International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Patients’ preferences for specialised health professionals in follow-up care. A cross-country and cross-disease comparison.
Sebastian Heidenreich, Diane Skatun, Mandy Ryan, Job van Exel, Elliot Robert, Christine Bond

Last modified: 28 March 2017


I Background and Objective

In 2015, average expenditure on health and health care in Europe accounted for 9.5% of the GDP; compared to 8.2% in 2000 (WHO 2016). Labour costs are the largest single item of this expenditure (Hernandez 2006). Despite the increases in expenditure, the pressures on European health budgets is intensifying, due to a rising health care demand that results from population ageing, and rising costs of drugs. In many countries, this has resulted in changes in the way health care is delivered. Prominent among these changes has been workforce redesign, which has raised questions about the optimal skill mix of the health workforce (Tsiachristas et al. 2015). Most European countries have extended roles for existing professions or introduced new health care professions. In some cases, this has resulted in a transfer of health care tasks from physicians to other health care professionals (Bosley and Dale 2008, Ross et al. 2012). While the introduction of these new health care roles aims to relieve the financial pressure on the health care budget and is supposed to increase the overall efficiency of health care systems, little is known about how patients value the adjustment of the skill mix in the health workforce.

In this paper, we use a discrete choice experiment to elicit patients’ preferences for the health professionals delivering follow-up care and whether or not they received specialist training. We also address two important methodological issues in the modelling of the DCE data: (1) The study is conducted across seven European countries and three disease areas; we explore how to control for spatial correlation in preferences and health care system dependency. (2) Patients’ evaluation of the opt-out alternative might depend on their risk preferences and their expected health outlook if they do not receive follow-up care. We control for such heterogeneity in respondents’ understanding of the opt-out using a latent variable approach. The final estimates will be used to evaluate different policies that aim to slow the increase in health expenditure by delegating tasks from specialist and generalist doctors to new or existing health professions.

II Study design and data collection

The DCE is part of the MUNROS project, funded under the FP7 framework of the European commission (Bond et al. 2016). Within the DCE, respondents are asked to choose between two unlabelled arrangements of their follow-up care and the option to have no follow-up care at all. The two follow-up care appointments are characterised by four attributes: (1) the length of a follow-up care appointment; (2) the frequency of appointments; (3) the consulting health professional; and (4) the monthly costs to patients. All attributes are generic characteristics of follow-up care appointments applicable across the three covered disease areas (heart disease, type-2 diabetes, breast cancer). Levels of the cost attribute are presented as rounded purchasing power parity equivalents in the home currency of participating countries and are based on values from a contingent valuation payment card that was included in a previous questionnaire in the MUNROS project.

A pilot study was undertaken in seven countries (UK, Netherlands, Germany, Italy, Poland, Czech Republic, Turkey) and the analysis is currently being conducted. The pilot was based on a D-efficient design (flat priors) consisting of 30 choice tasks that were split into two equally sized blocks. Estimates from the pilot will be used to generate an updated D-efficient design with Bayesian priors. The data for the main study will be sampled subsequently in 12 different hospitals and primary care sites in participating countries in autumn 2016.

III Statistical approach and methodological challenges

A random parameter (RP) logit will be used as a baseline model and extended subsequently to account for various methodological challenges in the data. In an initial specification step, we will account for correlation in preferences within each disease area and country by adding corresponding random effects to the model.

In a next step, we explore if respondents’ understanding of the opt-out alternative varies due to two reasons: (1) different respondents may assess the risk to their health outlook differently, if they choose to not receive follow-up care. For this purpose, respondents are asked to rate their health outlook for the next two years with and without follow-up care on a symmetric Likert Scale ranging from -5 to +5. (2) Respondents may also differ in their willingness to take these risks. Information on respondents’ risk aversion in a health context is collected using a series of validated rating questions (Van Osch & Stiggelbout 2004). The effect of respondents’ self-assessed risk of having no follow-up care and their general willingness to take risks with their health are integrated into the analysis within a latent variable framework to reduce the risk of endogeneity. Specifically, we will estimate a ‘subjective risk’ factor and a ‘risk aversion’ factor. These latent variables are interacted with the alternative specific constant of the opt-out option. The resulting hybrid choice model will be estimated simultaneously.

The fully specified choice model will be used to test if patients’ preferences vary between countries and disease areas. Hicksian welfare measures of alternative policies that either allow patients to choose between different follow-up arrangements and variants with no choice will be calculated from the final model specification.

IV Conclusion

This paper will present the first study eliciting preferences for new roles of health professionals in Europe. We will also make methodological contributions by accounting for respondents’ understanding of the opt-out in the context of subjective risk and by exploring how to combine stated preference data from different health care systems and disease areas meaningfully.


Bond C, Bruhn H, de Bont A, van Exel J, Busse R, Sutton M, Elliot R on behalf of the MUNROS team 2016. The iMpact on practice, oUtcomes and cost of New roles for health pROfeSsionals: a study protocol for MUNROS, BMJ Open, 6(e010511).

Bosley S, Dale J 2008. Health care assistants in general practice: practical and conceptual issues of skill-mix change, British Journal of General Practice, 58: 118-124.

Hernandez P, Dräger S, Evans DB, Tan-Torres Edejer T, Dal Poz MR 2006 Measuring expenditure for the health workforce: evidence and challenges, Evidence and Information for Policy, Geneva: World Health Organisation.

Ross N, Parle J, Begg P, Kuhns D 2012. The case for the physician assistant, Clinical Medicine, 12(3): 200-206.

Tsiachristas A, Wallenburg I, Bond CM, Elliot B, Busse R, van Exel J, Rutten-van Mölken MP, de Bont A, the MUNROS team 2015. Costs and effects of new health professional roles: Evidence from a literature review, Health Policy, 119: 1176-1186.

Van Osch SMC, Stiggelbout, AM 2004. Development of the Health-Risk Attitude Scale (H-RAS).  https:/, In: Van Osch, SMC. The Construction of Health State Utilities, 117–138, Leiden: Department Medical Decision Making, Medicine / Leiden University Medical Center (LUMC), Leiden, accessed: 15. August 2016.

World Health Organisation (WHO) 2016. Global Health Expenditure Database, nha/database, accessed: 15. August 2016.


Conference registration is required in order to view papers.