Mental health service diversity and work disability: associations of mental health service system characteristics and mood disorder disability pensioning in Finland

The study data consisted of three integrated data sets: (1) data on Finnish DP for mental disorders; (2) ESMS-R data on regional MHS from 113 Finnish municipalities within seven hospital districts; and (3) demographic information for the Finnish municipalities. The original mental disorder DP study data included all Finnish citizens granted either a temporary or permanent DP due to a mental disorder (ICD 10: F04-F69, F80-F99) for the first time between 2010 and 2015 (N = 50,728). The study data was collected from the registers of Statistics Finland, the Social Insurance Institution of Finland, the Finnish Centre for Pensions and the Finnish Institute for Health and Welfare (THL). During preliminary data analysis, we excluded the following subjects from the data: (1) individuals with any previous DP; (2) individuals aged under 18 or over 65; (3) individuals who had moved to a new hospital district during the last three years before receiving DP. After this exclusion process, the data included 36 879 subjects with a mental disorder DP. For a more detailed data analysis, see our previous study [33]. Lastly, for this study, we excluded the DP recipients who had been granted their DP for other than F30-39 mood disorder as their primary diagnosis and recipients living in other municipalities than those in the study area. Thus, the final data set included 13 783 first-time mood disorder DP recipients. The municipalities’ demographic characteristics were collected for 2015 from the Sotkanet Indicator Bank, an information portal provided by THL that offers essential population health and welfare data [37].

The seven hospital districts comprising the study area were Helsinki and Uusimaa (HUS), Kymenlaakso, South Karelia, Southwest Finland, Pirkanmaa, Kainuu and Lapland (Fig. 1). These hospital districts included 113 municipalities from which we excluded those with a population less than 2000 inhabitants, with the final data set having 104 municipalities. These municipalities cover approximately 60% of all Finnish citizens aged 18–64. The DP risk of these regions was reported in a previous study, which indicated that these districts are a representative sample of different Finnish regions with mostly stable DP risk [15].

The MHS of the study area were analyzed using the ESMS-R -tool [16, 17]. The ESMS-R's hierarchical taxonomy-based coding tree is described in Online Resource 1. The MHS data from HUS, Kymenlaakso, South Karelia and Southwest Finland was collected during 2012–2013 for the REFINEMENT project (Research on Financing Systems’ Effect on the Quality of Mental Health Care in Europe; for example [19, 29]). MHS in the Pirkanmaa, Lapland and Kainuu hospital districts were analyzed retrospectively for the years 2013 and 2014. These years 2012–2014 correspond with the DP data timespan of 2010–2015 in this study, as no major alterations were made to these municipal MHS during these years. This ESMS-R data is the most comprehensive available data concerning MHS system characteristics and resources in Finland.

The classification of local vs. centralized services is not initially coded in the ESMS-R taxonomy. We used a categorizing variable designed by Ala-Nikkola et al. [38] to identify local services with and without gatekeeping and centralized services, which were reclassified from the existing ESMS-R data. The Local Service variable was created using a modified Delphi panel procedure. Thus, five different MHS types from the ESMS-R -data were studied: (1) all MHS, (2) outpatient care (ESMS-R class O), (3) local services without gatekeeping, (4) local services with gatekeeping, and (5) centralized services. Outpatient services included only those services where patients were seen in an outpatient setting without the services being residential or day care services. Local services without gatekeeping included three service classes for outpatient care and four classes for day care. It also included almost all the information for care, accessibility to care and self-help and voluntary care services. Local services with gatekeeping included most of the outpatient care services, but also five classes of day care services, one information for care and one self-help and voluntary care service class. Centralized services mainly comprised day care and residential care and one type of regionally concentrated special outpatient care service class. For further details, see [38].

For these five different types of MHS, three different ESMS-R -service system characteristic factors were used in the analysis: (1) service resources as the number of personnel in full-time equivalents (FTE) allocated by municipality population, per 1000 inhabitants, (2) service richness as all the different ESMS-R classes available for the municipality’s inhabitants and (3) service diversity as the Gini-Simpson Diversity Index (GSDI) calculated with service richness and the available units for the municipalities [39, 40]. The formulation of MHS types and factors from the ESMS-R data is displayed in Fig. 2.

We first calculated the sum of FTE for each municipality to evaluate the effect of the number of personnel in the MHSs. In cases where the same MHS unit provided services to inhabitants in several municipalities, the unit’s FTE was divided and allocated based on the proportion of the population in each municipality (for example, MHS X provided services with an FTE of Y to four municipalities, one of them being municipality Z. Municipality Z’s inhabitants consisted of 20% of all the persons of the area to whom MHS X provided services. Thus 20% of the Y was allocated to the municipality Z’s FTE.) Secondly, we divided the sum of FTE for each municipality by 1000 inhabitants.

Although service richness can be used as a simple way to indicate diversity, it only reflects the number of service classes reported, regardless of the number of different available service points provided for the municipalities’ inhabitants. GSDI and similar diversity indices are commonly used for example in ecological and ecosystem service research to calculate species or class diversity in a given environment or area [41]. By combining and weighing service richness with the number of available units in the ESMS-R class, the GSDI defines an index of 0 to 1 for the municipalities’ MHS diversity, with a higher GSDI signifying higher diversity. The calculation of GSDI and an example are given in Online Resource 2. The GSDI gives a more multifaceted approach to diversity of services, considering evenness between service provision rather than the mere number of ESMS-R classes commonly used in previous MHS research and named service richness in this study. To our knowledge this study is the first to examine MHS diversity using the GSDI, made possible by the ESMS-R classification of MHS classes.

Services classified as 24 h service housing without a fixed term (ESMS-R classes R11-13) were excluded from the study. These residential services presumably do not affect the disability pensioning process, as they are targeted primarily to people already on a disability pension. Furthermore, the person-years of staff information was missing from 174 services, which were excluded from the analysis regarding the service resources. However, the services with missing person-years of staff information consisted mostly of 71.3% self-help and volunteer care services (ESMS-R class S), with only 6.9% outpatient care (O) services. The final mental health service data set included 1 088 MHS units in 104 municipalities.

The municipalities were divided into three groups based on the 2015 classification by Statistics Finland for describing the degree of urbanization of their residence: urban, semi-urban and rural. Urban municipalities included those in which at least 90% of the population lived in urban settlements or where the population of the largest urban settlement was at least 15,000. In semi-urban municipalities, at least 60% but less than 90% of the population lived in urban settlements, and the population of the largest urban settlement was between 4000 and 15,000 inhabitants. In rural municipalities, less than 60% lived in urban settlements, and the population of the largest settlement was less than 15,000 inhabitants, or between 60 and 90% of the population lived in urban settlements, and the largest settlement was less than 4000 inhabitants.

Incidence rate ratios (IRR) and 95% confidence intervals (95% CI) were calculated for the models. Because of varying multicollinearity between the different MHS types and municipality groups, the analysis for each MHS factor was modeled separately. The correlation between the municipality groups MHS factors and demographic characteristics are shown in Online Resource 3. In all statistical analyses, P values < 0.05 were considered statistically significant. GSDI values were calculated for the MHS factors with R version 4.0.1 [42], RStudio [43] and the R-package diverse [44]. The statistical analyses were performed with SPSS version 28.0.

Characteristics of the study’s catchment area and municipality groups on the macro-level are reported in Table 1. Urban municipalities comprised 29% of all the municipalities, but 79% of all DP recipients and 81% of all the catchment area population resided in them. Urban municipalities also had a lower ratio of mood disorder DP, mental health index and dependency ratio in the population, as well as a higher rate of higher education qualifications and population density. The characteristics of the semi-urban and rural municipalities were primarily similar, although semi-urban municipalities had lower unemployment rates. Rural municipalities also had the lowest rates of population density and higher education qualifications, but also the lowest rate of those not in education or training at age 17–24.

Means and standard deviations are reported for the MHS types and their factors in Table 2 and Online Resource 4. In all MHS types the municipality groups were statistically significant concerning service richness and diversity, but did not differ regarding FTE resources per 1000 inhabitants. The mean number of MHS FTE in all municipalities was 3.13 per 1000 inhabitants (SD 1.28). The mean for service richness of all MHS was 15.67 distinct ESMS-R classes offering services to a single municipality’s residents (SD 6.09). The mean value of GSDI for all MHS diversity was 0.88 (SD 0.04) between all municipalities. Concerning the municipality groups and MHS types, service richness and diversity were highest in urban municipalities, while being lower in semi-urban and typically lowest in rural municipalities (Table 2). Only in local services with gatekeeping did rural municipalities have a higher GSDI (0.69; SD 0.11) than semi-urban (0.66; SD 0.18).

Noticeable differences between MHS factors and mood disorder DP associations were observed (Table 3). The relationship between MHS factors and DP appears to be associated with the degree of urbanicity and the context of the municipalities. When all municipalities were studied together, a higher service richness and diversity in all MHS (richness IRR 0.995; 95% CI 0.991–0.998, and GSDI IRR 0.396; 95% CI 0.185–0.850), outpatient care (richness IRR 0.978; 95% CI 0.996–0.990, and GSDI IRR 0.687; 95% CI 0.511–0.924) and local services with gatekeeping (richness IRR 0.980; 95% CI 0.970–0.990, and GSDI IRR 0.641; 95% CI 0.500–0.821) were associated with lower DP risk.

In urban municipalities service richness was associated with lower DP in all five studied MHS types (all MHS IRR 0.993; 95% CI 0.988–0.998, outpatient care IRR 0.976; 95% CI 0.962–0.990, local services without gatekeeping IRR 0.985; 95% CI 0.970–1.000, with gatekeeping IRR 0.982; 95% CI 0.971–0.994, and centralized services IRR 0.986; 95% CI 0.974–0.998), as well as with service diversity in local services without gatekeeping (IRR 0.428; 95% CI 0.258–0.711).

Uniquely in semi-urban municipalities, a higher FTE per 1000 inhabitants indicated a lower DP risk in all MHS (IRR 0.941; 95% CI 0.897–0.988), outpatient care (IRR 0.818; 95% CI 0.726–0.922), local services with gatekeeping (IRR 0.881; 95% CI 0.792–0.980) and centralized services (IRR 0.925; 95% CI 0.859–0.996), but not in local services without gatekeeping. Furthermore, in outpatient care we found a lower risk of DP associated with higher service diversity (IRR 0.644; 95% CI 0.442–0.940), and in local services with gatekeeping a lower DP risk with higher service richness (IRR 0.943; 95% CI 0.913–0.975) and diversity (IRR 0.544; 95% CI 0.398–0.743). Thus, all studied MHS factors showed an association with lower DP risk in local services with gatekeeping, but not in local services without gatekeeping. Interestingly, we found no associations between rural municipalities' DP risk and MHS factors.

The strengths of this study include the use of comprehensive national-level data registers. Finnish population registers have high coverage and quality, allowing detailed and extensive epidemiologic research for MHS associations with mood disorder DP in this study [45, 46]. To our knowledge, there has not previously been a comprehensive study of the relationship between MHS types and factors, and mood disorder DP in different municipality contexts. The ESMS-R mapping tool was used for clear hierarchical taxonomy-based coding of MHS. This study also included an examination of the MHS context of service provision. Both are essential in researching complex MHS systems [9, 10, 16].

Our study setting includes some important limitations. One major limitation in this study was that because of the varying multicollinearity between the different MHS types and municipality groups, not all the MHS factors could be entered and adjusted in the same model. The correlation between the municipality groups MHS factors and demographic characteristics are provided in Online Resource 3. Secondly, the MHS units could be of different sizes, which does not affect the FTE but could affect GSDIs, which were calculated with the available units in the municipality and therefore reflected the number of components in the complex MHS system rather than the components’ size. Thirdly, some MHS units provided services to several municipalities, which could involve regional dynamics that were not comprehensively considered in this study. In these MHSs, the FTE resources were allocated to municipalities on the basis of their relative share of inhabitants. This factor is based on the assumption that all the municipality’s inhabitants used the MHS available to them in equal amounts. However, this might not be the case, although this was the best available estimate in this study. Fourthly, the MHS data does not include information about the co-operation of the services or pathways of care between them, or on whether the psychosocial treatment provided was grounded in evidence-based psychosocial treatment models and a specific philosophy/culture of psychosocial treatment provision. In future research, the treatment contents and cultures should be integrated to ESMS-R classification research, which could yield a more complex but truthful picture of MHS ecosystems and functioning concerning mood disorder treatment and DP prevention.

It is important to note that the ESMS-R classification tool is not all-encompassing, and there may have been subtle features of the classified MHS units that are not included in the analysis. The ESMS-R data in this study only includes public services, and it excludes Finnish occupational health care, private services or rehabilitative psychotherapy imbursed by the Social Insurance Institution of Finland. However, previous studies have indicated a lower mental disorder DP rate in occupational health care users compared to population statistics [47], and a higher rate of sick leaves for mental disorders in public service users compared to occupational health or private service users [48]. These findings indicate that public MHS have a crucial populational role in treating and pre-empting mental disorders and disability for most of the population.