Keywords Mental health - healthcare utilization - children and adolescents - deprivation
Introduction
Social determinants of health and well-being tend to cluster in specific
neighbourhoods [1 ]. The socioeconomic
status is known to be a strong social determinant of health at the individual level
[2 ]. Individual health is strongly
influenced not only by personal socioeconomic status, but also by the socioeconomic
status and living conditions of the environment [3 ]. Children and adolescents are strongly linked to their families
socioeconomic position and living environment [4 ]. The neighbourhood is of particular importance to children and
adolescents compared to adults, as they spend a significant amount of time in their
local environment due to restrictions on free exploration and access to different
environments [5 ]
[6 ]. Neighbourhood characteristics have been
emphasised as important for the development of children and adolescents in addition
to individual and family characteristics [7 ]. Among the various neighbourhood effects, neighbourhood deprivation
has been studied and recognised as a social determinant of young people's
health [6 ]. Deprivation at the individual
level is defined as the inability to participate in common activities and to access
the resources and opportunities that are considered customary or essential for a
reasonable standard of living within a particular society [8 ]. Deprivation thus refers to a relative
phenomenon of poverty and social exclusion [8 ]. Deprivation tends to cluster spatially, with groups with similar
socio-economic circumstances often living in close proximity to each other [9 ]. Research has identified neighbourhoods,
the social environment and area-deprivation as potential risk factors for mental
health among children and adolescents [6 ]
[10 ]. Moreover, an unequal
effect of the COVID-19 Pandemic on mental health between different population
groups, moderated by the distance to urban parks was observed [11 ]. During the COVID-19 pandemic, an
increase in psychosomatic diseases in children and adolescents compared to the
previous period was found [12 ] with an
disproportionately impact on children with disadvantaged social backgrounds [13 ].
Access and healthcare utilization in Germany
Access and healthcare utilization in Germany
Access to mental health care in Germany can be challenging despite acute illness and
can be accompanied by long waiting times [14 ]
[15 ]
[16 ]. One reason is commonly seen in the
mechanisms for care demand planning. In order to be able to provide and bill
services within the statutory health insurance (SHI), primary care physicians and
psychotherapists need a social security license (Kassensitz) . The number of
social security licenses are determined for a given region, based on demand planning
that was last adjusted in 2019 [17 ]. The
planning regions are divided into four levels of care. The planning regions for
psychologists are administrative districts or cities, and child and adolescent
psychiatrists are planned in larger so-called spatial planning regions [17 ]. However, a point of criticism of the
demand planning is the uneven distribution within planning areas, some of which
cover large areas [18 ]. This has been
studied with focus on general practitioners and paediatricians in the city of Essen,
showing a maldistribution in favour of more affluent districts and a resulting
social inequality in access to care [19 ].
Moreover, the resulting care situation was studied in particular in comparison with
regional differences, showing an unequal distribution especially in rural areas
[20 ]
[21 ].
This type of supply-side planning has particular implications for urban areas. The
city of Cologne, for example, is the fourth largest city in Germany with a
population of over one million, which raises issues of inequality and potentially
long travel times. For children and adolescents in particular, long journeys from
their place of residence to where healthcare is provided can be a barrier to
healthcare access. As an urban centre of a densely populated metropolitan region,
Cologne has strong links with neighbouring rural areas and towns. This creates a
large catchment area for specialised healthcare services. In this capacity, it is
an
example of the urban centre of a prototypical European metropolitan area within a
densely populated zone. Not least for this reason, it is important to examine the
role of the neighbourhood and the areal distribution of providers of mental health
care services for children and adolescents.
Focusing on children and adolescents aged 0 to 19 years living in the city of
Cologne, this article addresses following questions:
(1) How are prevalence of mental and behavioural disorders and mental health services
distributed across the city of Cologne?
(2) Whether and how are area deprivation and the availability of mental health
service providers providing services to patients covered by the SHI system
associated with the prevalence of mental and behavioural disorders and utilization
of mental health services?
(3) How do the associations between area deprivation, availability of mental health
service providers, prevalence of mental and behavioural disorders and health care
utilisation vary across space?
Methods
Data
Claims data
The research questions will be investigated using claims data from children
and adolescents residing in the city of Cologne and insured by four
statutory health insurances. These data are made available for research by
the CoRe-Net project (Cologne Research and Development Network) [cf. 22]. In
the CoRe-Net project, the insurance data of four statutory health insurances
cover around 50% of the population of the city Cologne [22 ]. In 2021, a total of 193,844
children and young people aged 0 to 19 lived in the city of Cologne. From
this database, the present research used 2021 data from children and
adolescents aged 0 to 19 years who were continuously insured with one of the
statutory health insurers in the observation year. Sex was included as a
binary category (male, female) due to data availability. Analyses focus on
prevalence of ICD-10 diagnoses (F00-F99) and utilization of mental health
services (EBM). Definitions of ICD-10 diagnosis and mental health services
are reported in the online-appendix (Table 3–4). In accordance with common
practice for claims data [23 ],
diagnoses were validated before analysis. Diagnoses coded by registered
physicians were included if they were documented in at least two of four
quarters (M2Q criterion (minimum 2 quarters)). Discharge/primary diagnosis
from the hospital were included. For the analysis of utilization of mental
health services, the study population includes all children and adolescents
with a prevalent diagnosis of mental and behavioural disorders.
The 45 ZIP-code areas in the analyses of prevalence and utilization of mental
health services reflect the places of residence of the children and
adolescents. All results are sex- and age standardized and adjusted to case
numbers per 1,000 children and adolescents based on the respective ZIP-code
areas. The adjusted rate provides a measure of the case count's
magnitude, irrespective of the number of children and adolescents in the
respective areas. This approach offers the advantage of facilitating
comparisons between ZIP-code areas of varying sizes and population
densities.
Further data sources
As the aim of the study was to evaluate, whether the diagnosis for mental and
behavioural disorders and the utilization of mental health services are
associated with area deprivation, the German index of socioeconomic
deprivation (GISD) was calculated [24 ]. The index includes the dimensions education, occupation and
income. Due to data availability, a modification of the GISD was necessary.
[Fig. 1 ] displays the
included indicators for each dimension. Data on school leavers without a
degree, employed persons with a university degree, unemployment and
employment rate are provided by the City of Cologne. The debtor rate is
based on the publication of the Debtor Atlas for the Cologne/Bonn
Metropolitan Region [25 ]. The data
is available at district level and in percentages. The 45 ZIP-code areas
were defined by the post office, with the 86 districts being determined by
the city itself. Both classifications are subject to other regulations and
only overlap to a limited extent. Due to the availability of data area,
shares of the city districts in the ZIP-code areas were calculated using
QGIS 3.30. In the next step, the indicators were calculated according to
their area share (e. g., 33% district X, 77% district Y) for the ZIP-code
areas. Based on this, the dimensions and finally the deprivation index with
a possible range from 0 to 100 was calculated for each ZIP-code area for the
year 2021.
Fig. 1 Study model for research question 2 and 3.
In order to account for availability and access to care, the ratio of child
and youth mental health service providers per 1.000 children and adolescents
per ZIP-code area is used. The numbers are based on providers reported on
the website of the Association of Statutory Health Insurance Physicians of
North Rhine (www.kvno.de ). The number
of providers by ZIP-code area was extracted on March 22th 2023.
[Fig. 1 ] shows the study models
for research questions 2 and 3.
Statistical analysis
In the first step, the distribution of the prevalence of diagnoses for mental
and behavioural disorders, the utilisation of psychotherapeutic services,
deprivation index and the distribution of mental health service providers
with social security license were mapped. For visualization, QGIS 3.30 was
used.
Analyses have been carried out using R 4.3.1/R studio. Before creating a
geographically weighted regression model, all significant explanatory
variables were identified through exploratory regression. Therefore, a
linear regression model (OLS) with robust standard error has been used. In
order to control for normal distribution as requirement of OLS, the
distribution of the untransformed and the log-transformed dependent
variables were evaluated. The untransformed dependent variable prevalence
was closer to the normal distribution, for healthcare utilization the log
transformation was closer to the normal distribution and was therefore
chosen as the final dependent variables in the analyses. The explanatory
variables identified for the final model are based on the following
criteria: (1) the variables are statistically significant (p<0.005) and
(2) free from multicollinearity. The correlation between the study variables
was tested as a sensitivity analysis (Table 5, online-appendix).
Correlations between the deprivation index and the distribution of mental
health service providers working within the SHI system (-0.51) as well as
correlation between deprivation index and utilization of mental health
services (-0.67) were found. The control variable share of children
aged<11 correlates strongly with the distribution of mental health
service providers (0.86). In order to control for lower diagnosis rates and
utilization of mental health services due to a younger average age between
ZIP-code areas, the proportion of children aged under 11 is included in the
analyses as a control variable.
One goal was to estimate the strength of association between the identified
explanatory variables and prevalence for mental and behavioural disorders/
utilization of mental health services. The socio-demographic composition and
deprivation of the population varies widely across the city of Cologne.
Therefore, it is hypothesized that a geographically weighted regression
(GWR) model provides a better explanation than a global OLS model [26 ]
[27 ]. GWR offers the capability to
quantify changes in relationships between predictors and outcome variables
across different spatial locations, all within a unified modelling framework
[28 ]
[29 ]. GWR has been carried out using
the R package spgwr. The data analysis followed established standards
of secondary data analysis [30 ].
Results
Cartographic visualization
The data was analysed broken down by ZIP-code and presented cartographically in
order to detect areas with elevated prevalence rates for mental and behavioural
disorders and utilization of mental health services (research question 1). [Fig. 2 ] shows the prevalence of
validated diagnosis for mental and behavioural disorders per 1,000 children and
adolescents. The utilisation of mental health services per 1,000 children and
adolescents in the Cologne ZIP-code areas are displayed in [Fig. 3 ].
Fig. 2 Geographical representation of the prevalence of diagnoses
for mental and behavioural disorders (F00-F99, M2Q) by children and
adolescents per 1,000 children and adolescents. Underlying data: Health
insurance data (claims data) from the CoRe-Dat database.
Fig. 3 Geographical representation of the utilisation of
psychotherapeutic services (EBM) by children and adolescents per 1,000
children and adolescents with prevalent diagnoses for mental and
behavioural disorders. Underlying data: Health insurance data (claims
data) from the CoRe-Dat database.
The overall prevalence of mental and behavioural disorders in the year 2021 was
195.22 per 1,000 children and adolescents aged 0 to 19 years. The spatial
distribution of the prevalence varies between the 45 ZIP-code areas. The
prevalence of mental and behavioural disorders ranges from 121.57 to 261.04 per
1,000 children and adolescents. There is a trend of higher prevalence in the
northern and southern areas of the city as well as the areas East of the river
Rhine, while the inner city and the west exhibits comparably lower
prevalence.
[Fig. 3 ] illustrates mental health
service utilization among children and adolescents with a prevalent diagnosis
for mental and behavioural disorders. The utilization rate ranges from 370.37 to
790.32 per 1,000 children and adolescents. The average utilization rate was
549.06 per 1,000 children and adolescents with prevalent diagnosis for mental
and behavioural disorders. Local clustering of service utilization can be
observed in the north western part of the city (low rates of utilization of
mental health services). Looking further at the geographical patterns, the lower
health service utilization rates on the eastern side of the Rhine contrast with
the high health service utilization rates in the southern parts of the city
centre on the western side of the Rhine.
The second aim of the study is to uncover potential links between area
deprivation, the availability of mental health service providers offering
services to patients covered by SHI and the prevalence for mental and
behavioural disorders as well as utilization of mental health services (research
question 2). For this purpose, the deprivation index ([Fig. 4 ]) and the distribution of
mental health service providers in 2023 ([Fig. 5 ]) were analysed cartographically.
Fig. 4 Geographical representation of the deprivation index. Based
on own calculations. Data used as a basis: Data from the City of
Cologne, Debtor Atlas for the Cologne/Bonn Metropolitan Region.
Fig. 5 Geographical distribution mental health service providers
for child and youth with social secext-linkty license per 1,000 children
and adolescents. Based on own calculations. Data used as basis:
www.kvno.de
(22.05.2023).
The deprivation index indicates higher levels of deprivation on the eastern side
of the Rhine and in the northern areas compared to the rest of the city ([Fig. 4 ]). Compared to this,
deprivation levels in the city centre on the western side of the river are
lower. Overall, deprivation index ranges from 23.42 to 45.00 with a mean of
31.85.
Mental health service provider for children and adolescents with social security
license are clustered in the city centre ([Fig. 5 ]). In 2023, no mental health service providers were registered
in the northern areas, southern areas and eastern areas of the city.
OLS regression
We identified the deprivation index as a significant predictor of the prevalence
of mental and behavioural disorders (research question 3, [Table 1 ], online-Appendix Table 6).
The model including the share of children ages under 11 years explains 11,17% of
the prevalence ([Table 1 ]). The
residuals of the model were significantly clustered (Moran’s I: SD=0.40,
p<0.01). However, by comparing the AICc (Akaike Information Criterion) of the
OLS and GWR model, the OLS outperforms the GWR (∆AICc=13.58) and is therefore to
be preferred. The OLS model shows a significant increase in the prevalence with
an increase of the deprivation index (ß=1.66, p<0.05, SE=0.80) ([Table 1 ]). Analyses by sex show no
significant effect of the deprivation index on prevalence (online-Appendix,
Table 7–8).
Table 1 OLS Regression for the prediction of diagnosis of
mental and behavioural disorder.
Model
Coefficient
Std. error (robust)
Intercept
139.37***
28.32
Deprivation Index
1.66*
0.78
Percentage of children aged<11 years
− 1478.13
7479.53
Adjusted R²
0.11
Global Moran`s I of residuals
I=0.40 (p<0.001)
Significance levels: *≤ 0.05; **≤ 0.01: ***≤ 0.001.
Explaining the utilization of mental health services among prevalent cases the
deprivation index, the share of children under 11 years old as well as
distribution of mental health service providers were identified as a significant
predictor with an explained variance of 61.32% ([Table 2 ]). A global OLS model can be
considered as appropriate for modelling the utilization of mental health
services within the city of Cologne based on the Global Moran`s I of residuals
(SD=-0.02, p>0.05, (∆AICc=1.75). The final model indicates an increase in the
utilization of mental health services with an decrease in the deprivation index
(ß=-0.02, p<0.001, SE=0.00) as well as an increase of healthcare utilization
with increasing ratio of mental health care providers (ß=0.15, p<0.05,
SE=0.07). Analyses by sex support the finding of a significant effect of the
deprivation index. The distribution of mental health service providers is
significant only for male (online-Appendix, Table 9–10).
Table 2 OLS Regression for the prediction of healthcare
utilization.
Model
Coefficient
Std. error (robust)
Intercept
7.04***
0.10
Deprivation Index
− 0.02***
0.00
Ratio of mental health care providers
0.15*
0.07
Percentage of children aged<11 years
− 420.87***
80.20
Adjusted R²
0.61
Global Moran`s I of residuals
I=0.06 (p>0.05)
Significance levels: *≤ 0.05; **≤0.01: ***≤ 0.001.
Discussion
Differences were found in the prevalence of diagnoses of mental and behavioural
disorders in children and adolescents between ZIP-code-areas, with a discernible
trend towards higher prevalence in more deprived areas. This is consistent with
previous studies that have identified neighbourhood deprivation as a potential risk
factor for mental illness in children and adolescents [6 ]
[10 ]. In addition, the COVID-19 pandemic was found to have a particularly
adverse effect on the mental health of children and adolescents from lower
socio-economic backgrounds [13 ]. Going in
line with previous research on general practitioners and paediatricians [19 ], an uneven distribution of mental
health service provider to disadvantage of deprived areas was found. Areas with a
higher density of mental health service providers have comparatively higher health
care utilisation. Area deprivation has been found to be associated with an increase
in the prevalence of mental and behavioural disorders and, among children and
adolescents with a prevalent diagnosis, with a decrease of healthcare utilization.
Moreover, there was a negative effect of the distribution of healthcare providers
on
healthcare utilization for male children and adolescents with a prevalent diagnosis.
The results show a clustering of increased need in deprived areas, with availability
of mental health service providers and healthcare utilization behaviour higher in
less deprived areas. From an economic point of view, this indicates an allocation
problem, but from a patient's point of view, it is a fundamental problem of
possible underuse and, in particular, the additional burden caused by the greater
distance to the nearest treatment centre. Additional burdens can be a barrier to use
and can be of a financial or psychological nature. In particular, socially
disadvantaged people are more likely to experience additional burdens due to a lack
of resources.
The analyses do not suggest that the associations vary across space. However, based
on previous research, a high number of unreported cases can be assumed, particularly
in areas with a negative association as young people in deprived areas are generally
more affected by mental and behavioural disorders than those in non-deprived areas
[6 ]. On the one hand, the
under-reporting can be explained by access to or general utilisation of medical
services. Secondly, it should be kept in mind that the data basis for the study is
the year 2021, which was a year influenced by the COVID-19 pandemic. An analysis of
the prevalence of the individual diagnostic groups of mental disorders before and
during the coronavirus pandemic revealed three different trends for the city of
Cologne [31 ]. On the one hand, there was
an increase in the prevalence of individual diagnostic groups that was not visible
before the pandemic or a continuous increase that was already present before the
start of the pandemic [31 ]. On the other
hand, there were decreases in the prevalence of certain diagnostic groups
potentially indicating under-reporting [31 ]. Additionally, fewer school entry examinations were carried out this
year, which puts deprived children at a particular disadvantage [32 ]. This reduction disproportionately
affects deprived children, as these examinations are crucial in identifying
developmental disorders, particularly in children with less contact with the
healthcare system. The lower general utilization of the healthcare services during
the COVID-19 pandemic may also play a role in this under-reporting.
Claims data of statutory health insurance funds have certain limitations. Diagnostic
codes only reflect administrative prevalence, i. e. cases of insured persons who
visited a service provider to receive treatment. Undetected and untreated cases are
not included. There is also a selection bias based on the collective of insured
persons of the four statutory health insurance funds providing the data. These do
not reflect the full socio-economic spectrum of all statutory health insurants in
the city. People with private health insurance (~8,7% of the population of Germany
[33 ]) are also not included in the
data set. Due to data availability, the deprivation index was calculated according
to area shares of city districts in ZIP-code areas. This may lead to a possible
underestimation of the deprivation of individual areas due to the heterogeneity of
neighbourhoods within the city of Cologne. It is possible that areas with lower
socioeconomic status of certain neighbourhoods are located exclusively in one
ZIP-code area and better-off areas of these neighbourhoods than others. Since the
values were reported for city districts and not ZIP-code areas, the effect would be
underestimated in this case. Another limitation is the formation of the deprivation
index. Due to the availability of data, it was not possible to include all of the
indicators provided for in the GISD, which is why it cannot be modelled in full.
ZIP-code areas are tailored to the demands of the postal service and do not reflect
socio-geographic logic. Therefore, these areas are heterogeneous in relation to most
variables used in this study. As no other geographical markers are currently
available in the data, the analyses were calculated on the basis of the ZIP-code
areas.
Conclusion
Healthcare utilisation was found to be higher in the city centre, with
comparatively lower prevalence and a higher density of mental health service
providers. The results suggest a lower healthcare utilization in areas with
higher deprivation and indicate a need for improved demand planning. Due to the
great heterogeneity of the population in the ZIP-code areas, this study only
provides initial indications of the link between the deprivation index and the
prevalence and utilisation of mental health services. Therefore, smaller-scale
data is required to analyse the topic in further studies. In order to make this
possible, additional steps must be taken to ensure data protection in the
transfer of data between the health insurance funds and the trust centre.
Moreover, further studies should look at the temporal development of deprivation
and the prevalence of mental and behavioural disorders as well as utilization of
mental health services in order to gain deeper insights into the interplay
between these.
Data availability
The specific dataset used for the present study is not publicly available due to
strict data protection regulations for social data. Researchers can apply for
general use of the CoRe-Dat database, subject to certain criteria and an approval
procedure.
Contributions
AP, LM, IM contributed to conception and design of the study. AP wrote the first
draft of the manuscript and performed the statistical analysis. LM, IM, TKP
supervised the work. All authors contributed to manuscript revision, read, and
approved the submitted version.
This article is part of the DNVF supplement “Health Care
Research and Implementation”
