13–14) which could better represent the perspective of a ‘facilitator liberal state’.

5 In DALY, the 0–1 scale is inverted because QALY is a measure of health benefit (1 is ‘perfect health’) while DALY is a measure of disease burden (0 is absence of disability), see (Bognar, 2010, pp. 394–395).

6 For an overview of the flaws of standard utilitarian evaluation and subsequent metrics of WB, QL see (Sen, 1979; Sen & Williams, 1982; Sen, 1982; Hausman, 2012, 2015).

7 On the inadequacy of the value-free ideal in positive economics, see (Reiss, 2017).

8 For substantial critiques of QALY/DALY as methods applying CUA/CEA see (Nord, 1999, Chapter 2–3; Anand et al., 2004; Hausman, 2015, chapter 15–17).

9 Dilemmas for fairness raised by the fair innings argument (also in connection to QALY) have been discussed by (Rivlin, 2000; Anand, 2005; Nord, 2005).

10 For an overview of current debates see Williams, 1985; Chang, 1997; Broome, 1999; Nussbaum, 2003; and Thomson, 2003 cited in Oswald, 2015, pp. 38–40.

11 Against welfarism and particularly, the reductionism/simplicity of standard economics value-free assumptions i.e., the standard rationality, see (Sen, 1986).

12 About the importance of ‘non-utilitarian information’ for valuing welfare, interpersonal comparability and redistributive justice, see (Sen, 1982).

13 Another difficulty for obtaining accurate measurements concerns the reference to a standardised measure of life expectancy – that of Japan having the highest LE worldwide – for all countries.

14 Discussing the aptness of individuals’ preferences to represent people’s WB and whether is possible to meaningfully measure WB, health-related QL, health states etc. through survey-based methods eliciting preferences exceeds the scope of this paper analysis. For an overview of on-going debates, see (Hausman, 2012; 2015, Chapter 5–7–8; Hersch, 2015).

15 A review of empirical methods for priority setting in healthcare from an ‘empirical ethics’ perspective is offered by (Hasman, 2003).

16 On the epistemic import of robustness analysis in economic modelling see, Kuorikoski, Lehtinen, and Marchionni (2010). A taxonomy of varieties of robustness is offered by Woodward (2006). Discussing the insights of the specific kind of robustness analysis on which the FOD method is constructed upon exceeds the scope of this paper. We have maintained that FOD robustness is valuable for getting epistemically justified beliefs about results and analogously, for attaining robust evidential knowledge – further functionally based on a broader multidimensional informative basis. Thus, in FOD applications, errors can be reduced, and replicability of experimental results augmented. See Cenci & Hussain, 2019).

17 For an introduction and synthesis of different definitions, see Østerdal (2010) cited in Hussain et al., 2016, p. 744.

18 The practical applicability of the FOD method is not a primary issue in this paper since it has been widely established by several recently published FOD-based empirical studies in leading economic journals See: Arndt et. al. 2012, 2016; Hussain et al., 2016; Arndt, Mahrt, Hussain, & Tarp, 2018; Permanyer & Hussain, 2018; Hussain, Siersbæk, & Østerdal, 2019, etc.

19 For an overview of how standard approaches in health economics deal with multidimensionality, see Drummond 2005; Feeny et al. 2002; UHF 2012; Kohn 2012; HCP 2012; and WHO 2013 cited in Hussain et al., 2016, p. 740.

20 On the advantages for economic evaluation of complexity-plurality over reductionism-simplicity, see (Sen, 1986, p. 22).

21 For other robust measures see (Atkinson and Bourguignon, 1982; Atkinson, 1992; Bourguignon & Chakravarty, 2003; Gravel et al., 2009; Muller & Trannoy, 2011; and Gravel & Moyes, 2012 cited in Arndt et al., 2017).

22 This minimum ‘value requirement’ of FOD (‘more welfare is better than less’) offers a criterion to compare two given welfare distributions in order to establish which one is better. It establishes an important difference compared to standard economics measures (e.g., liberal welfarism) where the same assumption (accompanied with the identification of utility and welfare) has often been improperly used to derive operational strategies that in occasions, do not logically or do not necessarily follow from them. That is, to reduce the complexity-plurality of evaluations, to restrict the informative basis to utilitarian information and to establish the total commensurability of every aspect relevant for welfare in utility-monetary terms (Sen 1970, 1979, 1982; Anand et al., 2004; Hansson, 2007, etc.). This applies to multidimensional evaluations as well. For example, the weighting approach (i.e., assigning a numerical value to dimensions) is still based on ‘welfarist’ tenets like the commensurability-substitutability while, at an operational level, there is prevalent use of utilitarian information (elicited by means of individual preferences-based surveys). Conversely, in FOD, welfare and analogously, health or relative disadvantage are understood broadly and the use of objective data and non-utilitarian information is frequent (see this paper Section 5).

23 For a mathematical demonstration of the FOD procedure, see Hussain et al., 2016, pp. 744–5.

24 On the insights for fairness-equity of applying the method of pairwise comparison namely in the view of rejecting the pure consequentialist logic of standard economics measures, see this paper Section 5.

25 Firstly, FOD can produce indeterminate outcomes (i.e., distribution A does not dominate B, and B does not dominate A) namely FOD could not provide complete and transitive orderings – which constitute the standard requirement of rationality and rational choice. At an empirical level, it has been solved by applying a Copeland approach (1951 cited in Arndt et al., 2017, p. 7–9) which always provide a dominating-dominated ranking. Anyhow, recent advancements in decision theory demonstrate that partial agreements based on incomplete orderings of preferences are good enough for rationality since valuational unrankability is not a sign of irrationality but rather, a testimony of the structural diversity and plurality of choices in real contexts (Sen, 2018). It entails huge discussions that cannot properly be addressed here. But following what said above, we can affirm that whether FOD generates partial orderings, at least at theoretical level, could not suppose a problem for the rationality of evaluations and comparisons. Secondly, FOD valuation does not directly provide information with respect to the intensity of dominance (i.e., to what extent A dominates B or vice versa). Here, a bootstrapping procedure, a method to deal with statistical uncertainty, is applied to get the ‘probability’ or ‘depth’ of dominance which expresses the likelihood that A dominate B or vice versa and, to what extent (e.g., Hussain et al., 2016; Arndt et al., 2018).

26 The method that combines both deontological and consequential evaluation is known as goal-rights system (Sen, 2000).

27 This expanded analytical framework relies on broader philosophical foundations such as the ethical rationality (Sen, 1977) and the positional objectivity (Sen, 1993, 1994).

28 For an account of Capabilitarian sufficiency, see (Nielsen & Axelsen, 2017).

29 For an opposite view that is frequently used to justify privatisation of healthcare facilities, see (Dworkin, 1993).

30 Advantages for interpersonal comparability, thus, redistributive justice, of espousing CA pluralism, objectivism instead of standard economics utility-based consequential evaluation espousing monism, subjectivism, individualism are illustrated by (Nielsen, 2012).

31 A unique index is consistent with Martha Nussbaum’s Capabilities Approach (a so-called objective-list ethical theory) and her list of ten valuable capabilities for human life (Nussbaum, 2000). However, it is not consistent with Sen’s procedural-deliberative version of the CA and Sen’s idea of a wide Capability set containing achieved and not achieved functionings (i.e., people’s actual possibilities-positive freedom for welfare).

32 For a comparison of the characteristics of welfarist and extra-welfarist accounts and how they respectively evaluate alternative social/health states, see (Brower, Culyer, Van Exel, & Rutten, 2008).

33 A popular capability-based measure adopting a ‘weightings’ approach to multidimensional poverty is the Alkire’s and Foster’s model (2011) which is implemented by UNDP (United Nation Development Program). The advantages for consistency of results and rankings of complementing standard measures namely MPI (Multidimensional Poverty Index) with robust measurements from FOD has been shown by Permanyer and Hussain (2018). They use a FOD approach to uncover the socio-economic gradient that exists between countries while not having to rely on many of the normative binding assumptions that underpin the construction of the MPI index.

34 The choice of focusing on education instead of directly on income is explained at (Hussain et al., 2016, p. 747). Most likely, in contexts with higher socio-economic inequalities than Denmark and not having a universal healthcare system, this variable should be different.

35 As suggested by Alkire (In Kauffman (Ed.)2005, pp. 133–152), CA-oriented deliberative-participative methods as ‘participatory planning’, which is mostly used in poverty analysis, can functionally be used to extrapolate ‘group values’ also in other research areas.