Factors influencing utilization
Psychotherapeutic services for depression – is it all a question of region?

Depression is one of the most frequently diagnosed mental illnesses both among the general population in Germany and among those insured by AOK Nordost. Previous studies estimate the prevalence of depression in Germany at 8.1% to 10.2%, depending on the respective study design and the underlying data source (Busch et al. 2013; Erhart and von Stillfried 2012). A continuous increase in the prevalence of depression has been observed for years (Thom et al. 2017). Even though treatment rates are constantly increasing, it is estimated that between 35% and 80% of all patients with mental disorders do not receive treatment (Larisch et al. 2013). It can be assumed that this also applies to the treatment of depression to a similar extent. In addition to individual factors as well as social and natural factors, the treatment rate also depends on the availability of service providers (Rommel et al. 2017; Helbich et al. 2018). Accordingly, the current needs planning of psychotherapists is also the subject of controversial discussion (IGES 2016; Peikert et al. 2011).

>> The demand planning for psychotherapists in Germany is still based on ratios at the level of the districts, which were determined in 1999 by the ratio of all residents of the entire Federal Republic - i.e. West and East Germany combined - to all licensed psychotherapists and are based on the degree of urbanization of the districts judge. In comparison, the ratios for general practitioner demand planning are based on the ratio of residents to practicing general practitioners in 1990 only for West Germany. Since in 1999 there were still many psychotherapist offices in the approval process, especially in the new federal states, the IGES Institute considers this ratio to be underestimated from today's perspective (IGES 2016).
The only potential influencing factor on the frequency of mental illnesses is the age distribution of those under 65 years of age using a demographic factor (IGES 2016). The waiting times between the request and the start of guideline psychotherapy show, among other things, that the current requirements planning does not adequately take regional care needs into account. There are considerable differences in the area of responsibility of AOK Nordost - i.e. the federal states of Berlin, Brandenburg and Mecklenburg-Western Pomerania. While in Berlin patients only have to wait an average of 13.4 weeks for therapy to begin, in Brandenburg it is 23.4 weeks and in Mecklenburg-Western Pomerania 20.2 weeks (Chamber of Psychotherapists 2018).
Accordingly, a readjustment of the psychotherapeutic ratios is also required (IGES 2016). As part of a simulation study, it was discussed how the ratios of psychotherapeutic demand planning could be recalculated based on the results of the study on health in Germany currently (GEDA) by the Robert Koch Institute. As a result, in addition to determining a uniform ratio, the socio-economic structure of the districts should also be taken into account (IGES 2016). Previous study results from primary data indicate a connection between the use of psychotherapeutic services and the regional availability of service providers. However, analyzes based on primary data have the disadvantage that they only cover a small part of the population and, due to the small sample, only have a rough regional resolution (Rommel et al. 2017; Jacobi et al. 2016; Thom et al. 2017).
Here, data from statutory health insurance companies (GKV) has the advantage that it covers a large part of the population, is available on a small scale and across the board, and that both socio-demographic characteristics and information on previous illnesses are available for the insured person. However, small-scale studies using geographic information systems (GIS) and spatial-statistical methods on the connection between the use of psychotherapeutic services and the regional availability of service providers based on billing data are still missing for Germany. With this work we want to close this gap. The aim of our study is to shed more light on regional differences in diagnosis prevalence and the use of psychotherapeutic services for depression in northeast Germany. In addition to individual factors, the connection to the availability of psychotherapists is also examined in more detail.
Data basis and methods
For this study, anonymized billing data from AOK Nordost at the individual level for 2016 was used. AOK Nordost insures around 1.6 million insured people in the federal states of Berlin, Brandenburg and Mecklenburg-Western Pomerania. All insured persons over the age of 18 were included in the study (n = 1.45 million). Depression was defined as the presence of a confirmed diagnosis based on the ICD-10 codes F32, F33 and F34.1. To calculate the proportion of insured persons who used psychotherapeutic services for a diagnosed depression, those insured persons were used who received psychotherapy for depression in the calendar year 2016 in accordance with the guidelines of the Federal Joint Committee on the implementation of psychotherapy (Federal Committee 2009 ).
The selection of explanatory variables is taken from the insured person master data, the data from the 2011 census at the municipality and district level and calculated indicators on the availability of service providers. Age, gender and citizenship were taken from the insured person master data. To measure socioeconomic status, the insured types of unemployed and long-term unemployed were combined. Chronic somatic diseases are also considered risk factors for depression, although the direction of the relationship has not yet been clearly identified (Pieper et al. 2008). Therefore, for this study, the number of diagnoses for none or several of the diseases type 2 diabetes mellitus (Pieper et al. 2008), coronary heart disease (Pieper et al. 2008) and chronic obstructive pulmonary disease (COPD) was included (Schneider et al. 2010 ). The proportion of one-person households was considered as an area-related variable as an indicator of a lack of social support (Wittchen et al. 2010).

Gravity model for measuring the density of general practitioners and psychotherapists
So far, the planning of service providers has taken place at the level of administrative units. However, this approach does not take into account a) the accessibility of the service providers, b) co-care from nearby service providers and c) the willingness to travel to a doctor decreasing with distance. For these reasons, so-called gravity models are better suited to measuring resident-physician ratios. To put it simply, a gravity model measures the population-doctor ratios in certain catchment areas (according to the Federal Joint Committee (GBA), for example 15 minutes for general practitioners and 30 minutes for psychotherapists (Sundmacher et al. 2018)), taking into account a) the accessibility of the service providers, b) Co-care from nearby service providers and c) the willingness to travel to a doctor decreases with distance. This means that the resident-physician ratios measured using this approach are also better suited to determining connections between utilization and availability of service providers. In order to look at the connection between diagnosis, utilization and the availability of service providers, the ratio of general practitioners per 100,000 inhabitants in distance-weighted catchment areas of 15 minutes and the ratio of psychotherapists to residents in distance-weighted catchment areas of 30 minutes were calculated. The methodology for calculating the gravity model is taken from the G-BA's report on the further development of requirements planning (Sundmacher et al. 2018) and was implemented using ESRI ArcGIS 10.6. To calculate the psychotherapist density, all doctor groups relevant to the psychotherapist requirements planning (child and adolescent psychotherapist, psychological psychotherapist and psychotherapeutic doctor) were included. The other groups of doctors who can also provide psychotherapeutic services (child and adolescent psychiatry, neurology, psychiatry and psychotherapy, psychosomatic medicine and psychotherapy) can belong to neurologists or psychotherapists, depending on the main share of the service provided and licensing status.

Cartographic representation of prevalence and treatment rates
The population of Germany was used to represent the age- and gender-standardized prevalence of depression. Since the number of diagnosed depressions varies greatly within the communities and districts, a Bayesian smoothing procedure according to Besag-York-Mollie was also used (Besag, York, and Mollie). The BYM model always weights the rates within a community to the average of neighboring communities and the global average. When presenting the treatment rate, age standardization was omitted because the actual treatment proportion was of interest.

Statistical analysis
Our analysis is based on an individual-level Bayesian logistic regression model. The advantage of Bayesian models is the flexibility to assume different effects for certain variables. For example, the probability of developing depression does not increase linearly with age (Erhart and von Stillfried 2012). For this reason, the age of the insured was integrated into the model with a non-linear effect (random walk prior). This allows the modeling and visualization of non-linear relationships (Helbich et al. 2017). Due to the different waiting times between the request and the start of guideline therapy between the federal states, it can be assumed that there is a regionally varying connection between the use and availability of psychotherapists. The model was therefore calculated twice: once as a global model, in which the psychotherapist density only received a global coefficient, and once as a local model, in which the psychotherapist density received a locally varying effect. Such a local effect was taken into account in the model using a Besag prior (Blangiardo and Cameletti 2015). All other variables were assumed to have a) regional dependencies and b) random effects based on small regional units.
The calculation of the smoothing procedure and the Bayesian logistic regression model was carried out using the R package INLA (Rue, Martino, and Chopin 2009). The results of the regional effects were presented with ESRI ArcGIS 10.6 and the results of the demographic effects with the R package ggplot2 (Wickham 2016).
Results
Regional distribution of psychotherapist density
Overall, there is a very clear difference from Berlin to the surrounding rural regions (Fig. 1): While in West Berlin there are over 51 psychotherapists per 100,000 inhabitants, in some regions of Brandenburg there are fewer than 15 psychotherapists per 100,000 inhabitants. Overall, the density of psychotherapists in the rural regions of Mecklenburg-Western Pomerania is higher than in the rural regions of Brandenburg. The gravity model illustrates the strong differences within the districts; This is particularly clear in the districts bordering Berlin. This also makes the weakness of the planning at the district level clearly visible: The majority of psychotherapists in the districts bordering Berlin tend to settle in the Berlin area - where there is already a high density of psychotherapists - and not in the rural communities, where accordingly According to previous findings, the need for care is not as low as the differences in the ratios between rural and urban planning regions suggest (IGES 2016).

Regional distribution of depression prevalence and treatment rates
The age- and gender-standardized prevalence of depression was 16.1% in 2016 (Fig. 2a). By far the highest prevalence is in West Berlin. In rural regions, the prevalence of depression is generally rather homogeneously distributed - with a few exceptions. The differences between the rural communities are less pronounced than between the living environment-oriented areas in Berlin. The use of psychotherapeutic services when diagnosed with depression was on average 48.8% and therefore corresponds to the values from previous studies (Wittchen et al. 2010). The strong contrast between the federal states is striking (Fig. 2b): While in Mecklenburg-Western Pomerania the treatment rates are predominantly above average, in Brandenburg they are predominantly below the AOK Nordost average. The small-scale analysis at the postal code level shows that the highest treatment rates can be found in the Mecklenburg Lake District. The lowest treatment rates are in the Uckermark district in Brandenburg.

Factors influencing the diagnosis of depression
Essentially, our analyzes confirm previous results on risk factors for depression: As expected, the age of the insured is not linearly associated with depression (Fig. 3a). In our study, too, the risk of depression increases continuously until the age of 55, remains at the highest level until 65 and then decreases again (Erhart and von Stillfried 2012; Busch et al. 2013). Women have a 136% higher risk of depression than men, unemployed insured people have a 35% higher risk than non-unemployed insured people, and as the number of chronic illnesses increases, the likelihood of depression also increases: Insured people with a somatic illness have a 65% higher risk, insured people with two somatic diseases a 112% higher risk and insured people with three somatic diseases a 182% higher risk than insured people without somatic diseases (Table 1). However, with regard to citizenship, our results differ from previous findings (Baschin et al. 2012). Insured people with foreign citizenship have a 4% lower risk of being diagnosed with depression than insured people with German citizenship. The proportion of one-person households - as an indicator of a lack of social support - is significantly associated with diagnosis prevalence. With one percent more one-person households in the insured person's place of residence, the probability of depression increases by 29%. Our results also differ from previous findings when it comes to the connection between diagnosed depression and the density of general practitioners. Overall, the density of general practitioners is not significantly associated with the probability of diagnosis. However, it must be clearly taken into account here that there are strong differences between Berlin and the regional states. For Berlin alone there is a significant, slightly positive connection to the density of general practitioners, but not for the regional states. The influence of population density is not significant; Based on our results, it cannot be concluded that depression is higher in urban regions – with the exception of the “outlier” Berlin – than in rural regions.

Factors influencing the use of psychotherapeutic services for depression
The factors influencing utilization essentially reflect previous findings from the GEDA study (Rommel et al. 2017): Among insured persons diagnosed with depression, utilization decreases continuously with age (Fig. 3b). Women are 55% more likely to use psychotherapeutic services than men. Unemployed insured people have a 10% higher probability than unemployed insured people. Insured people with foreign citizenship are less likely to have a 9% than insured people with German citizenship. As the number of chronic illnesses increases, utilization also increases: insured persons with one somatic illness have a 5% higher probability, insured persons with two somatic illnesses have a 9% higher probability, and insured persons with three somatic illnesses have a 16% higher probability of using psychotherapeutic services as insured persons diagnosed with depression without the presence of one of the three somatic illnesses. However, the influence of the proportion of one-person households in the place of residence differs here: While one-person households are positively and significantly associated with the diagnosis, the association with utilization is negatively associated and no longer significant.
It can be said that our results reflect the findings of previous publications on the risk factors of depression and the resulting use of psychotherapeutic services (Wittchen et al. 2010; Rommel et al. 2017): Essentially, age, gender, and the presence of somatic illnesses and socioeconomic status are the most important determinants for the diagnosis of depression and for the use of psychotherapeutic services. The main difference is related to age: While the probability of diagnosis increases continuously up to the age of 50 and then remains at an elevated level even in those over 65, the use of psychotherapy decreases continuously with age.

Influence of the density of psychotherapists on the use of psychotherapeutic services for depression
According to previous findings, it would have been expected that utilization would correlate positively with the density of psychotherapists, especially in cities (Rommel et al. 2017). However, based on our results, this can only be confirmed for Potsdam, Schwerin, Rostock, Cottbus and some postcode areas in Berlin - especially for the postcodes in West Berlin (Fig. 4). In these cities, the probability of utilization increases by up to 2.5% with one more psychotherapist per 100,000 inhabitants. In Frankfurt (Oder) and Brandenburg an der Havel, however, the probability of utilization decreases by up to 2.5% with one more psychotherapist per 100,000 inhabitants.
The differences are even more pronounced in the
Area states: In many postal codes in Mecklenburg-Western Pomerania, the probability of use increases as the density of psychotherapists increases. In the Mecklenburgische Seenplatte district, the probability of utilization increases in some areas by up to 5% to 7.3% with one more psychotherapist per 100,000 inhabitants.
Although there are also some areas in Brandenburg in which the density of psychotherapists correlates positively with utilization, there is also a relatively area-wide
covering negative connection. Particularly in the districts of Uckermark, Prignitz, Havelland and Oder-Spree, there are some strongly negative connections. In some regions, the probability of utilization decreases by 2.5% to 5% with one more psychotherapist per 100,000 inhabitants. In some regions even around 5% to 8.8%. A negative correlation between the probability of use and the density of psychotherapists arises when, for example, supply capacities are exhausted, waiting times seem too long for the patient or the treating psychotherapists only work part-time and can therefore accept fewer patients. Previous studies show that the scope of care differs significantly between regions: In regions with high levels of care, only about half as many cases are treated per psychotherapist as in regions with a lower density of psychotherapists (Jacobi et al. 2016) - which means corresponding differences in waiting times between Clearly show the request and start of guideline psychotherapy (Chamber of Psychotherapists 2018). Accordingly, it is not surprising that negative correlations can be found, particularly in sparsely populated regions with a lower number of psychotherapists. The negative and relatively strong correlation with the density of psychotherapists suggests that psychotherapeutic care is inadequate in large parts of Brandenburg.
However, it must be expressly emphasized here that crucial information about the psychotherapists is not available: In the billing data used here, for example, there is no information about the care order, the number of consultation hours and the work volume of the respective psychotherapist. Each psychotherapist entered the analysis with an entire care assignment. Therefore, our result should be repeated with complete data on the respective service providers. Of course, other groups of doctors can also provide psychotherapeutic services that are not currently taken into account in the psychotherapists' needs planning (e.g. general practitioners with quality-assured additional qualifications in depth psychology-based psychotherapy or psychosomatic physicians). It must also be taken into account in this analysis that depression can also be treated with medication by other specialist groups that are not taken into account in the psychotherapists' needs planning.
discussion
What does the result mean for the needs planning of psychotherapists?
Our results suggest that there needs to be continued discussion about changing the psychotherapeutic ratios. Based on the assumption that mental illnesses are evenly distributed across cities and rural areas, the IGES Institute, on behalf of the Federal Chamber of Psychotherapists, proposes a uniform ratio that should be regionally adjusted using a needs index (IGES 2016). Due to the lack of correlation between depression prevalence and population density, a uniform ratio - with the exception of Berlin - seems quite plausible based on our results. However, our study cannot make any statement about how high this uniform ratio should be. An analysis of mental illnesses based on all those insured by statutory health insurance would be necessary here. When considering whether the ratios should be adjusted based on socioeconomic factors, it must be noted that although the association between unemployment and prevalence is high, the association with the use of psychotherapeutic services is rather low. The extent to which an adjustment of the ratios by socio-economic factors actually leads to higher utilization in rural regions cannot be clarified in this work. Further discussions with all those involved in demand planning would be necessary to discuss this.
However, planning based on ratios - regardless of whether they are adjusted or not - is methodologically difficult: ratios based on administrative units do not take into account accessibility or co-supply effects from nearby service providers. When planning at the district level, it is assumed that residents within a certain district will also visit the respective service provider in this district. However, this approach does not take commuter movements from the periphery into the big cities into account. It would make more sense here to measure the ratios using a gravity model, as is also used in the report on the further development of the G-BA's requirements planning and in this study, in order to also take into account the accessibility and co-supply effects of nearby psychotherapist locations (Sundmacher et al. 2018 ).
The application of the gravity model has made it clear that there are strong differences within the districts bordering Berlin - which only represent one planning area. While the density of psychotherapists is very high in the Berlin area, it is very low in the rural communities in the same district and cannot be justified based on need. This problem must be taken into account and cannot be solved by adjusting the ratios alone. It would make more sense to continue to measure the supply situation in the future using the gravity model developed by the G-BA. However, with the data available to us, our results can only provide an indication of differences in care; for a closer look, more detailed information on the care mandate and the scope of care provided by the individual psychotherapists would be necessary.
conclusion
Our results make it clear that the current planning of psychotherapists' needs cannot be justified on the basis of epidemiological results. On the contrary: the current ratios tend to lead to an undersupply in rural regions. A uniform ratio would therefore be entirely plausible based on our results. Whether adjustment based on socioeconomic factors would also lead to increased utilization remains to be discussed. However, our results also make it clear that simply changing the ratios would not necessarily lead to an improvement in care. It would make more sense here to measure the supply based on the gravity model developed by the G-BA in order to prevent possible oversupply in the Berlin area and thus also an undersupply in the rural communities in the districts bordering Berlin. <<