Healthcare decentralization in the Netherlands

(1)

Healthcare decentralization

in the Netherlands

Studying the differences in quality between municipalities after the

reform

Master Thesis, 9 June 2018 Leiden University, the Netherlands

Faculty of Public Administration

Author: Bart Winkels (s1123300) Supervisor: Jelmer Schalk

(2)

1

Chapter 1: Introduction

1.1 Introduction

Twenty years ago, discussions began in the Netherlands about modernizing the General Special Medical Expenses Act (Dutch: ‘Algemene Wet Bijzondere Ziektekosten’; AWBZ) (TK, 1998/1999). In the coalition agreement of the second ‘purple cabinet’ in 1998, a specific process for modernization of the AWBZ was decided. The text of the agreement states that "The planned modernization of the AWBZ aims to achieve one system of claims that better reflects the developments in the demand for care. The room for flexibility and differentiation in the offer is increased by the claims to make the AWBZ more flexible." (Coalition

Agreement 1998: 44-46).

Twenty years later the earlier laws governing healthcare under the AWBZ have been brought under the umbrella of new laws such as the new Long-term Care Act (Dutch: Wet langdurige zorg; Wlz) and the new Youth Act (Dutch: jeugdwet), as well as previously existing laws such as the Social Support Act (Dutch: Wet Maatschappelijke Ondersteuning; WMO) and the Healthcare Insurance Act (Dutch: ‘Zorgverzekeringswet’; Zvw) 1_{. Together,} they form the current basis of the Dutch healthcare system. One of the main effects of the reorganization has been the decentralization of healthcare tasks from the national level to municipalities. The WMO is primarily intended to provide aid for citizens who experience health issues such as the elderly, the chronically ill and people with a physical or mental disability2.

Decentralization is often intended to bring the decision making process closer to those it affects, however the process was not without criticism. Municipalities especially were very concerned about the way the decentralization of tasks to municipalities would take shape. In the often extensive content reactions, municipalities refer to the lack of clarity about budgets, too much non-commitment in cooperation with health insurers and the delay of legislative process. Despite the criticism the WMO 2015 was adopted on 24 April by the House of Representatives (VNG, 2014).

1_{AWBZ in 2015 alle wijzigingen. (n.d.). Retrieved May 1, 2018, from}

https://www.zorgwijzer.nl/zorgverzekering-2015/awbz-in-2015-alle-wijzigingen 2_{WMO. (z.d.). Retrieved 23 May, 2018, from}_{https://www.zorgwijzer.nl/faq/wmo}

(4)

3

1.2 Problem definition and research question

A key governance problem with decentralizations in the Netherlands arises from the complex layers involved in Dutch governance as a result of the state being a decentralized unitary state (Breeman et al., 2010). Before examining the problem definition and research question we should go into what decentralization actually entails. Decentralization involves the transfer of tasks and powers to lower levels of government. This results in the problem that different rules may arise between municipalities. Within a unitary state coherence and coordination from above are central to the system of law. Municipal regulations may not conflict with the national laws and regulations (Breeman et al. 2010: 27). The policy drawn up by a

municipality for decentralization must thus match with the policy goals that the national government links to decentralization. This is why decentralizations are often coupled with supervisory and accountability mechanisms and policy monitors from the government. The accountability pressure for municipalities in the process decentralization can be large.

Sometimes this is at the expense of the effort required to give substance to the other requested tasks of a municipality (Boogers et al. 2008: 33).

One interesting governance aspect of the decentralization involved with the WMO is that the law does not require a specific method through which supervision of the execution of the WMO is to be conducted. Effectively, municipalities are free to choose a method of conducting the supervision. The main method of supervision, chosen by over half of the municipalities, is to have supervision carried out by the Municipal or Common Health Service (Dutch: ‘Gemeentelijke of Gemeenschappelijke Gezondheidsdienst’; GGD) (VNG, 2017). This brings us to the main research question of this thesis, as differences in supervision of the WMO could potentially result in different outcomes for the decentralization process between municipalities, such as locally different levels of quality of the health care.. Thus the key research question for this study is:

Does supervision executed by the GGD, rather than done by the municipality itself, affect the quality of healthcare in the context of decentralization, and if so, how?

1.3 Scientific and academic relevance

This research will contribute to and build upon current public administration research in a number of ways. While there is already a significant body of research done on the effects of decentralization itself (Boogers et al.; Breeman et al.), no research has so far been done on the effect of outsourcing supervision within a decentralization process. And while earlier research

(5)

4

(Clough-Gorr et al., 2015) has shown that tourism, gender, marital status, nationality, urbanization and income inequality can influence a population’s health, the effect of outsourcing or keeping supervision “in-house” has yet to be examined. If there is a link between outsourcing supervision and a population’s health then this could be used to increase the population’s health. Therefore the aim of this study is to examine the link between

supervision, healthcare and decentralization, establishing potential causality between the three, by using a pooled regression with clustered standard errors.

1.4 Thesis structure

This thesis has been designed in a series of chapters to allow for structured reading and navigation. Before examining the results of this research, some background is required to fully understand the laws of the WMO that are involved. As such the next chapter will examine the history of the WMO, as well as go into how the organizational structure of the WMO is regulated and which forms of supervision are allowed within the framework of the WMO. The following chapter will then provide a review of the relevant literature surrounding decentralization and supervision. Chapter four will outline the design and empirical approach of this research , including the operationalisation of the variables. Chapter five will present the results and analyses stemming from the conducted research. Chapter six will contain the discussion and summary of the study’s findings. Finally a complete list of references will be included.

(6)

5

Chapter 2: The case of the WMO

2.1 History of the WMO

On January 1st, 2007, the WMO was enacted. However, in the coalition agreement called Building bridges (Dutch: 'Bruggen slaan') between the political parties VVD and PvdA, signed on the 29th of October 2012, several policies were included which had the effect that, starting in 2015, a number of tasks previously assigned to the AWBZ would be terminated. These tasks were decentralized from the national government to the municipalities, and regulated in the new WMO. In April 2013 a healthcare agreement was reached between the cabinet and social partners associated with caregiving, such as labor unions and employers.. Following the care agreement cutbacks on domestic help were reduced from 75% to 40% from 2015 onwards. Recently, the government has decided on a scheme for domestic help allowance. Municipalities will receive 75 million per year in 2015 and 2016 . The purpose of the money is to stimulate the demand for domestic help in order to retain full-time

employment as much as possible. It's up to the municipality, together with the providers, to explore where possibilities for improvement and investment can be found. The main idea behind the expansion of the domestic help allowance is that municipalities are better at seeking and finding the connection with local needs. The other cutbacks decided upon by the cabinet initially remained intact. Specifically, guidance was cut by 25%, the generic discount by 25 million, transport to and from daytime activities had already been cut by more than 40% since 2013, from 2014 onwards a discount on AWBZ PGB's was cut by 2.5%; tools were cut by 50 million, and finally financial customization provision income support was cut by 750 million (VNG, 2014).

In September 2013, the government decided to transfer personal care towards the Zvw. After the government decided to transfer personal care to the Zvw the VNG Members' Meeting virtually took over the process. The VNG Members’ Meeting is a branch of the Dutch Municipalities Association (‘Dutch: Vereniging Nederlandse Gemeenten’). They unanimously adopted a resolution which called on the VNG to renegotiate with the cabinet as they considered the conditions for decentralization to no longer be acceptable due to the cuts in personal care.

After consultation with the State Secretary, a report was drafted containing positive results of the meeting, which was subsequently sent to the members of the VNG. This

(7)

6

nursing and care in the insured package prescribed on the basis of the Zvw, as well as an extra 200 million per year from 2015 and onwards. Moreover it resulted in a decision regarding district nursing, creating the policy which ensured every neighbourhood team gained the district nurse function. The report also states four functions of the district nurse, namely a preventive, signalling, coordinating and consultation function. The budget for the availability of the district nurse for participation in the social district team was earmarked in the funding system. This also applies to the funding of activities such as coordination, signalling,

coaching and individual, indicated and care-related prevention. Effectively this means that health insurers must enter into agreements with municipalities regarding the deployment of the district nurse and coordination between social support and care in the district. VNG and Health Insurers Netherlands (Dutch: ‘Zorgverzekeraars Nederland’; ZN) are developing a cooperation agenda, in which agreements about a reconciliation model between health insurers and municipalities will be decided (VNG, 2014).

However, not all members were happy with the results of the consultation between the State Secretary and the VNG. In a reaction to the consultation result two thirds of the members spoke out against the resulting deal, often with extensive motivation. This motivation showed that the municipalities, even those who opposed the results of the

consultation, still want to take the new responsibility (VNG, 2014). However, municipalities were very concerned about the way in which the decentralization of tasks to municipalities was taking shape. For instance in the often extensive content reactions, municipalities referred to the lack of clarity surrounding budgets, too much non-commitment regarding cooperation with health insurers and delays in the legislative process. The VNG board finally concluded from the reactions of its members that the conditions package for the decentralization was inadequate for a responsible introduction of the new tasks. In a letter to the cabinet dated January 17th, 2014, the VNG asked for an adequate response to the concerns of its members.

The government sent its response to the VNG on the 29th of January 2014. The cabinet argued that the financial effects of the transitional law in 2015 were to be monitored closely and that initial results would be discussed that same year. Aside from that the

response stated that agreements regarding the deployment and financing of the district nurses in the social district team were laid down in the Decision care claims (Dutch: ‘Besluit

zorgaanspraken’) and in the policy rules from the NZA. BZK and VWS proceeded to set up a system of broad periodic consultation with municipalities. They stated that this so-called

(8)

7

'Control table decentralisations' should lead to a strengthening of the control on the three decentralizations processes (VNG, 2014).

Yet the VNG was not the only institution which had its reservations regarding the decentralization process. The Council of States was also critical of a number of aspects incorporated in the bill. This criticism was largely in line with the concerns and objections previously expressed by the VNG. According to the Council of States, the bill provides a first contribution mastering the development of collective spending on long-term care (VNG, 2014).

Prior to the handling of the bill, the government entered into an care agreement with the parties D'66, Christen Unie and SGP. This led to an improvement of the financial

preconditions for the start of the WMO in 2015. Municipalities would receive an additional € 195 million in 2015 in addition to the previously promised 200 million, another € 165 million in 2016 and then finally € 50 million in 2017, decreasing to € 30 million structurally from 2019. The State would also make additional money available (rising to € 70 million from 2019) to soften the extramuralization. As a result, municipalities were able to realize a more gradual turnaround in the first years, and therefore be able to focus on innovation instead of cuts (VNG, 2014).

On 24 April the House of Representatives adopted the WMO 2015 by 102 votes in favor and 48 against. Furthermore 32 amendments and 6 motions were adopted. On the 8th July the Senate passed the bill with 38 votes in favor and 37 against (VNG, 2014).

2.2 Governance structure

The introduction of the WMO 2015 has resulted in many changes regarding how social support is regulated by law. The core purpose of the 'old' WMO and the WMO 2015 are the same: providing social support to people with disabilities, to maintain, and hopefully increase, their level of social responsibility and social participation. One of the effects of the WMO 2015 is a change in the composition of the target group. Since the bill’s introduction the composition of the target group has included more people with a mild intellectual disability and psychiatric problems, which previously fell under the AWBZ. Another change is the inclusion of new types of support, such as protected living and living guidance, as well as an increased call on social networks surrounding patients to provide aid themselves (SCP, 2017).

(9)

8

Municipalities conduct the WMO 2015. They are responsible for the design of access to the WMO 2015, the provision of care and enlisting support for the involvement of citizens in the development of the WMO policy. In addition, the municipalities have ample policy space, which has resulted in a great deal of variation between local areas in how access to the WMO 2015 is arranged and how municipalities provide for support and people (SCP, 2017).

Municipalities receive money from the government from the municipal fund to implement the WMO. Anyone who receives support from the WMO has to pay a personal contribution into this fund. The municipalities themselves determine the personal contribution for general provisions, such as services and facilities that are accessible to all elderly people. The personal contribution for tailor-made provisions, such as a wheelchair or adjustments to a home, are determined by the municipality and collected by the Central Administration Office (CAK). The relative amount of the personal contribution is also dependent on income, assets, age and family composition. There is an upper limit to the personal contribution, the amount of which may differ per patient situation. Similarly to the 'old' WMO, the law requires specific rules for personal contributions are made by a general, and thus national,

administrative order. These rules are necessary to ensure that there is a uniform system which determines income and capital dependent contributions (TK 2013 / 2014a). The height of the contribution can differ per municipality and the height of the personal contribution cannot exceed the maximum period contribution. Moreover the height of the personal contribution is not allowed to exceed the cost price paid by the municipality (SCP, 2017).

In many municipalities access to social support is organized by social

neighbourhood teams (Van Arum and Schoorl 2016). This research shows that nine out of ten municipalities worked with social neighbourhood teams (n = 234 municipalities). Smaller municipalities are more likely to not have a social (neighbourhood) team than larger ones. Schoorl’s research also examined the different models for district teams between

municipalities, placing them in four distinct categories. Model A1 is a broad-based neighbourhood team (e.g. youth plus adults, where all help questions are picked up). This model was used in half of the municipalities surveyed. Model A2 is found in around 25% of municipalities, and is also a broad-based neighbourhood team, but focussed specifically on complex and multiple help questions. Model B, which is found in slightly over 10% of municipalities, concerns domain / target group-specific teams (e.g. youth and adults separately)Finally, model C, which is found in slightly less than 10% of municipalities,, a generalist team which functions as an outpost where staff mainly refers. A very small

(10)

9

proportion of the municipalities surveyed employed a district team which did not fit one of the models. (SCP, 2017).

2.3 Supervision of the WMO

The organization and implementation of the supervision is free to be determined by municipalities. The WMO 2015 only determines that the College of B & W designates

persons who are charged with monitoring compliance with the provision by or pursuant to the law. How to specifically form and carry out the supervision, is left to the college. The

Explanatory Memorandum states that it is the responsibility of the municipalities to take care of the quality and continuity of the social support. This effectively means that municipalities need to monitor compliance with the requirements imposed on the implementation of the law and enforce it if necessary (VNG, 2017).

The law states that the executive board of a municipality points out persons that are in charge of supervising the compliance in accordance with the law (VNG, 2017). The law states supervision as the collection of information about whether an act or case meets the requirements imposed on it, forming a judgment based on this information, and when

necessary intervene in the process(VNG, 2017). In 2005 the Dutch government published the framework vision ‘Less burden, more effect. Six Principles of good supervision’(Dutch: ‘Minder last, meer effect. Zes principes van goed toezicht’). It states that supervision is

important, as it promotes compliance with policy and regulation rules. In addition, supervision provides information about the quality of the execution of public tasks by private

organizations. Based on the information gathered by supervisors, policies can be adjusted by the minister if necessary. Supervision informs the minister, parliament and society on

developments in the practice and the effects of political policy. Supervision bears the ministerial responsibility and has a social function.

Two types of supervision can be distinguished within the WMO. Firstly there is the supervision of compliance and quality. This involves monitoring compliance of the (quality) requirements as stated in the laws and regulations of the care in kind and PGB's. The law makes it possible to monitor the quality of PGBs. A number of municipalities employ social workers in a district team themselves. In these cases the municipality is a care provider themselves. The role of the municipality as referrer can also be part of an investigation in the context of supervision. Secondly, there is the supervision of legality and fraud. Municipalities must, pursuant to Article 2.1.3, paragraph 4, of the WMO 2015 in the Regulation, draw up

(11)

10

rules on combating wrongly received tailor-made facilities or PGB’s. In addition the law requires municipalities to carry out periodic re-assessments of allocated customized facilities and PGBs. The frequency at which these investigations must take place is not regulated in the WMO 2015. Municipalities can set their own priorities (SCP, 2017).

The law gives municipalities a lot of room to arrange the WMO supervision. Supervision can be divided into reactive and proactive supervision. Reactive supervision is supervising triggered by a report or a signal. Proactive supervision takes place on the basis of, for example, risk analyses or a (multi-year) planning of the supervisory activities. Although the WMO 2015 is formulated very openly when it comes to the implementation of

supervision, there are a number of general principles of good supervision that are also important for the WMO supervisor. It concerns the six principles of good supervision as formulated in the government-wide vision on supervision ‘Less burden, more effect. Six Principles of good supervision’. The six principles are:

- Selective

Municipalities and providers are hesitant about more regulatory pressure. If

supervision is necessary, the municipality will ensure it is done correctly. However, supervision is often not required, for example when providers have a relevant quality mark.

- Decisive

The WMO supervisor stimulates social support providers to do the supervision themselves to achieve the desired result. The starting point is that the WMO

supervision takes a “soft” approach where it can and a “hard” one where necessary. Supervision must be decisive, especially if the situation requires it. It is therefore important that communication with the enforcement department of the municipality occurs frequently so that measures can be taken quickly.

- Collaborative

Municipalities often cooperate on the supervision of social support. A lot of

providers of care and support work regionally and often have both young clients, as well as clients from the Wmo, Wlz and Zvw. That makes coordination and

cooperation between local WMO supervisors and national inspections necessary. This collaboration is useful because the task of supervision of the WMO 2015 for the municipality is new. By working together municipalities learn from each other

(12)

11

and from the national inspectorates and this increases the efficiency of supervision. In addition, good cooperation with fellow supervisors WMO or the IGZ is a

requirement to avoid unnecessary supervisory burden for providers. The cooperation here also concerns the cooperation with the providers of social support. Good

agreements with providers can contribute to better quality of the Social Support.

- Independent

Independence relates to the material independence of the supervisory civil servant. The WMO supervisor must conduct his actual activities independently and be independently able to execute his responsibilities without being influenced by others who, for example, are involved with policy, procurement or provision. In the WMO the responsibility for policy, purchase, financial responsibility and supervision lies with the municipality. It is therefore important that the supervision is positioned independently.

- Transparent

The WMO supervisor sets priorities. What choices does he make within the assignment of the municipality? What does the supervisor focus on? The regulator communicates about those choices with the College of B and W, with other municipalities and with the providers of social support. The WMO supervisor also makes his findings public where possible. And afterwards, the regulator justifies the choices and results achieved.

- Professional

Supervising is a profession. The supervisor is a professional. This means that the WMO supervisor reflects on the choices made, requires collegial supervision and is testable. He also permanently works on improvement in consultation with

colleagues, the College of B and W and supervised providers. It is important that municipalities encourage professionalism with their WMO supervisors. In addition, the supervisory organization and the professional group also play a role in the professionalization of supervision.

The law does not state a specific method by which supervision is to be conducted therefore municipalities are free in deciding which way supervision is applied. In November 2015 the VNG did a survey among municipalities about how they conducted the supervision. This survey led to five different possibilities. The first one is that the municipality organizes the supervision themselves. The second possibility is an organization with other municipalities

(13)

12

which conduct the supervision across the involved municipalities. The third possibility is that the supervision is carried out by the GGD. The fourth possibility is for the municipality to conduct the risk-based supervision itself or in cooperation with other municipalities, and have the emergency based supervision carried out by the GGD. Finally the last option is to have the supervision carried out by an external office (VNG, 2017). The guide states pros and cons of each of the possibilities, but it does not state the consequences for the quality of the WMO for each option. Yet it is interesting to look into that relationship, as it could be that the type of supervision affects the quality of healthcare. The range of this study was decided due to scope, as more than half of the municipalities’ preferred method of supervision to having it carried out by the GGD.

The supervision conducted by the GGD has several benefits. The first benefit arises due to the fact that the GGD is more independent than municipalities and thus better reflects the WMO. Secondly if independence is at stake the GGD supervisors can replace each other to guarantee independent supervision. Thirdly the GGD is experienced with supervision, as they prior to the institution of the WMO they were conducting supervision on childcare and toddler daycare. Lastly, the GGD can conduct the supervision spread across a wide area without having to set up a separate structure, as the GGD already works in 25 regions. Of course there are certain downsides to outsourcing the supervision to the GGD as well. For example, it is possible that local GGD personnel are closely involved in the aid of a struggling family. In such a situation GGD personnel can no longer be considered independent. To guarantee the independency and in the context of good governance the GGD can then outsource supervision to another WMO supervisor if the GGD itself is part of the tasks that are being supervised (VNG, 2017).

If supervision is conducted by the municipality itself it has the benefit of ensuring a rapidly growing understanding of the health concerns of its citizens and WMO supervision standards. Another benefit are short lines between the municipal organization and providers. However there are also some downsides to having municipalities conduct supervision. For instance, from an efficiency point of view municipality supervision is an expensive option, as properly regulating supervision requires a great deal of knowledge and attention. Specifically in the case of smaller municipalities costs may rapidly rise. Another negative aspect is that smaller municipalities may only employ one supervisor, which potentially leaves the whole supervision construction vulnerable. Similarly to the GGD, municipality supervision comes with some concerns surrounding independence. Care providers may well work in multiple

(14)

13

municipalities and if each municipality has their own supervisor this would require close cooperation with supervisors from neighbouring municipalities. Thereby it could potentially be the case that the municipality is involved in the assistance of families that are being

supervised. To guarantee independence this would require a clear independent position for the WMO supervisor (VNG, 2017).

When comparing the pros and cons of each type of supervision it is interesting to note that outsourcing the supervision to the GGD has more benefits than the other options. Thus supervision by the GGD has fewer downsides compared to municipalities, and a solution which is already in place for place for its issues regarding independence. As such it makes sense that more than half of the municipalities have opted for outsourcing their supervision to the GGD. This makes it both the most interesting type of supervision to research and the focus of this paper.

(15)

14

Chapter 3: Decentralization, supervision and healthcare quality: how

do they connect?

3.1 What is quality of healthcare?

In order to evaluate the effectiveness of the decentralization of the WMO in terms of health outcomes, and specifically the effect of the type of WMO-supervision on the healthcare quality, we need to define, first of all, what quality of healthcare is.

The WHO definition of health is a state of complete physical, mental and social well-begin and not merely the absence of disease or infirmity (WHO, 1946). Healthcare is defined as the entirety of care providers (and support staff), institutions, resources and activities that are directly aimed at maintaining and improving the state of health and / or the ability to control yourself, and at reducing, eliminate, compensate and prevent shortages (van der Meer et al., 1997; Nuy & Bex, 1986).

Healthcare expenditures as a percentage of the Gross Domestic Product has been growing rapidly over time (Rosen & Grayer, 2010). Possible reasons include the aging of the population, increases in income, influence of third party payments and technological change (Rosen & Grayer, 2010). Newhouse (1992) argued that the most important reason for the increase in spending on healthcare in the United States is improvements in medical

technology. For instance, consider that the treatment of heart attacks is much more expensive today than it was decades ago. However, treatment of a heart attack is not the same treatment as in 1950. Back in the 50s the treatment involved prescribed rest, morphine for the pain and oxygen (Cutler, 2004). The treatment has since changed radically, and since the 50s

cardiovascular mortality has declined by over half, while the probability of dying after a heart attack has decreased by nearly three-quarters (Cutler, 2004). Therefore Newhouse (2001) argues that while innovations raised expenditures they actually reduced the prices of obtaining numerous health outcomes, such as surviving hospitalization due to a heart attack.

According to Rijn (2013), not only are the expenditures rising, but people’s living requirements are changing. People generally prefer to live independently at home as long as possible (Rijn, 2013). A global trend that we can be certain of is that populations are rapidly ageing (Crimmins & Beltrán-Sánchez, 2011). And while we live longer than we have in the past, it is unclear whether these extra years are lived in good health (Chatterji et al.,2015). Recent analysis argues that relative to disease burden, the global funding burden for health

(16)

15

has shifted to younger age groups (Skirbekk et al., 2017). As populations age and the disease burden shifts more and more to chronic disorders, this imbalance will become increasingly obvious (Dieleman et al., 2016). Investment in health does not need to come at the cost of disinvesting in the health of younger age groups. People that are older need health systems that provide person-centred services which are located near their homes. These services can be appropriate for all ages. Creating systems that support caregivers, such as home care, enables care-dependent elderly to live lives of meaning and dignity, while it allows caregivers to pursue other aspirations. Such systems can create jobs and a care economy, and does not require extensive government funding (Beard et al.,2017). In addition, Elkan et al. (2001) argues that home visiting can reduce the mortality of older people. Admission to institutional care among older people can also be reduced by increased home visiting.(Elkan et al. 2001). Beswick (2008) argues that community-based interventions lead to a decrease in the number of hospital admissions for older people. Some benefits of homecare include reduction in outpatient visits, quicker patient discharge, involvement of patients in their own care and reduced risk of cross-infection (Baker & Swana, 2017). Patel et al. (2015) argues that the cost of hospital treatment compared to home-based therapy is higher. This finding is consistent with World Health Organization guidelines which examines the effectiveness of managing pneumonia in a home setting (Patel et al. 2015). The literature stated above shows that homecare can increase the quality of healthcare and moreover that it does not requires extensive government funding.

Expenditures in healthcare have risen, but, as argued above, the quality has increased. As stated above home care can tackle those expenditures and maintain or even improve quality. How can quality in healthcare be measured, specifically in home care? Hirdes et al. (2004) answered this question by giving two recommended indicators. The twenty-two indicators are separated in twenty-two groups, prevalence and failure to improve/incidence. The review of the health indicators was focused on the following considerations. The first

consideration is about the relevance of the measure in regard of the quality in home care, second if the prevalence or incidence rates were retained because of their clinical importance and lastly if variability between agencies would be of little interest (Hirdes et al.,2004). Therefore Hirdes health indicators can be used to evaluate the quality in home care.

(17)

16

3.2 The relationship between supervision and decentralization

In order to evaluate the decentralization of the WMO, and specifically the effect of the type of WMO-supervision on the quality of the WMO, we need to examine what supervision is and how it relates to decentralization.

The Netherlands is a decentralized unitary state (Breeman et al., 2010).

Decentralization involves the transfer of tasks and powers to lower levels of government. The result is that different rules may arise between municipalities. In a unitary state coherence and coordination from above are central. Municipal regulations may not conflict with the national laws and regulations (Breeman et al. 2010: 27).

In the relationship between decentralization and a unitary state, the concepts of autonomy, co-management and supervision are central (Breeman et al. 2010: 27). Municipalities and provinces have the authority to carry out duties in their own territory. Municipalities and provinces can act within their own jurisdiction and determine their own rules. These rules are called autonomous regulations. These autonomous regulations may differ per municipality. The competence of province and the municipality to establish its own rules is called autonomy (Breeman et al. 2010: 27-28).

Municipalities and provinces also draw up rules on behalf of a higher regulation. An example of this is that municipalities must draw up zoning plans on the basis of the spatial planning law (Dutch: ‘Wet ruimtelijke ordening’). Municipalities and provinces are usually free to determine the content, but must take account of higher order rules (Breeman et al. 2010: 28). Decentralization within the unitary state is determined by autonomy and co-management. Supervision takes place from above, at the level of the national government. The national government can destroy all decisions made by lower authorities, for instance in cases where decisions are in conflict with the national law or general interest. The term general interest is described as insurance of the unity of government policy. Yet the national government is very cautious in intervening in this area and only sporadically uses these powers (Breeman et al. 2010: 28).

The policy drawn up by a municipality for decentralization must match with policy goals that the national government links to decentralization. That's why decentralizations are coupled with supervisory and accountability mechanisms and policy monitors from the government. The accountability pressure for municipalities in decentralizations can be large.

(18)

17

Sometimes this is at the expense of the effort required to give substance to the other requested tasks of a municipality (Boogers et al. 2008: 33).

Research by Boogers et al. among civil servants of municipalities shows that two thirds of the respondents believe that the accountability pressure for decentralizations is disproportionately large compared to the actual size of the decentralized tasks. Eighty percent of respondents believe that the pressure has increased over the past ten years (Boogers et al. 2008: 33). A similar percentage is of the opinion that the mechanisms of accountability and supervision from the central government take too little account of the local conditions of municipalities. In the case of supervisory and accountability mechanisms the focus is on the situation in large cities. No attention is given to the problems faced by small municipalities in the process of decentralization. The research shows that in small municipalities (below 50,000 inhabitants) the feedback on accountability and monitoring data are of limited use. In larger municipalities, information about the quality of the task performance is considered more useful. In most cases this also leads to an adaptation of the policy in larger municipalities (Boogers et al. 2008: 34).

The strong accountability pressure comes mainly from a lack of trust from the

government, which lack confidence that municipalities are able to perform decentralized tasks correctly (Boogers et al. 2008: 34). To assuage the fear that decentralization will be at the expense of the quality of the policy, and that decentralization will have unintended negative side-effects, the government will conduct major supervision and create accountability instruments. After some time, when it is clear how municipalities carry out the transferred tasks and the policy adjustments are implemented, there will be more confidence in local government. Supervision and accountability are then relaxed (Boogers et al. 2008: 34).

The literature argues that decentralization entails supervision from the central government on local governing bodies. Yet supervision can also lead to increased accountability pressure. Accountability pressure is defined as the pressure placed on

professionals to comply with rules, and the freedom restrictions that this entails. The concept of accountability pressure is often used in practice as a synonym for the concept of

administrative burdens (De Vries et. Al. 2006). This definition is analogous to the definition of 'regulatory pressure', which was drawn up in 2006 by an interdepartmental working group within the framework of the 'usable legal system' program. The definition of regulatory pressure is 'the investment and effort that citizens, companies and institutions have to perform

(19)

18

and the freedom restriction they undergo to comply with regulations' (Interdepartementale werkgroep regeldruk, In regels gevangen, 2006).

The problem therefore lies not only in the existence of rules, but also in the obligation to comply with them. Accountability is not only caused by the existence of rules to which professionals must comply, but also by the (changing) context in which those professionals work. For example, tasks and responsibilities in healthcare and education are increasingly being decentralized to more executive levels, such as institutions and schools. This results in pressure on the decentralized institutions to account for what they do. On balance, this can lead to an increase in accountability for professionals, and thus to a limitation of the freedom of action of these professionals. A healthcare professional who is faced with the duty to account for what he does, has less time for his primary tasks (De Vries et. Al. 2006).

Several studies by research and consultancy firms conclude that the (experienced) regulatory pressure among healthcare providers is high (Ernst & Young 2012, Care

Committee No Market 2013, Sira Consulting 2015, PWC 2016). For example, a recent survey among healthcare professionals in PWC's long-term care shows that professionals spend 62.7% of their time on clients in the current situation. In the ideal situation, that is 69.4%. As such reducing regulatory pressure likely results in more time spent on clients. . At the same time, the interviewed professionals also indicate that they need time for training and

development and that overhead is also needed to deliver good care to some extent (PWC 2016).

Regulatory pressure has an impact on the quality of care delivered. As stated, rules can contribute to good quality of care, but if regulatory pressure arises, rules can also reduce the quality of care. This partly has to do with the fact that rules can create registration pressure; if care workers are busy with all kinds of registrations of activities, they cannot spend this time on direct healthcare (Bland 2007, Banerjee, Armstrong et al. 2015). In addition, rules regulate care in a way that does not always provide the best quality of care for the patient. Rules have a reductionistic effect; they emphasize certain aspects of quality of care, often shifting focus on what can be made measurable. These are not always the aspects of care that are paramount for patients (van de Bovenkamp & Dwarswaard 2017). For example, rules and routines from the organization can ensure that patients have little room to make choices which are important to them. For example, when patients choose to do things that make them happier but which go against the treatment prescribed in guidelines (for example less dialysis because you want to

(20)

19

work or occasionally drink a beer while you know that you need extra dialysis) then they may clash with the standards of their healthcare providers. If caregivers are convinced by the patient to go for an alternative treatment choice, they are often hesitant, as they are accountable if the treatment goes wrong (van de Bovenkamp & Dwarswaard 2017).

The literature about supervision says that decentralizations are coupled with supervisory and accountability mechanisms. This is because the policies drawn by a municipality must match with the policy goals of the national government. Such

supervisory/accountability mechanisms increase the accountability pressure. Moreover it states that if accountability and regulatory pressure arises it can potentially reduce the quality of healthcare.

(21)

20

3.3 Earlier research on the effect of the WMO on the quality of healthcare

In order to evaluate the decentralization of the WMO and especially the effect of the type of WMO-supervision on the quality of the WMO we will examine the earlier research conducted on the WMO.

The Netherlands Institute for Social Research (Dutch: ‘Sociaal en Cultureel Planbureau’; SCP) conducted research on the WMO and came up with several

recommendations. One of the recommendations concerns the lack of knowledge among local authorities/assessment officers about certain distinct groups (SCP, 2017). This lack of

knowledge can obviously influence the quality of the WMO. When assessing someone you do not have the knowledge for can be dangerous. It could lead to wrong assessments which has an direct effect on the quality of the care received.

In the research of the SCP (2017) only a small proportion of the applicants used a digital application. This suggests that digital contact is not the best way to communicate with the WMO target group. Therefore local authorities should be wary of focusing solely on digital communication, as only communicating through digital means results in less people being helped. (SCP, 2017). Other recommendations are about the informal carers, many of whom felt unheard and felt a desire for a greater say in the overall care provided (SCP, 2017). Informal carers can influence the quality of the WMO, because when they feel noticed and have a say in the overall care they are more willing to perform the tasks asked of them. This is especially important in the allocation of care times and in determining who provides care how often. Almost sixty percent of the informal carers felt that they needed some form of support that they were not receiving. Often this meant financial support, as well as a replacement who sometimes could take over the care from them (SCP, 2017).

Unaffordability and absence of (extra) support from their own network are the main reasons for applicants to be dissatisfied with the solution that is offered to them by the WMO consultant (SCP, 2017). A lack of income/wealth can influence the quality of the WMO, because if an applicant cannot afford the additional costs they will not receive treatment. A lack in social networks can influence the quality as well, because if there is a shortage of informal carers, patients often rely more on formal care. Yet formal care is not always an option, which can lead to dissatisfied applicants and thus a lack of care quality.

Research from van der Heijden & Schalk (2016) finds that investing in relationships with client interest organizations is beneficial to achieving the service outcomes as described

(22)

21

in the formulated policy goals. Heijden & Schalk (2016) found a positive indirect effect through physical self-reliance has been found. A possible explanation is that client-interested networking can support a client’s personal network. This support helps their network (family and friends) in fulfilling their role as informal caregivers.

Other research from Schalk et. al. (2014) outlines key recommendations for

municipalities concerning how to handle the WMO. One of the recommendations is to create synergies by cooperating with the similar municipalities across social domains. Schalk’s research (2014) also found that there are well-established patterns of cooperation between service providers and municipalities. Schalk et al. (2014) expect that after the transition there will be more manoeuvring room for new cooperation agreements with service providers that have currently remained below the surface. The degree of networking with other

municipalities and/or service providers can influence the quality of the WMO. Schalk et. al. (2014) recommends creating synergy by cooperating with other municipalities. Van der Heijden & Schalk (2016) found that investing in relationships with client interest

organizations is beneficial to the service outcomes. Better networking with municipalities and/or service providers (e.g. client interest organizations) should lead to a higher quality of the WMO.

Earlier research (Schalk et al., 2014) provided several recommendations to improve the quality of the WMO. The only one that can relate to supervision of the GGD is that better networking can lead to a higher quality of the WMO. If the GGD conducts the supervision then this can be seen as a form of networking.

The literature argues that decentralizations are coupled with supervisory and

accountability mechanisms and policy monitors from the government. These mechanisms can lead to large accountability pressures for municipalities. Sometimes this is at the expense of the effort to give enough focus to the requested tasks of a municipality (Boogers et al. 2008: 33). Accountability pressure can be seen as the pressure placed on professionals to comply with rules, and the freedom restriction that this entails (De Vries et. Al. 2006). This pressure leads to a limitation of the freedom of action of these professionals. A healthcare professional who is faced with the duty to account for what he does, has less time for his primary tasks (De Vries et. Al. 2006). All these rules can contribute to good quality of care, but if regulatory pressure arises, rules can also reduce the quality of care (Bland 2007, Banerjee, Armstrong et al. 2015). In the case of the WMO municipalities can choose who will conduct the

(23)

22

supervision. Van der Heijden & Schalk (2016) found that investing in relationships with client interest organizations (e.g. GGD) are beneficial to service outcomes. Therefore one can argue that if the WMO-supervision is conducted by the GGD instead of the municipality itself they experience less accountability pressure. The decrease in pressure could lead to more time for care workers to spend on direct healthcare and therefore could lead to an increase in the quality of healthcare. Therefore this thesis hypothesis is:

If the supervision is formally executed by the GGD instead of the municipality itself, the quality of healthcare is higher.

No previous research has studied this relationship. The next section will go into the research design which was created to, research this relationship.

(24)

23

Chapter 4: Research design

4.1 Case selection

This study’s focus is on Dutch municipalities. The number of municipalities examined was 388 when the data was retrieved in November 2017. Therefore, these 388 municipalities are the units of analysis in this study.

4.2 Research and data collection

Data about the municipalities has been found through a website which is an initiative of the VNG and the Quality Institute for Dutch Municipalities (Dutch: ‘Kwaliteitsinstituut

Nederlandse Gemeenten’; KING)3_{. VNG is an association of Dutch municipalities and KING} is a quality assurance institute for the Dutch municipalities. The data they have is retrieved from several sources, such as the Central Bureau of Statistics (Dutch: ‘Centraal Bureau voor de Statistiek; CBS’), the GGD and citizen polling (Dutch: ‘burgerpeiling’). Municipalities deliver data on their own as well. To examine how the municipalities conducted their

supervision, the GGD GHOR was contacted for data. They provided information concerning the municipalities in which they conduct the supervision. They conduct the supervision for 235 municipalities. The variable GGD supervision is a dummy variable with 1 if the GGD conducts the supervision and 0 if the GGD does not conduct the supervision.

4.3 Operationalisation of key variables

In order to evaluate the effect of supervision by the GGD on different aspects of the quality of care, several dependent variables are analysed. There are four dependent variables and each of them covers a different aspect of the quality of the WMO. According to Rijn (2013), people’s living requirements are changing and people would like to live at home as long as possible. In regard to these changing living requirements it is essential to see if these changed

requirements result in better care. Therefore the first dependent variable is the subjective

health score residents give themselves. This is a score between 1 and 10. The data for this

variable is available for 47 municipalities in 2014 and for 46 municipalities in 2015.

How else can quality in healthcare be measured, specifically when it concerns home care? Hirdes et al. (2004) answered this question by giving twenty-two recommended indicators. Two of the prevalence indicators are the number of falls and hospitalization. Therefore the second dependent variable is the number of 65 plus residents that fall with

(25)

24

hospitalization as a result. The data of this variable is available for 388 municipalities in 2014

and for 350 municipalities in 2015. We expect that the number of falls is lower in the municipalities where the GGD conducts the supervision.

Working together with service providers should lead to a higher quality of the WMO (Heijden & Schalk, 2016). Therefore the third dependent variable is the percentage of

residents that are satisfied with the health services. The data of this variable is available for 47 municipalities in 2014 and for 50 municipalities in 2015. This variable is interesting, as residents themselves actually rate the care they receive. We expect that more residents are satisfied if the municipality works together with the GGD.

The classic decentralization argument entails that local decision makers’ access to information on local conditions is better than central authorities (Tiebout, 1956), which enables them to make more effective policy decisions. For example this knowledge allows for better services and public spending in accordance with the local needs and preferences

(Musgrave, 1959). Therefore the fourth dependent variable is the percentage of residents that have confidence in the way the municipality is governed. The data for this variable is

available for 47 municipalities in 2014 and for 46 in 2015. This variable is interesting,

because it says something about the trust of the residents in their municipality. We expect that after the reform there is more confidence in the way the municipality is governed than before the reform. Almost all the dependent variables have data available for 2016 only the number of falls has no data available for 2016.

The treatment variable is in line with the hypothesis that if the supervision is formally executed by the GGD instead of the municipality itself, the quality of healthcare is higher. Therefore the treatment variable is supervision conducted by the GGD, yes or no. Because it has only two options, yes or no, it is a dummy variable. We expect that all four dependent variables are positively correlated with the treatment variable.

All the independent variables have an N of 388 and are available for 2014 and 2015. The first independent variable is population and this is the total population of an municipality divided by 1.000. The second independent variable is age and it is the average age of

residents in an municipality. The third dependent variable is grey pressure and this the percentage of 65 year and older residents respective to the percentage of 15-64 year old residents. The fourth independent variable is immigrants, this is the percentage of immigrants compared to the entire population of a municipality. The fifth independent variable is low

(26)

25

educated this is the percentage of people per 1.000 inhabitants that have an low education.

The sixth independent variable is middle educated this is the percentage of people per 1.000 inhabitants that have an average education. The seventh independent variable is high educated this is the percentage of people per 1.000 inhabitants with a high education. The eighth

independent variable is expenditure per inhabitant this are the expenditures in regard of health and environment of an municipality. The total expenditures are divided between all

inhabitants, which gives the number of euros per inhabitant. The ninth and last independent variable is debt per inhabitant this is the net debt of a municipality. The total debt is divided between all inhabitants, which gives the number of euros per inhabitant.

Almost all the independent variables have data available for 2016 as well, only the charges in regard of health and environment of an municipality and the net debt of an municipality have no data available for 2016.

(27)

26

Chapter 5: Research results

5.1 Analytical strategy

In order to test the hypothesis, we should first define the structure of our data. The observed data has a set of observations at a different point of time and it is collected for several individuals (municipalities). This is called panel data and therefore it could be that although the same municipalities are analyses in multiple periods, the values of the variables in the analysis are averages, which can be based on slightly different populations. Yet we can assume that variance of each population is the same for all municipalities. To gain a higher precision estimate of variance than the individual sample variances the data for the years 2014 -2016 is pooled for municipalities. The higher precision leads to increased statistical power. Since we pool the panel data the dataset can be seen as pooled panel data. In order to test pooled panel data, several multiple regression analyses can be used. Regression analysis is a technique that estimates the relationships among variables. Therefore it helps to understand why the value of the dependent variables changes when one of the independent variables changes, while the rest of the independent variables are held constant. It can be used to understand the relationship between the dependent and independent variables4_{. The problem} with this analysis is that it assumes that the quality of the WMO is completely independent from one municipality to the next. This is unlikely to be true since a lot of municipalities are working together in regards to the WMO5. In order to generate valid standard errors and homoscedasticity a regression clustered on municipalities can be the solution. By clustering the data on municipalities it would yield the same regression coefficients but it allows for differences in the standard errors due to arbitrary intra-group correlation (Cameron & Miller, 2013). In order to test for a difference before and after the reform a variable for time is added. This variable is a dummy variable that is one for 2015 and zero for 2014. Given that we create two groups for time and we cluster for municipalities we call the analysis conducted a pooled regression with clustered standard errors.

4_{What is Multiple Linear Regression? (n.d.). Retrieved May 1, 2018, from} https://www.statisticssolutions.com/what-is-multiple-linear-regression/

5 _{De Groen, B. (n.d.). Intergemeentelijke samenwerkingsverbanden in de Wmo. Retrieved May 1, 2018, from} https://www.vilans.nl/producten/intergemeentelijke-samenwerkingsverbanden-in-de-wmo

(28)

27

The dependent variables are tested in different models to uncover the relationship between the key independent variable GGD supervision and the dependent variables. The four dependent variables encompass the quality of care with respect to the WMO.

For some dependent variables, the number of cases (municipalities) is limited. To determine how many independent variables can be used in single model, the ‘one in ten rule’ (Peduzzi et al., 1996) is used. This is a rule of thumb for how many independent variables can be used in a regression analysis. The lowest number of cases for any dependent variable in the analysis is 46 (dependent variables health services and confidence). Therefore we can use four independent variables in these regression models.

Because there are nine independent variables, these cannot be included in the same model. Therefore, we cluster these independent variables into three meaningful subgroups, namely demographic, educational, and financial. Demographic variables include population, age, grey pressure, and immigrants. Educational variables include the percentages low educated, middle levels of education, and highly educated. Financial variables, finally, include charges and debt.

In the first model, subjective health score is predicted by GGD supervision, in three different regressions. In each regression another group of independent variables is taken into account. The sample size is low and this gives us two problems. The first is external validity, because there are probably not enough cases in the sample to be representative. The second problem concerns the causal analysis, whether or not there are enough cases to show a

difference between the treatment and control groups (Toshkov, 2016). The statistical power of this variable is likely relatively low. However this variable is interesting to evaluate, because it actually rates how healthy residents are feeling.

In the second model, the number of falls that lead to hospitalization is tested against the treatment variable in three different regressions. In each regression another group of independent variables is taken into account. This variable has a big sample size and therefore the statistical power is likely higher. However while the statistical power is high the

explanatory power of this variable can be questioned. This is because the number of falls can be explained by other things than the quality of healthcare. For example the number of falls can be higher within an municipality with a lower budget for housing. Therefore the number of falls cannot be directly related to quality of healthcare. However it is a prevalence indicator as stated by Hirdes et. al. (2004).

(29)

28

In the third model the percentage of residents which are satisfied with the health services is tested against the treatment variable in three different regressions. In each regression another group of independent variables is taken into account. The sample size is low and this gives us the same problems as in model 1, which results in a lower statistical power. However this variable is interesting to evaluate, because residents actually rate the health services themselves.

In the fourth model the percentage of residents that have a lot of confidence in the way the municipality is governed is tested the against the treatment variable in three different regressions. In each regression another group of independent variables is taken into account. The sample size is low and this gives us the same problems as in model 1 which has resulted in lower statistical power. However this variable is interesting to evaluate, because it shows us if residents are happier with the way the municipality is governed after the reform and if enlisting the GGD as a supervisor has an effect on confidence levels.

5.2 Results

In order to evaluate the hypothesis, the regression was completed for all dependent variables. GGD supervision was significant in only one of the twelve regressions (B= -3.128; p < .05). It was significant in the regression of rating of health services with demographic independent variables. It says that having supervision conducted by the GGD decreases the rating of health services by 0,03. In all tables HScore refers to the dependent variable subjective health score,

HService refers to the dependent variable health services, Confi refers to the dependent

variable confidence and falls refers to the dependent variable number of 65 plus residents that fall with hospitalization as a result. Moreover Demo refers in all tables to the set of

demographic control variables, Educ refers to the set of educational control variables and

(30)

29

Table 1. Pooled regression analysis of quality of healthcare in the context of the WMO (2014-2015). HScore Demo (n=93) HScore Educ (n=93) HScore Fina (n=93) HService Demo (n=97) HService Educ (n=97) HService Fina (n=97) Supervision -0.033 (0.037) 0.025 (0.040) 0.022 (0.045) -3.128** (1.570) -1.613 (1.799) -1.564 (1.920) Age -.053*** (0.013) -3.343*** (0.751) Grey pressure 0.013** (0.005) 1.128*** (0.267) Number of inhabitants -0.003*** (0.001) -0.038 (0.024) Number of immigrants 0.704** (0.276) 15.308 (11.184)

Percentage Low educated -0.002

(0.004) -0.406* (0.220) Percentage Middle educated 0.002 (0.004) 0.075 (0.185)

Percentage High educated 0.008***

(0.002) 0.148 (0.158) Charges 0.000 (0.000) 0.001 (0.015) Debt -0.000 (0.000) -0.000 (0.001) Year dummy -0.048 (0.032) -0.082** (0.031) -0.081** (0.035) 2.038 (1.476) 1.161 (1.54) 0.998 (1.712) Constant 9.508*** (0.461) 7.454*** (0.206) 7.681*** (0.052) 186.9*** (24.780) 82.5*** (12.743) 81.4*** (4.157) R-squared 0.353 0.233 0.087 0.240 0.137 0.018

Notes: Errors are allowed to be correlated within clusters (municipalities) ***p < 0.01

**p < 0.05 *p < 0.10

How is it that supervision was only found to be significant in one model? When researching this question an answer was found within the model which was significant. The only other significant variables were age and grey pressure. Therefore I started researching the

interaction effect between age and supervision. I plotted an interaction table between age and supervision. The crossed lines on the graph suggest that there is an interaction effect. The graph shows that health service satisfaction is higher for younger aged municipalities if the supervision is conducted by the GGD. It also shows that health service satisfaction is lower for higher aged municipalities with GGD supervision. For example at an average age of 30 the percentage of people satisfied with the health services in their municipality is 90 percent if the GGD conducts the supervision. If the GGD does not conduct the supervision and the average age is 30 then the percentage is around 87 percent. Meanwhile at an average age of 48 the average satisfaction is higher for municipalities without supervision by the GGD (79%

(31)

30

versus 76%). This suggests that the effect of supervision by the GGD depends on the average age of the municipality. The graph shows that at an average age of 40 there is no difference between municipalities with or without supervision by the GGD. Therefore an interaction dummy variable was created; if the age is 40 or older and the GGD conducts supervision in the municipality.

(32)

31

Table 2. Pooled regression analysis of quality of healthcare in the context of the WMO (2014-2015). Confi Demo (n=93) Confi Educ (n=93) Confi Fina (n=93) Falls Demo (n=734) Falls Educ (n=734) Falls Fina (n=734) Supervision -1.932 (2.900) 0.044 (2.920) 0.003 (2.943) 7.650 (9.659) 8.901 (9.368) 12.690 (9.397) Age -3.380*** (0.980) -0.594 (3.002) Grey pressure 0.951** (0.382) 0.736 (1.212) Number of inhabitants -0.063 (0.039) 0.010 (0.041) Number of immigrants 5.236 (14.773) 110.6*** (39.802)

Percentage Low educated 0.203

(0.272) -0.308 (0.640) Percentage Middle educated -0.045 (0.263) -2.68*** (0.566)

(0.182) -0.623* (0.375) Charges -0.021 (0.015) 0.137*** (0.049) Debt -0.000 (0.001) -0.001 (0.002) Year dummy 4.315* (2.352) 2.580 (2.295) 2.837 (2.253) 50.8*** (6.970) 51.043*** (6.617) 49.053*** (6.637) Constant 141.5*** (30.799) 14.369 (15.525) 32.651*** (4.209) 116.959 (94.070) 248.5*** (26.687) 95.143*** (12.923) R-squared 0.180 0.090 0.053 0.144 0.163 0.1439

Notes: Errors are allowed to be correlated within clusters (municipalities) ***p < 0.01

**p < 0.05 *p < 0.10

Including this interaction variable in all models gave some interesting results. In eight of the 12 models the interaction variable was significant (p < .10). These results show that there is an interaction effect between age and supervision. Implementing the interaction variable gave three significant (p < .10) results for supervision. Results show that in the regression the rating of health services is not significant anymore, while age and grey are still significant in this model. This suggests that age is a negative confounder which means that the observed association was biased towards the null. In the regressions with other dependent variables age was a positive confounder which means that the observed association was biased away from the null. The R-squared of the models stayed the same in model 2.1 (HService Demo) and in all other models the R-squared increased. This means that with the addition of the interaction variable the models explain more of the variation of the variables than without the interaction variable.

(33)

32

Table 3. Pooled regression analysis of quality of healthcare in the context of the WMO with the

interaction variable included (2014-2015).

HScore Demo (n=93) HScore Educ (n=93) HScore Fina (n=93) HService Demo (n=97) HService Educ (n=97) HService Fina (n=97) Supervision -0.014 (0.085) 0.162** (0.073) 0.145 (0.090) -3.172 (3.655) 2.896 (3.157) 2.433 (3.540) Age40*supervision -0.022 (0.083) -0.180** (0.073) -0.161* (0.088) 0.051 (3.745) -5.719* (3.118) -5.051 (3.518) Age -.051*** (0.015) -3.347*** (0.744) Grey pressure 0.013** (0.005) 1.129*** (0.265) Number of inhabitants -0.003*** (0.001) -0.038 (0.024) Number of Immigrants 0.684** (0.269) 15.352 (11.127)

Percentage Low educated -0.001

(0.004)

-0.382* (0.214)

Percentage Middle educated 0.003

(0.005)

0.102 (0.183)

(0.002) 0.182 (0.135) Charges 0.000 (0.000) 0.003 (0.014) Debt -0.000 (0.000) -0.000 (0.001) Year Dummy -0.049 (0.033) -0.082*** (0.031) -0.081** (0.035) 2.040 (1.463) 1.160 (1.545) 0.998 (1.723) Constant 9.442*** (0.496) 7.362*** (0.215) 7.666*** (0.051) 187.1*** (24.891) 80.2*** (11.916) 80.8*** (3.902) R-squared 0.354 0.311 0.150 0.240 0.173 0.040

Notes: Errors are allowed to be correlated within clusters (municipalities)

***p < 0.01 **p < 0.05 *p < 0.10

With the addition of the interaction variable the hypothesis can be evaluated. The hypothesis was:

If the supervision is formally executed by the GGD instead of the municipality itself, the quality of healthcare is higher.

In order to evaluate the hypothesis four dependent variables were chosen. Only the dependent variable confidence seems to have a correlation with supervision conducted by the GGD (see table 2). There seems to be a positive effect on confidence in a municipality when the GGD conducts the supervision. Even more interesting is that there seems to be a negative effect on confidence in a municipality when the GGD conducts the supervision and the average age is 40 or older in a municipality. This suggests that when the GGD conducts the supervision this

Healthcare decentralization in the Netherlands