Assessment of the priority target group of mental health service networks within a nation-wide reform of adult psychiatry in Belgium

We carried out an observational study to assess which patients’ sociodemographic and clinical variables were associated with being given priority for care in the network by clinicians. At the time of the study (2013), ten mental health service networks for adults with mental health problems were established, covering urban and rural areas across the country. All networks were included in the study. The individual networks included between 14 and 140 different services, with a total of 635, and all services were invited to participate. Services were asked to select ten users by systematic sampling. The inclusion criteria were: being adult (between 18 and 65 years of age), being in contact with the service during the data collection period, and having psychological distress. No specific diagnosis was required. All included patients gave a written informed consent for participation. Clinicians of the services, knowing the users, were asked to complete an online questionnaire about the sociodemographic and clinical characteristics of each user selected (see Fig. 2). In addition, clinicians were requested to state whether, yes or no, the patient was, at that time, considered as a priority for care within the network. The wording of the question was: “Is this patient a priority for the care circuit?” Criteria for defining the priority were deliberately left undefined in order to assess how the priority group was constructed. The priority status of users had to be discussed during a staff meeting involving clinicians and health service managers in each individual service. Staff decision determined the inclusion of users into the priority group.

The dependent variable was “inclusion in the priority group for care in the network (yes/no)”. We assessed the association between the inclusion of the user in the priority group and the user’s social and clinical vulnerability. In particular, the level of social integration was measured using a SIX score [22]. The SIX is an index that measures the social integration of long-term psychiatric adult patients in terms of employment, housing, and social relationships. Clinical information included DSM-IV diagnoses (when available), and the Health of the Nation Outcome Scale (HoNOS), which captures the severity of psychosocial malfunctioning [23].

Additional information about the type of service and network characteristics was also included. On the one hand, previous research suggested that some types of services have a broader spectrum of user profiles, which may affect their capacity to prioritise [24]. Hence, services were classified in five groups, following the functional model of the reform: [1] primary care and community mental health services, [2] crisis resolution and assertive outreach teams, [3] rehabilitation and social care services, [4] psychiatric hospitals and psychiatric wards of general hospitals, [5] sheltered accommodation services for psychiatric patients. On the other hand, some network structures may also be best adapted for some users’ specific profiles [19]. Thus, networks were classified according to their size (number of services included) and the type of services included. Four small networks included up to 41 services; four medium networks included between 54 and 75 services; and two large networks included more than 100 services. Regarding service types, two networks included a majority of primary care and community mental health services (hereafter “primary care-centred”), five networks included a majority of rehabilitation and social care services (hereafter “social care-centred”), and three networks included a majority of psychiatric hospital wards (hereafter “hospital-centred”).

Tabulations and chi-square tests described how the inclusion of users in the priority group was associated with users’ sociodemographic and clinical characteristics, as well as with service and network types. Patients were also compared across network types. Then, nested logistic regressions were used to analyse which covariates were associated with users’ priority status. All statistical analysis was performed using SAS 3.4 for Scientific Linux.

As mentioned in the introduction, there is a debate about whether networks should pursue depth of relationships (a higher degree of integration of care delivery between some services for all users) or breadth of relationships (a higher degree of integration of care delivery between all services for a narrowly-defined target group of patients) [4, 19, 25]. However, the reform programme supported the establishment of service networks that include all types of services for the needs of the whole adult population [12, 13]. Our analysis suggests that the policy may have been over-ambitious: the reform has targeted a specific group of users in practice: users with psychiatric and personality disorders and poor psycho-social functioning, i.e. a group related to chronic and severe mentally ill (SMI) patients. These SMI patients were designated as those who would be more likely to benefit from care delivered within service networks, it has been suggested elsewhere [26‐28].

At the same time, our results suggest that different types of services address different types of user profiles. Thus, services do not share a common vision about the type of users that should be a priority for care within the service network. This may generate tensions across services and, as a result, reduce the capacity of these services to collaborate and to provide care in partnership. Such tensions may affect service networks and other care coordination schemes, and might explain, for example, the reluctance of primary care and community mental health teams to get involved in service networks [24, 29‐31]. In this kind of service network, primary and community mental health care services are asked to play a key coordination role in maintaining continuity of care for SMI patients and reducing the resort to hospitals [13, 27, 28, 32]. However, they basically address a broader group of patients, which may hinder their capacity for collaboration with services specifically catering for a more severe group of patients. This coordination role is a daunting task, carrying a huge opportunity cost to clinical face-to-face activity with patients. It may also affect the patient-doctor trust through the “three-is-a-crowd” mechanism: the more additional bodies a clinician has to coordinate with, the less the patient may trust his decision-making, particularly in the context of a health care reform [33]. The feedback that we received from the social services during the data collection also confirmed this: for example, some social care services did not want to assess the mental health status of their patients, arguing either that they had no training in the use of instruments such as the HoNOS, or that they did not want to stigmatise their clients because of their mental illness.

This ambiguity is reflected in the policy framework itself. Indeed, although the reform programme is intended for the whole adult population with mental health needs, the policy blueprint of the reform used different terms and implicitly narrower definitions of the target group population [12]. Generic mental health terms (“psychological support”, “mental health”, “mental disorders”, etc.) were most frequently used to describe users and symptoms, whereas other terms related to psychiatry (“psychiatrist”, “psychiatric problems”, “mental illness”, etc.) were used to explain previous policy reform measures and to describe care supply. Thus, although the case for the new policy was made on the basis of a broad definition of mental health problems, the proposals for its actual implementation referred to more severe psychiatric disorders.

A second important finding is that users with the lowest level of social integration were not assigned priority for care in the service networks. This may appear to contradict the reform’s goal of reducing the resort to psychiatric hospitals. However, it is consistent with previous research on the deinstitutionalisation process that showed that the most severe patients were left behind [34, 35]. One possible explanation for this apparent contradiction is that the less socially integrated patients are found in sheltered accommodation and long-term residential services, and may not be perceived as eligible for care within a network of services. However, this raises an important issue and a possible key limitation to the deinstitutionalisation process that the reform is supposed to strengthen. From a system perspective, it has been suggested elsewhere that patients require different levels of care integration according to the severity of their disorders [19]: coordination of care between different services would be best adapted to patients with mild to moderate disorders, whereas the most severe patients would require fully integrated services. Our findings suggest that mental health service networks might not be an optimal solution for SMI patients with very poor social functioning.

Our study had some limitations. Firstly, participation rates in the survey varied between networks, with a minimum of 20 % of the expected number of patients. However, the global mean participation rate (62 %) and the size of the sample (2,490 patients in total) were sufficient to perform quantitative analyses. Nevertheless, a selection bias cannot be totally ruled out: in particular it is likely that the networks with the lower participation rates were also those striving the hardest to implement the reform, and possibly, those that had the greatest discrepancies in terms of target group definition. Secondly, psychiatric hospitals proved more likely to participate in the survey than social and primary care services, leading to a possible bias in patient recruitment. Primary care, community mental health, and social care services accounted for 37.1 % of the total sample of users, which could be an underestimation of their overall contribution to mental health care nationwide. However, this is also an indication of their actual involvement in the development of the service networks, and reflects their understanding of such networks’ implicit target group. Thirdly, the study design deliberately left priority status undefined. Hence, it is likely that different clinicians and staff may have used different standards and criteria to make their decision, as well as elements related to their local context.

Finally, the number of networks involved (ten) was too low for a robust analysis of network-level effects. Moreover, these networks had only been operating for 2 years at the time of the study. The reform process is still ongoing; new networks have been established since then and more services have been included. Accordingly, it is likely that practices and views of priority target groups will evolve over time.