Effective policy mechanisms for an equitable geographical distribution of general practitioners: a qualitative comparative analysis of the accessibility of primary care in Europe

Abstract

Objectives

To explore the effectiveness of four different policy mechanisms in achieving a more equitable geographical distribution of general practitioners (GPs) in European countries. The following mechanisms were analysed: (1) interventions during medical training; (2) financial incentives; (3) quotas to allocate GPs to regions and (4) capitation-based remuneration.

Methods

A macro-comparative method, namely, fuzzy set qualitative comparative analysis, was employed to explore the distributional effectiveness of the four mechanisms. A literature review yielded information on the use of these mechanisms in the 21 European countries included, while country-specific equity in the geographic GP distribution served as the outcome variable.

Results

Quotas determining the number of GPs per region proved to be highly effective in producing an equitable GP distribution if calculated based on health care needs. Remunerating GPs largely through capitation payments also proved to be an effective policy mechanism. Financial bonuses to GPs practising in under-served areas and interventions during medical training had little or no impact.

Conclusion

Several high income countries have a maldistribution of primary care physicians to the detriment of rural or socially deprived areas. Policy makers have instituted a variety of policies to counter this. This study helps to identify mechanisms which are likely to be more and less effective.

Keywords

geographic variations health equity primary care

Introduction

Theoretical background and research question

Ensuring accessible and needs-oriented health care is an integral aim of health policy in many countries. Primary care plays a crucial role in fulfilling this objective, since it provides the first local contact point to health care services.¹ Accordingly, it is well recognized that access to primary care providers should be guaranteed for everyone, with priority to those most in need. In other words, there should be equity in access to primary care.²

Access to health care can be hampered by a variety of factors, ranging from financial affordability of care to its physical availability.³ This analysis focuses on local accessibility of primary care. Accessibility is most difficult to achieve in low- and middle-income countries where health care workers are scarce.

However, the issue of accessibility to primary care has also reached the agenda in high-income countries. Rural, remote and socially deprived areas have the greatest problems attracting staff. Therefore, the Organisation for Economic Co-operation and Development (OECD)⁴ regularly devotes a chapter in its Health at a Glance reports to inequity in access to health professionals. In many European Union (EU) countries, there is a national debate about the adequacy of the local health care professional workforce. To date, several high-income countries have implemented their own regulations and programmes to optimize primary care accessibility; however, there is no common understanding on best practice.

To fill this gap, this study aimed to evaluate four different policy mechanisms with regard to their effectiveness in inducing an equitable geographical distribution of primary health care suppliers. The analysis focused on two policy mechanisms recommended by the World Health Organization (WHO),⁵ in addition to two other distributional mechanisms used in 21 European countries. For each of these countries, a thorough literature review was conducted examining the existence and degree of implementation of the four mechanisms which may affect supplier distribution. In addition, the degree of equity of the distribution of primary care providers was computed. Since primary care is provided by many different health professional groups whose roles vary largely across different countries, this analysis focused on general practitioners (GPs) who play a central role in most health systems. Equity of GP distribution was calculated for each country based on GP distribution between sub-national entities. In order to bring together the rich qualitative data from the review and the quantitative information on supplier distribution, a fuzzy set qualitative comparative analysis (fs/QCA) was used to evaluate the effectiveness of the policy mechanisms. Fs/QCA is a macro-comparative research method which allows the combination of qualitative, case study-based information and quantitative data in order to identify causal relations between (combinations of) conditions and outcomes.⁶

Hypotheses and the outcome variable

Guided by systematic reviews, public health literature and policy papers, four policy mechanisms were identified as potentially effective in achieving greater geographical equity in GP distribution. The first two were derived by clustering the policies recommended by the WHO expert council⁵ into two mechanisms. They focus on attracting health workers to underserved areas. The subsequent two do not specifically address underserved areas but take a wider distributional approach.

Interventions during training

Several studies have found evidence that mechanisms related to the recruitment or the curricula of medical students have a positive effect on attracting physicians to rural and remote areas (RRAs); the following are the most prominent^7–10:

preferential recruitment of students with a rural background or affinity;

early exposure to rural practice during medical studies;

provision of scholarships or loan forgiveness for students committing to rural practice; and

location of universities and training facilities outside metropolitan areas.

Proposition 1 (P1): Health care systems which use policy mechanisms targeted at achieving an equitable geographical distribution of physicians during their training are more likely to have a more equitable distribution of GPs.

Direct financial and non-financial incentives

Several studies have identified an effect of financial and non-financial incentives on GP distribution.^8,10,11 They include:

permanent or temporary remuneration bonuses for practice in RRAs;

financial bonuses for agreeing to work in an RRA covering costs of moving and housing;

advisory services regarding practice location, practice set-up, housing, etc.;

family support, e.g. regarding spouse’s job search, child care, schooling, etc.;

mentoring during practice start-up.

Proposition 2 (P2): Health care systems that offer financial and non-financial incentives aimed at increasing the attractiveness of practice in an RRA will have a more equitable GP distribution.

Quotas for physician contracts and practice licenses

Several European countries limit the number of GPs per district or other sub-national entity that may (fully) charge the health insurer.⁷ Measures include:

allocation of practice licenses according to minimum and/or maximum quotas of GP density per sub-national entity; and

selective contracting by health insurers with GPs, e.g. according to needs assessment or minimum/maximum quotas per sub-national entity.

Proposition 3a (P3a): The inequity of GP distribution will be lower in those health care systems where the compensation or reimbursement of GP charges by the health insurer is only (fully) granted to those GPs who have a license or contract, and where the geographic allocation of such licenses/contracts is limited in number.

Quotas are usually defined in terms of absolute physician numbers or densities. Thus, if there is a considerable overall shortage of GPs, they may not have a redistributive effect. In order to control for this aspect, the following supplementary condition was introduced:

Supplementary Condition 1 (SC1): This mechanism will be more effective in countries with a high average GP density.

Regulators define quotas per sub-national entity using a variety of criteria, e.g. based on an assessment of health needs, population numbers, historical demand or historical supply. However, the outcome is likely to differ based on the applied criterion. For instance, if a quota is based on the past pattern of supply, an inequitable distribution of physicians may be conserved. Instead, only those quotas that are defined by health needs should lead to higher equity. The sophistication in defining health needs may vary considerably – it may range from a basic per capita calculation to a multivariate needs index.

Proposition 3b (P3b): Quotas for the geographic allocation of billing rights are more effective if the quota is based on an estimation of health care needs.

Indirect financial incentives from GP remuneration schemes

A fourth category comprises those mechanisms that result (intentionally or not) from the remuneration scheme for GPs. In the reviews cited above, only those remunerative mechanisms are discussed which are intended to influence GP distribution directly, such as bonuses for rural practice paid in addition to usual compensation. Yet many European countries have adopted a complex mix of remuneration mechanisms for reasons of volume control and quality incentives, involving capitation, fee-for-service, per case payments, quality-based payments and salaries.¹² Capitation fees are deemed particularly effective in avoiding supplier-induced demand and countering oversupply since they do not reward the GP for services provided, but for the number of residents they are responsible for.^13,14 Since the earning potential of GPs is, therefore, clearly restricted by the population size in that location, capitation remuneration may also help avoid a geographic concentration of GPs and, thereby, lead to a more equitable GP distribution.

Proposition 4 (P4): Health care systems in which GP remuneration relies mostly on capitation will have a more equitable GP distribution.

However, there are two supplementary conditions to this: first, capitation will be more effective in regulating GP distribution in those health systems in which patients are required or incentivized to register with a GP and where the GP has a gatekeeping role. Otherwise, GPs might take on more patients than they could care for and thereby induce patients to bypass the GP.

Supplementary Condition 2 (SC2): This mechanism is more effective in health care systems with a family doctor system (i.e. mandatory registration) and an effective GP gatekeeping.

Second, supplementary condition SC1 (as defined under P3a) also applies: namely, capitation may be more relevant in systems with a high GP density, since income potential in oversupplied regions will become very limited.

Methods

Qualitative comparative analysis

A qualitative comparative analysis (QCA) was conducted in order to evaluate the effectiveness of the four mechanisms in achieving an equitable distribution of primary care providers. QCA is a method that combines the richness of qualitative data collected for a medium to high number of cases with the advantages of quantitative data analysis.¹⁵ The aim of QCA is to identify the relevance of causal conditions (independent variables) in causing a certain outcome (the dependent variable). In contrast to regression analysis, QCA is based on the idea that there may be several parallel causal paths leading to an outcome. Each case in the sample provides a configuration of conditions. By applying Boolean algebra, the necessary and sufficient conditions required for the outcome can be derived.^16,17

Sampling of cases

The sample comprised member countries of the EU and the European Free Trade Association (EFTA). The rationale was that the EU provides a system for identifying comparable geographic areas for these countries, the so-called nomenclature of territorial units for statistics (NUTS). This analysis was based on NUTS 2 regions, which range between 800,000 and three million residents.¹⁸ Given that primary care should be provided close to home, this is a relatively large population area. However, many countries do not report GP numbers for smaller areas. In order to include a large sample and, thereby, strengthen the validity of the results, basing the analysis on NUTS 2 seems a reasonable compromise. The term ‘region’ used in this study thus refers to NUTS 2 regions.

All EU and EFTA countries were included unless one of the following criteria applied:

Countries that consist of less than four NUTS 2 regions had to be excluded since otherwise an analysis of geographical distribution among sub-national entities would not have been possible.

The four French overseas regions were excluded due to their great distance from the mainland, which makes a common market for health workers doubtful. Also, the regulatory framework of their health systems partly differs from the mainland’s.

Based on these criteria, 21 countries comprising 270 regions were included in the sample.

Definition and calculation of the outcome variable

The outcome variable was defined as the degree of equity in the geographic distribution of GPs in each of the included countries. The concept of ‘equity’ relies on the notion that health care services should be provided according to the level of health care ‘need’ of each individual.^19,20 However, there is neither a consensus as to how to define ‘need’ nor a common understanding of valid need indicators.¹⁹ Difficulties also arise from a lack of comparable international data.

In this study, the health care ‘needs’ of the regional population were, therefore, approximated by taking into account its age structure. Since elderly people have a higher need for health care services, regions with an older age structure require a higher density of GPs. This was accounted for by calculating a needs-weighted population in a manner similar to Matsumoto et al.²¹ Several statistical sources from different countries confirm that there is an increase in physician consultation rates by a factor of 2.5 for people aged 65 years or above compared to people younger than 65 years of age.^22–24 Therefore, the needs-weighted population was calculated as follows:

W_{ij} = (1 - e_{ij}) \times P_{ij} + 2.5 \times e_{ij} \times P_{ij}

where P_ij reflects the population size of region i of country j and e_ij is the share of regional population aged 65 years or older. Equity was then measured by contrasting the number of GPs per region with the needs-weighted population size W_ij in that region.

Indicators measuring distribution are subject to distortion based on the number of regions entering the analysis and the existence of extreme values.²⁵ In order to limit these distortions, two indicators were combined, namely the Gini coefficient and the decile ratio. The Gini coefficient is relatively robust with regard to outliers, but it is sensitive to the number of data points. In contrast, the decile ratio is less sensitive than the Gini coefficient to the number of data points but responds more strongly to outliers.²⁵

For the calculation of both measures, NUTS 2 regions of each country j were ranked from lowest to highest by their weighted GP density; i.e. the ratio of the regional number of GPs to the needs-weighted regional population.

The country-specific Gini coefficient was then derived by plotting (in rank order) each NUTS 2 region’s cumulative share of weighted population on the horizontal axis against each region’s cumulative share of all GPs in the country. The Gini coefficient was then given by the ratio of the area between the curve derived from the plotted values (Lorenz curve) and the 45° line compared to the entire area under the 45° line.²⁶ Where the weighted GP densities across NUTS 2 regions of a country were the same, the Gini coefficient would equal 0; in cases where all GPs were concentrated in one region, it would be 1.

The decile ratio was computed by comparing the number of GPs available to the weighted population decile in regions with the lowest weighted GP density and the number of GPs accessible by the population decile with the highest weighted GP density. The ratio of highest to lowest weighted GP densities amounts to the decile ratio.²⁷ This indicator takes on values between 1 (total equity) and infinity (total inequity).

Each country’s Gini coefficient and decile ratio were multiplied together. The resulting value is hereafter called the ‘equity indicator’.

Data on GP numbers per region in 2008 were taken from national statistical databases, websites of the national health ministries, health yearbooks, physicians’ chambers and similar sources.

Review process and data coding

A thorough literature review was conducted for each country regarding implemented policies falling under one of the four mechanisms elaborated above. Since there were few academic articles, the review was expanded to a variety of sources. The starting point for all countries was the Health Systems in Transition series by the European Observatory on Health Systems and Policies.²⁸ Other sources were the websites and publications of health ministries, sickness funds, national health services, physicians’ associations, regional governments and independent research institutes. Further information was available through health legislation, parliamentary documents and policy papers. The information was collected in a standardized template for each country.

In order to make these data usable for QCA, they were condensed into ‘conditions’ (as explanatory variables are called in QCA) that were suitable for testing of hypotheses. For each proposition, one condition was defined to account for the degree of effective usage of each policy instrument in each country; the condition Education (E) signified the existence of interventions during the training of GPs; the condition Incentives (I) reflected the use of incentives; Quota (Q) stood for quota-based allocation of GPs; Needs_Quota (N) indicated a needs-based quota and Capitation (C) indicated the use of capitation-based remuneration. The two supplementary conditions were captured by GP_Density (D), which reflected a high number of GPs per 100,000 inhabitants, and Gatekeeping (G), which indicated the degree of gatekeeping in a health system.

Since conditions in QCA must be in a range of {0; 1}, conditions with a value of ‘1’ reflected the full use of the specific mechanism in the respective country, whereas ‘0’ signified the absence of that mechanism. A partial use of a specific mechanism was reflected by two further values lying in-between. Thus, the variables derived from the review were coded on a four-point scale {0; 0.33; 0.67; 1} in order to strike a balance between an adequate granularity, while avoiding an over-precision that would impede reproducible and transparent coding results. Values were assigned based on a codebook that was developed beforehand, which was designed to reflect the range of each variable (see the online Appendix to this article for further details). The equity indicator, derived from the computation of GP distribution, was calibrated to the range {0; 1} using logit transformation.²⁹

Data analysis

Each country displays a specific configuration of present and absent conditions. These are listed in a so-called truth table (see the online Appendix to this article). By identifying those conditions or combinations of conditions that consistently lead to a positive outcome for all cases, the relevant pathways can be derived. By way of illustration, the method is demonstrated by a Venn diagram showing four hypothetical conditions (see Figure 1).³⁰ Each circle A, B, C and D contains the entire set of cases (in this analysis: countries) which fulfill the hypothetical condition A, B, C or D, respectively; while the circle Outcome contains all cases with a positive outcome (e.g. in the case of this study, the outcome is Equity). Where circles overlap, cases fulfill both conditions. In Figure 1, condition A is always present where the outcome is also positive. Hence, condition A is identified as a necessary condition for the outcome. Yet, A is not a sufficient condition since the mere presence of A may also lead to a negative outcome. Each condition B or C is, by itself, also not sufficient. However, the presence of both conditions B and C always leads to a positive outcome. In QCA, this would be written as ‘B*C → Outcome’. Since the intersecting set of B*C fully coincides with the circle Outcome, B*C represents a sufficient solution term with a consistency of 1. Condition D also has a large, (but not full) overlap with the Outcome. Therefore, it could also be recognized as a sufficient condition yet with a consistency slightly less than 1. The threshold for the sufficiency of a causal condition was set to ≥0.75 in this analysis, in accordance with Schneider and Wagemann.¹⁶ A second important measure is the coverage ratio. It reflects the relevance of a solution for the outcome. In the illustrated diagram, the coverage ratio of B*C would be relatively low due its small area compared to the total area of the circle Outcome. The solution term D, on the other hand, would have a larger coverage. Thus, the consistency ratio may be interpreted as similar to the significance level in multivariate quantitative analyses, while the coverage ratio signifies the degree of explanatory power of the solution term.¹⁶

Figure 1.

Illustrative Venn diagram of necessary and sufficient conditions. Circles A, B, C and D represent hypothetical sets of cases that fulfill the hypothetical condition A, B, C or D, respectively. The circle ‘Outcome’ contains cases with a positive outcome. Source: Author’s illustration based on Ragin.³⁰

This process illustrated graphically in Figure 1 was performed using the fs/QCA software.³¹ This software essentially uses the same logic, yet it takes into account fuzzy sets; i.e. cases which are not only full members or non-members, but also those which have an incomplete (non-)membership.

The analysis was conducted in two ways: in addition to studying the measures which led toward the outcome, Equity, a cross-check was performed by analysing conditions resulting in the contrary outcome, Inequity.

Results

Figure 2 displays the country-specific geographical variations in weighted GP densities in 2008. Each data point represents one NUTS 2 region; data were scaled by setting the national average weighted GP density to 1. The chart clearly shows that some countries, such as Greece, Poland and Portugal, have strong variations in sub-national GP densities while others show little variance between sub-national entities (e.g. Denmark, Hungary, Netherlands). Therefore, the sample provides a good basis for this analysis since it contains cases of both relatively equitable and inequitable distribution.

Figure 2.

Weighted GP densities in 2008 per NUTS 2 region scaled by country average. AT: Austria; BE: Belgium; BG: Bulgaria; CH: Switzerland; CZ: Czech Republic; DE: Germany; DK: Denmark; ES: Spain; FI: Finland; FR: France; GR: Greece; HU; Hungary; IT: Italy; NL: Netherlands; NO: Norway; PL: Poland; PT: Portugal; RO: Romania; SE: Sweden; SK: Slovak Republic; UK: United Kingdom. Source: Author’s calculations and illustration.

The analysis of possible necessary conditions showed that none of the conditions was strictly necessary since the recommended consistency threshold of ≥0.9 was not met by any of the conditions.¹⁶

Therefore, the analysis focused on sufficient conditions. Out of 128 logically possible configurations of present and absent conditions, 17 were found in the sample. Out of these configurations, nine proved to lead to the outcome of high Equity. The solution terms displayed a high overall consistency of 0.83 and a high coverage of 0.85. The cross-check for the outcome Inequity yielded eight configurations.

Training-related interventions did not prove to be a relevant instrument in the quest for a more equitable GP distribution. In fact, the condition Education was not part of any of the solution terms (see Table 1). The cross-check, on the other hand, listed the absence of training-related interventions as a driver toward a higher inequity. However, this finding is relatively weak evidence since only three countries made active use of training-related instruments. Therefore, there was insufficient evidence to confirm proposition 1.

Table 1.

Summarized results table.

Outcome	Overall consistency/coverage	Term no.	Solution terms	Consistency	Unique coverage
Equity (O)	0.83/0.85	1.1	QND +	0.83	0.38
	1.2	CGq*d +	0.85	0.17
	1.3	CGQNI	0.87	0.08
Inequity (∼O)	0.91/0.80	2.1	Cqn*D +	0.89	0.10
	2.2	Cne*D +	0.94	0.11
	2.3	Cgqni +	0.92	0.06
	2.4	Qnegd +	0.84	0.03
	2.5	Qnied +	0.93	0.21

Source: Author’s calculations.

The effect of the condition Incentive remained unclear. Solution term 1.3 (see Table 1) implied a positive impact on the equity of GP distribution. However, in all cases underlying this solution term, incentives were always combined with capitation and a (needs-based) quota. The dataset contained only one case, France, which did not combine incentives with capitation or quota-based GP allocation. Yet, France had high inequity (1.0). Thus, the effect of incentives alone remained unclear and it would require further evidence to prove that financial and non-financial incentives have a relevant impact on the equity of GP distribution.

With regards to proposition 3a, a generic quota-based allocation of licenses or contracts did not prove to be effective. Instead, this mechanism only had a positive effect on equity when it relied on a needs- or population-based computation. In QCA nomenclature, a tilde (‘∼’) before the condition signifies the full or partial absence (value < 0.5) of the respective condition/outcome. Therefore, solution term 1.2 Capitation*Gatekeeping*∼Quota*∼GP_Density → Equity suggested that an absence of a general quota combined with capitation-based remuneration might have a positive effect on equity. Such a finding was also confirmed by the cross-check. The solution term 2.5 implied that the existence of a quota that was not needs-based could increase the inequity of GP distribution. Therefore, proposition 3a could not be confirmed.

With regards to the more rigorous proposition 3b, the results showed that a needs-based quota could be used in combination with other instruments such as incentivizing measures and capitation remuneration (1.3) or without any other instruments (1.1). The sufficiency of a needs-based quota alone could only be confirmed for countries with a high GP density. Due to the limited diversity of the dataset, the proposition could not be verified in countries with a low GP density. However, the cross-check showed that none of the countries with a needs-based quota allocation had an inequitable GP distribution. Therefore, needs-based quotas appear to have a strong positive effect on the equity of GP distribution.

Proposition 4 concerned the effect of capitation remuneration on geographical GP distribution. As Table 1 shows, the condition Capitation was found to exert a positive effect on equity. In countries with a low GP density and a gatekeeping system, Capitation proved effective in promoting an equitable GP distribution without the use of further instruments. Unfortunately, the dataset did not provide any cases of high GP density countries that used capitation remuneration without combining it with a needs-based quota allocation. Thus, it was not possible to assess the individual effect of capitation for these countries. Therefore, the analysis implied that in countries with a high GP density, the use of capitation remuneration required to be combined with a needs-based quota (and with incentives) in order to be effective. The cross-check for inequity confirmed the positive impact of capitation, since an absence of capitation proved to reinforce inequity in several configurations (Solution terms 2.1–2.3). Thus, the analysis provides evidence for a positive effect of capitation remuneration on more equitable primary care distribution. As the high consistency ratio of solution term 1.3 demonstrates, this positive effect was strongest when capitation remuneration was combined with a needs-based quota allocation (and, possibly, incentives).

Discussion

This study provides a first exploration of the relative effectiveness of policies to attain equity in the distribution of GPs as important providers of primary care. Figure 3 summarizes these findings. Needs-based quotas and capitation remuneration proved to be very effective in encouraging an equitable distribution of GPs. Countries not using either of these two mechanisms had an inequitable distribution of GPs. Furthermore, countries using a quota-based allocation of GPs, where the quota was not based on any estimate of health care needs, were likely to demonstrate an inequitable distribution of GPs. Interventions during GP training and incentives did not prove to have a major effect.

Figure 3.

Pathways towards (in-)equity of geographical GP distribution. Sweden’s (SE) GP distribution is just slightly above the threshold for inequity and its configuration of conditions contains many fuzzy values. Therefore, it does not entirely fit the displayed pathway and is shown in light grey. Source: Author’s illustration.

Based on these findings, it is possible to draw three policy conclusions in European countries:

A quota-based geographic allocation of primary care physicians is a highly effective mechanism for achieving a more equitable distribution. Yet, it is essential that this quota is determined on the basis of an estimation of regional health care needs.

In health care systems that require the registration of patients with a GP, the remuneration of GPs through capitation payments has a positive effect on an equitable distribution of GPs, especially when combined with quota-based physician allocation.

Financial bonuses for physicians practising in a remote or rural area and interventions during medical training have little or no impact.

Like all studies, there were a number of weaknesses in the analysis. First, it was only possible to compare countries using data for large population areas. Second, there were limitations regarding the depth of policy analysis. As Walt and Gilson³² point out, the content of a policy mechanism, which was the focus of this analysis, is only one of several relevant elements which shape policy effectiveness. If, for example, a mechanism was successful in achieving greater distributional equity in the short term, it could hypothetically suffer from a lack of acceptability among future GPs and, thereby, lower the attractiveness of becoming a GP in the long term, thereby inducing shortages. Therefore, Walt and colleagues^32,33 argue that a policy analysis should also include an analysis of the implementation process, the central actors involved, the acceptability of the mechanism and the wider policy context. These aspects were not included in this analysis for reasons of feasibility in a macro-comparative study. Further research is required into these elements and long-term effects.

Third, the study was based on the assumption that a lack of regional primary care suppliers results from an inequitable distribution rather than from an overall lack of physicians. Consequently, any possible structural gaps between the available and required GP workforce in the different European countries were ignored in this study. Further research could consider this, taking account of the different primary care structures of each country.

Despite these limitations, by conducting a QCA, this study was able to provide evidence on the relative effectiveness of different policies in creating a more equitable distribution of key primary care professionals, namely GPs.

References

International Conference on Primary Health Care. Declaration of Alma Ata, Geneva: World Health Organization, 1978.

Whitehead

. The concepts and principles of equity and health. Health Promot Int 1991; 6: 217–228.

Penchansky

Thomas

. The concept of access: definition and relationship to consumer satisfaction. Med Care 1981; 19: 127–140.

OECD. Health at a glance 2009: OECD indicators, Paris: Organisation for Economic Co-operation and Development, 2009.

WHO. Increasing access to health workers in remote and rural areas through improved retention: global policy recommendations, Geneva, Switzerland: World Health Organization, 2010.

Rihoux

. Qualitative comparative analysis (QCA) and related systematic comparative methods: recent advances and remaining challenges for social science research. Int Sociol 2006; 21: 679–706.

Simoens

Hurst

. The supply of physician services in OECD countries, Paris: OECD, 2006.

Grobler

Marais

Mabunda

. Interventions for increasing the proportion of health professionals practising in rural and other underserved areas. Cochrane Database Syst Rev 2009; (1): CD005314–CD005314.

Wilson

Couper

Vries

E de

. A critical review of interventions to redress the inequitable distribution of healthcare professionals to rural and remote areas. Rural Remote Health 2009; 9: 1060–1060.

10.

Dolea

Stormont

Braichet

. Evaluated strategies to increase attraction and retention of health workers in remote and rural areas. Bull World Health Organ 2010; 88: 379–385.

11.

Günther

Kürstein

Riedel-Heller

. The role of monetary and nonmonetary incentives on the choice of practice establishment: a stated preference study of young physicians in Germany. Health Serv Res 2010; 45: 212–229.

12.

Langenbrunner

Orosz

Kutzin

. Purchasing and paying providers. In: Figueras

Robinson

Jakubowski

(eds). Purchasing to improve health systems performance, Maidenhead: Open University Press, 2004, pp. 236–264.

13.

Gosden

Forland

Kristiansen

. Capitation, salary, fee-for-service and mixed systems of payment: effects on the behaviour of primary care physicians. Cochrane Database Syst Rev 2000; (3): CD002215–CD002215.

14.

Zweifel

Breyer

Kifmann

. Health economics, Berlin, Heidelberg: Springer-Verlag, 2009.

15.

Ragin

. The comparative method: moving beyond qualitative and quantative strategies, Berkeley, London: University of California Press, 1987.

16.

Schneider

Wagemann

. Qualitative comparative analysis und fuzzy sets: Ein Lehrbuch für Anwender und jene, die es werden wollen, Opladen: Budrich, 2007.

17.

Rihoux

. Qualitative comparative analysis (QCA) and related techniques: recent advances and challenges. In: Pickel

Pickel

Lauth

(eds). Methoden der vergleichenden Politik- und Sozialwissenschaft, Wiesbaden: VS Verlag für Sozialwissenschaften, 2009, pp. 365–385.

18.

EC. Regulation (EC) No 1059/2003 of the European Parliament and of the Council of 26 May 2003 on the establishment of a common classification of territorial units for statistics (NUTS).

19.

Culyer

Wagstaff

. Equity and equality in health and health care. J Health Econ 1993; 12: 431–457.

20.

van Doorslaer

Koolman

Puffer

. Equity in the use of physician visits in OECD countries: has equal treatment for equal need been achieved. In: OECD (eds). Measuring up: improving health system performance in OECD countries, Paris: OECD Publishing, 2002, pp. 225–248.

21.

Matsumoto

Inoue

Bowman

. Geographical distributions of physicians in Japan and US: impact of healthcare system on physician dispersal pattern. Health Policy 2010; 96: 255–261.

22.

GEK. GEK-Report ambulant-ärztliche Versorgung: Auswertungen der GEK-Gesundheitsberichterstattung, St. Augustin: Asgard-Verl, 2008.

23.

NHS Information Center. Trends in consultation rates in general practice 1995/1996 to 2008/2009: analysis of the QResearch® database, http://www.ic.nhs.uk/cmsincludes/_process_document.asp?sPublicationID=1251287163774&sDocID=5315 (2009, accessed 11 April 2012).

24.

Eurostat. Consultation of a medical professional during the last 4 weeks by sex, age and educational level, http://appsso.eurostat.ec.europa.eu (2008, accessed 12 April 2012).

25.

Speybroeck N, Ebener S and Sousa A. Inequality in access to human resources for health: measurement issues. Evidence and information for policy, http://www.who.int/hrh/documents/inequality_access_measurement_issues.pdf (2006, accessed 12 April 2012).

26.

Atkinson

. On the measurement of inequality. J Econ Theor 1970; 2: 244–263.

27.

Kawachi

Kennedy

. The relationship of income inequality to mortality: does the choice of indicator matter? Soc Sci Med 1997; 45: 1121–1127.

28.

European Observatory on Health Systems and Policies. Health systems in transition, http://www.euro.who.int/en/who-we-are/partners/observatory/health-systems-in-transition-hit-series (2012, accessed 6 June 2012).

29.

Ragin

. Measurement versus calibration: a set-theoretic approach. In: Box-Steffensmeier

Brady

Collier

(eds). The Oxford handbook of political methodology, Oxford: Oxford University Press, 2008, pp. 174–199.

30.

Ragin

. Set relations in social research: evaluating their consistency and coverage. Polit Anal 2006; 14: 291–310.

31.

Ragin CC and Rihoux B. fs/QCA. Tucson, AZ: University of Arizona, http://www.u.arizona.edu/∼cragin/fsQCA/software.shtml (2009, accessed 10 February 2012).

32.

Walt

Gilson

. Reforming the health sector in developing countries: the central role of policy analysis. Health Policy Plan 1994; 9: 353–370.

33.

Walt

Shiffman

Schneider

. ‘Doing’ health policy analysis: methodological and conceptual reflections and challenges. Health Policy Plan 2008; 23: 308–317.