Priority setting in the German healthcare system: results from a discrete choice experiment

Priority setting in healthcare decision-making is inevitable. Limited health resources, growing health expenditures, combined with technological advances and an ageing population, continue to put pressure on healthcare systems (Fleck, 2001; Emanuel, 2000; Fuchs, 2010, Borck et al., 2012). Policymakers are encouraged to consider priority setting in healthcare, albeit such rationing decisions may not always be based on transparent processes, but are ad-hoc (Baltussen & Niessen, 2006) or history-based (Kapiriri & Martin, 2007), possibly leading to a suboptimal use of resources.

Complex developments in healthcare are widening the gap between new technological possibilities and advances in therapy and diagnosis and the financial feasibility of the German Healthcare system (Borck, 2012; di Costanzo, 2020). In that regard, concepts such as rationalisation and prioritisation are intensely debated in the quest to implement an equitable allocation of resources (Schmitz-Luhn & Katzenmeier, 2016).

For Germany, the controversial debate on healthcare expenditures and limits on the availability of healthcare services offered by the system started relatively late compared to other countries (Sabik & Lie, 2008).¹ Germany’s health expenditures have steadily increased from 9.4 per cent of its Gross Domestic Product (GDP) in 1992–11.7% in 2019, placing Germany in 6th place in the world rank and on top of the EU27 (8,3%) (OECD, 2021, Statistische Bundesamt, 2022). In international terms however, the German healthcare system stands out for a generous benefit package with high levels of capacity and relatively low cost-sharing (Beske & Drabinski, 2005). As in most high income countries, the future levels of expenditure on GDP, due to demographic and epidemiologic changes (e.g. increase in life expectancy, a diminishing mortality rate) (Institute for Health Metrics & Evaluation, 2016; World Bank, 2016) combined with the increasing costs of medicine are likely to be significantly higher. This has implications for the statutory health insurance benefits as the Federal Joint Committee (G-BA) decides over what adequate, appropriate, economic healthcare entails (Federal Joint Committee, 2017). However, rationalisation in terms of efficiency gains alone will hardly be enough to prevent a further divergence between the feasible and the financially viable.

The complexity of such decisions suggests the need for explicit criteria to be used (Alliance for Health Policy & Systems Research, 2004; Baltussen & Niessen, 2006; Chalkidou et al., 2016; Cromwell et al., 2015). In this respect, discrete choice experiments (DCEs) allow for the simultaneous assessment of multiple policy criteria and preferences elicitation of decision-makers when facing trade-off decisions (de Bekker-Grob et al., 2012; Hansen & Devlin, 2019; Lancsar & Louviere, 2008; Ryan & Gerard, 2003; Thokala et al., 2016).

Past studies support the feasibility and acceptability of DCEs in eliciting public preferences towards societal values and attribute-driven interventions (Genie et al., 2020; Green & Gerard, 2009; Krauth et al., 2021; Krinke et al., 2019; López-Bastida et al., 2019; Luyten et al., 2015). Empirical studies in Germany eliciting priority rankings for the treatments of determined groups, find treatments for children are ranked highest (Raspe & Stumpf, 2013) whereas Diederich (2011) finds little evidence that the German public accepts age as a priority criterion for healthcare services, although there is wide agreement to prioritise specific age groups.

A combination of efficiency and societal values tends to be predominantly considered in priority setting (Baeten et al., 2010; Kenny & Joffres, 2008). With the advent of numerous new initiatives in the health sector, decision-makers are expected to choose between competing healthcare interventions and explicitly consider equity and efficiency trade-off (Baltussen & Niessen, 2006). Transparent and informed decision-making in healthcare requires national level criteria are to be established (Mitton & Donaldson, 2004; Ottersen et al., 2016; WHO, 2007) encouraging a more open social and political discourse on the priorities in medical care to guide policy making on which healthcare technologies should be publicly financed at the different levels of the Social Health Insuracne (SHI) system (Diederich, ; Gerst, 2014; Norheim, 2016).

This paper presents the results of a DCE conducted in 2015, aiming to explore key stakeholders’ preferences for different features of healthcare policies and interventions in Germany, and to show how the latter are prioritised according to such preferences. The paper builds on past work (Defchereux et al., 2012; Mirelman et al., 2012; Mentzakis et al., 2014; Paolucci et al., 2015; Baji et al., 2016) and extends the pool of countries with available data for cross country comparisons.

The main pillar of the German system is the Statutory Health Insurance which is inspired by strong solidarity principles providing the normative basis to pursue the objectives of equity and comprehensiveness, and which represents the framework of regulations on the provision and the financing of healthcare services (Oduncu, 2013). The SHI regulations aim for cost-containment and sustainable financing mechanisms, managed competition, as well as improved efficacy and quality (Blümel & Busse, 2017; Sauerland, 2001).

At the federal level, the Social Code Book V gives a foundation for entitlements, rights and duties of insureds covered by SHI. However, it does not lay down specific guidelines but, instead, sets a framework for policy interfaces (Dannecker, 2009). The scope of benefits is subject to negotiations between the latter bodies and the associations of payers and providers (Blümel et al., 2020). In an international comparison, the German benefit catalogue shows to be quite extensive, leading to reform efforts and the question about how far the solidarity should go. The Reform Act in 2004 showed early attempts of rationing certain benefits from the catalogue, for instance medications for the treatment of erectile dysfunction, hair loss or smoking cessation (Burkhardt, 2012).

Ever since its resumption, the various aspects of prioritisation have been discussed following different political strategies and institutional procedures (Friedrich et al., 2009; Groß et al., 2010; Heil et al., 2010; Müller & Groß, 2009; Oduncu, 2012; Peacock et al., 2006; Raspe & Meyer, 2009; Schöne-Seifert, 2006, Heyers, 2016, Petri, 2015). The Central Ethics Committee for Observance of Ethical Principles in Medicine (ZEKO) and the German Medical Association have promoted the concept of prioritisation (ZEKO, 2007; ZEKO, 2000; Bundesärztekammer, 2014; Borck et al., 2012; Raspe & Schulze, 2013; Diederich et al., 2011) with specific focus on its legal, ethical and economic aspects and have supported the use of pre-defined criteria to evaluate medical services and benefits (Marckmann, 2009; Gordijn & Have, 2013; Oduncu, 2012; Borck et al., 2012; Storz & Egger, 2010; Kliemt, 2006).

As such, the three major criteria of prioritisation ‘‘medical need’’ (severity and urgency), ‘‘proven benefit and fitness for purpose’’, and ‘‘cost–benefit-effectiveness’’ have been proposed by the central ethics committee (ZEKO, 2007).

A discrete choice experiment (DCE) was employed to assess the relative weight of various criteria in setting priorities in the German health arena. DCE are commonly used in healthcare for prioritisation decisions (Lancsar & Louviere, 2008; Ryan & Gerard, 2003; Ryan et al., 2008) using public preferences. Health economists have acknowledged the benefit of the approach especially when facing health policy, planning and resource allocation decisions in high-income countries. In that regard DCEs are widely applied to a range of policy questions (Baji et al., 2016; de Bekker-Grob et al., 2012; Ryan & Gerard, 2003; Whitty et al., 2011) and priority setting frameworks (Baltussen et al., 2006; Peacock et al., 2010; Razavi et al., 2020). These include the elicitation of views on diagnosis, treatment and care (van de Schoot et al., 2017, Koopmanschap et al., 2010, Clark et al., 2017, King et al., 2007; Kjaer & Gyrd-Hansen, 2008), access to services (Longo et al., 2006; Mengoni et al., 2013), consumer (health) preferences (Czoli et al., 2016) and the employment preferences of health personnel (Mandeville et al., 2014; Wordsworth et al., 2004).

Respondents’ preferences are elicited in a survey adapted from previous studies (Defechereux et al., 2012; Mirelman et al., 2012; Paolucci et al., 2015). Respondents’ were asked to choose among a set of hypothetical alternative interventions presented in choice sets, with each alternative described in terms of six criteria. To every criterion, values have been assigned over a range of pre-defined levels.

Table 4 presents the mixed logit estimates for equity and efficiency attributes among German decision-makers. Magnitude is not directly interpretable and therefore we discuss sign and significance of the coefficients in the first instance; a positive sign suggesting utility increasing characteristics and conversely for a negative sign. With the exception of ‘middle age group’ and ‘willingness to subsidise’, all coefficients were statistically significant at 1% (p < 0.01).

Respondents appear to prefer interventions addressing the young (baseline) over those targeting high age groups. Moreover, interventions requiring public support are not favoured. On the other hand, there seems to be a strong preference towards interventions that target the severely-ill as well as towards interventions with substantial health effects for those treated. Not surprisingly, interventions that are beneficial for a larger proportion of the population and those which prove to be cost-effective are favoured.

Moving on to Table 5, with regards to equity criteria, ‘severity of disease’ increases the probability of selection for an intervention by 7.23% (95% CI 6.23–8.22) as compared to the baseline. All other equity attributes reduce the probability of selection as compared to the baseline. However, the effects for ‘middle aged’ and ‘willingness to subsidise others’ are statistically insignificant. Looking at efficiency criteria, all three criteria exhibit large, significant and positive effects. The probability of selecting interventions that entail substantial health benefits increases by 6.11% (95% CI 5.17–7.05) compared to the baseline, while the corresponding probability for interventions that provide benefits to a larger share of the population is 6.25% (95% CI 5.28–7.21), and 5.83% (95% CI 5.00–6.66) for interventions that are cost-effective. With regards to aggregate criteria along with the calculated equity/efficiency trade-off, interventions appear to be strongly preferred when improving efficiency, independent of the age group that is targeted. This, however, is especially true for interventions targeting young and high age groups. Except for aggregated equity criteria for high age groups, all coefficients are significant.

This study draws attention on the use of discrete choice experiments to devise rational frameworks for priority setting, taking explicitly into account the concerns for different societal objectives. In this regard, the results of the experiment on a sample of relevant stakeholders in the German health system allow one to discuss some interesting findings.

German decision makers consider severity of disease, individual health benefits, cost-effectiveness, as well as number of potential beneficiaries as important criteria for priority setting. The absolute values of the regressions reflect their relative importance in priority setting. Based on their respective weights, severity of disease, number of potential beneficiaries, individual health benefits and cost-effectiveness appear as the most important criteria for German decision makers within the sample population, displaying great preference towards efficiency.

Besides the general pro cost-effectiveness attitude, respondents associate a higher utility to interventions targeting younger age groups. This is in line with empirical studies in Germany, such as Raspe and Stumpf (2013), which find priority setting in favour of treatments for children. Willingness to subsidise others appears insignificant, which again confirms a priori expectations in high-income countries such as Germany that are characterised by universal coverage (cf. Norway and Austria). Overall, all efficiency attributes are favoured over equity criteria, except for severity of disease (Diederich et al., 2012). Although equity concerns seem to be comparably less important in healthcare resources allocation decisions in Germany, the two objectives (i.e. efficiency and equity) are in conflict with each other and are equally needed in a deliberative process (Culyer, 2006, 2015). The estimated ratios between equity and efficiency support a general preference for efficiency over equity criteria for all age groups, with much stronger results for interventions targeting younger and higher age groups.

These findings show a large overlap with the prioritisation discussion in Germany and are aligned with what was proposed by influential bodies in the German health community. For instance, in a second plea in favour of a priority setting debate in 2007 (first in 2000) the ZEKO addresses the importance of defining the best relative weight for the much-needed prioritisation criteria. Almost all revealed criteria, namely proven benefit/fitness of purpose, cost–effectiveness and medical need (urgency and severity), support our findings. Nevertheless, the general preference assigned to efficiency does not involve a lack of concern for equity. Indeed, basic equity in terms of financial protection is guaranteed through the basic solidarity principle grounded within the SHI. This principle entitles every individual to the same services irrespective of their insurance status or the contributions paid (Deutsche Sozialversicherung, n.d.).

The CLT results can be considered indicative when prioritizing among interventions. Largely, the resulting rankings reflect the National Health Goals (Bundesministerium für Gesundheit, Gesundheitsziele.de, 2022) concerning Type 2 diabetes, breast cancer, depressive disorder, healthy ageing, reduction of alcohol and tobacco consumption and enhancing health competence (Federal Ministry of Health, 2007). These main goals are a complementary governance tool in healthcare and seek to improve the health of individuals or specific groups to tackle the conditions of highest urgency. Yet, the high prioritization of mental disorders is not in line with Schröter and Diederich (2013) who reported that the German population considers mental health of lower importance for prioritisation of medical resources compared with physical health. Nevertheless, such discrepancies in preferences for mental health could largely depend on the specific context.

Together with the National Health Goals, the Information System of the Federal Health Monitoring and the Federal Joint Committee identified the disease burden as one of the most relevant determinants in setting healthcare priorities. This approach is in line with the WHO “2013–2020 Global Action Plan for Prevention and Control of Non-communicable Diseases” that underlines the need to urgently address prevention problems, and to allocate more resources to the early treatment of chronic NCDs and mental illnesses (WHO, 2013).

Similarly to the German results, preferences for efficiency and equity criteria elicited across countries have displayed individual benefits, severity of disease and cost-effectiveness as the most significant priorities in high income countries, HICs (Baji et al., 2016; Defechereux et al., 2012; Mentzakis et al., 2014). In low-income countries (LICs) like Ghana or Nepal (Baltussen et al., 2006, 2007) instead, number of beneficiaries, individual benefits, cost-effectiveness, severity of disease, and middle-aged people are found to be the preferred criteria—showing more balanced equity/efficiency preferences. Results of a Chinese study (Paolucci et al., 2015) disclose a much closer profile resembling that of the mentioned high income countries where universal health coverage is in place.

Comparing the CLTs obtained for Germany with those for Austria and Norway (as instances of comparable HICs) findings are largely comparable (Defechereux et al., 2012; Mentzakis et al., 2011). In those studies, countries share a similar disease burden, comprising mainly mental disorders and NCD, including diabetes, cancer, and cardiovascular diseases. This holds true for the CLT that assigns relatively high rankings for respective disease areas. Compared to other HICs, German decision makers seem to rank higher those interventions affecting young or middle-aged people.

One of the limitations of DCEs is their hypothetical nature. Due to the explorative nature of the study the findings cannot be directly implemented into national policy making but could act as a first step and guide. In fact, results can contribute towards the German debate on setting priorities in healthcare. Further, we note that while our sample size is not small and allows for robust estimation, the low response rate suggests caution in inference and limited generalizability, while future research should explore the congruence of preference between stakeholders and general public. Nonetheless, the methodology is generalizable and can be transported to other countries and settings when the required conditions for successful multi-criteria decision analysis (MCDA) in health are met. Apart from age, non-linear effects were not incorporated in the analysis. Our design identifies individuals’ direction of preferences rather than the exact shape of their function, while future research could focus on non-linearities. The survey and attribute levels were taken from a larger DCE project targeting many countries and as such reference levels in the dichotomization of the attributes were taken to meet international standards and ensure cross-country comparability. Yet, such dichotomization and use of attributes with different scales could introduce vagueness and affect their perception by respondents and conceal potential difference in the relative importance of attributes. Future studies could increase the number of levels for relevant attributes and obtain preferences over a range of discrete attribute values.

Establishing criteria for equitable and efficient resource allocation in healthcare is a political task with a number of dimensions including medical, economic, ethical and legal ones. The complexity of the issue makes it impossible to achieve a complete consensus between all those involved. Nevertheless, principles ought to be formulated in which existing structures and processes must be measured, not at least in the sense of a future-oriented perspective.

In conclusion, this explorative study details how multiple criteria can guide a transparent and systematic priority-setting process by allowing for the simultaneous assessment of multiple policy objectives of decision-makers. With German decision makers stating a preference for efficiency, such an approach can help to support the priority setting processes and may contribute to a more informed and participated debate on priority setting between different health interventions in Germany.