Background

About three million children in Germany have at least one parent with mental illness (PMI) [1, 2]. Children concerned have a three to seven times higher lifetime risk of develo** a mental illness themselves [3, 4]. When parents are mentally impaired, children have an increased psychosocial risk of experiencing socio-economic descent, interpersonal conflicts, separation of parents, or negligence [1], and mental health needs of children might not be recognised or get the necessary attention [5, 6]. This makes children of parents with mental illness (COPMI) more likely to develop mental health problems compared with children from the general population [4]. Therefore, measures to detect early signs of mental disorder in offspring of PMI and interventions to support families with PMI are recommended in the literature [7, 8]. However, research results regarding the effectiveness of such interventions are ambiguous: some interventions seem to be effective in terms of symptom reduction or decreased risk of diagnosis in children [9], while others did not find intervention effects on children’s mental health or social functioning [10]. Some detected medium to large effects on parents’ symptom severity and parenting behaviour [10], whereas no difference in the effects on children’s mental health could be found when comparing interventions for both parents and children, or interventions targeting parents only [9]. The approach of this study builds upon the recommendation of Bee et al. [10] to develop feasible and acceptable child- and family-based interventions. However, the cost-effectiveness of such interventions is rarely investigated and up to now, there had been only one in Germany [11].

As a basic requirement for health economic evaluations of health-related interventions, the whole spectrum of health-related costs incurred by the study participants must be estimated [12]. In case of interventions related to mental health this includes not only healthcare costs but also costs for psychosocial support, such as accommodation support and occupational rehabilitation [13]. Otherwise there is a risk of disregarding externalisation effects, caused by the shift of costs from the healthcare system to the social care system. For children and adolescents with mental health problems, comprehensive cost assessment must also consider costs of child welfare services and services provided by schools for behavioural problems [14].

Results from international studies indicate that COPMI use health as well as child and youth services more frequently than other children [15, 16]. In Germany, the use and the costs of health and social care services in COPMI are rarely investigated [10, 15]. In contrast to other countries (i.e. the UK), there is no comprehensive unit cost list for (children and adolescent mental) health and social care services in Germany. Based on a sample of the German population, Weschenfelder and colleagues (2018) recently estimated health costs for standard treatment of children with mental illness who are not attending school as being € 8020 for 12 months [11]. However, because healthcare costs cover only a part of the total societal costs of mental health problems in children and adolescents, this study provides insufficient basis for the health economic evaluation of preventive interventions.

This paper aims to: (a) investigate the whole spectrum of health and social services used by COPMI in Germany; (b) to provide a list of unit costs for these services and to estimate the corresponding costs for the society and the health system; and (c) to identify clinical and psychosocial characteristics which affect costs and service use.

Methods

Study design and participants

The sample of this investigation includes children and adolescents (CA) who participated in a randomised controlled trial on the evaluation of a prevention programme for families with at least one parent with mental illness (PMI, see study protocol [17]). Participating families were recruited at six study sites located at hospitals or hospital departments for adult or child and adolescent mental health in Germany between April 2014 and June 2017. Sites were selected on the basis of their particular interest in supporting families with PMI known from previous cooperations. Recruitment was carried out by means of posters, flyers, information during ward rounds, personal approach and patient-parent groups, as well as through newspaper advertisements. Families were included if they had at least one child aged between three and 19 years and if at least one parent reported having been diagnosed with a mental illness (F10 to F69 ICD-10) that was currently being treated or where treatment had finished recently. Diagnoses have been cross-checked with the patient records if available. Excluded from study participation were parents or children who experience severe psychopathological symptoms or suicidal thoughts indicating the need for acute inpatient care.

Because of the expected differences between the health and social care services used by COPMI and PMI, a comprehensive investigation of costs for both groups would go beyond the scope of this article. Therefore, the current paper focuses only on the costs of COPMI.

Diagnostic assessment of children was done by trained psychologists and psychotherapists. All other assessments were carried out by trained research workers on four separate occasions: before randomisation (t0), and at 6 (t1), 12 (t2), and 18 (t3) months follow-up. Health service use by all children was assessed with the help of their parents. The diagnostic interview, severity of mental illness, and functioning was assessed by asking parents about the health status of their child(ren) and also, for children aged ten years or older, by themselves. After providing informed consent, participants were randomly assigned to either the intervention group or the control group with treatment as usual.

Instruments

Health service use and medication was measured with the Child and Adolescent Mental Health Service Receipt Inventory (CAMHSRI) [18]. The CAMHSRI questionnaire was adapted to the German psychiatric care system for CA. The CAMHSRI consists of eight parts: inpatient care, outpatient care, inpatient social services, outpatient social services, other inpatient services, school help, type of school, and medication. All types of services assess the number of consultations for the previous three months, except the parts about inpatient care, medication and type of school attended. Inpatient care and school type are assessed for the previous 12 months. The section about medication assesses the type, dosage and frequency of medication taken for the previous month. All services were assigned the corresponding cost per unit and afterwards extrapolated to 12 months. The CAMHSRI assessment took about ten minutes.

Diagnosis of children was assessed by semi-structured interviews with the German version of the Kiddie Schedule for Affective Disorders and Schizophrenia [19, 20]. Children below the age of ten have been assessed on the basis of their parents’ reports, while children from the age of ten years and up have been assessed directly. The German scale for assessing psychiatric disorders in CA (Skala zur Gesamtbeurteilung von Kindern und Jugendlichen — SGKJ) [21], the global assessment of functioning for adults (GAF) [22], the global assessment of relational functioning for adults (GARF) [23] and the clinical global impression score for adults (CGI) [24] were used for further analysis.

The SGKJ assesses current individual psychosocial functioning in CA on a hypothetical continuum and corresponds to the GAF for adults (Cronbach’s alpha 0.74 [25]), which also assesses current individual psychosocial functioning. The GARF assesses current relational functioning in adults. SGKJ, GAF and GARF are rated on a scale from 1 (dysfunctional) to 100 (fully functional). The SGKJ and the GAF distinguish in ten sections of ten points each, graduating individual functioning with higher values rating better functioning. The GARF is used as an observational instrument and addresses three major constructs (problem solving, organisation, and emotional climate) in five clinical vignettes (Cronbach’s alpha from 0.72 to 0.97 [26]). The CGI rates with a single item the severity of the mental illness of the adult patient at the time of inquiry. It is rated on the following seven-point scale: 1 = normal, not at all ill; 2 = borderline mentally ill; 3 = mildly ill; 4 = moderately ill; 5 = markedly ill; 6 = severely ill; 7 = among the most extremely ill patients. All scales enter analyses as raw scores.

Costs and use of resources

As there is no unit cost list in Germany, costs for each service have been obtained from several sources. Information on inpatient costs has been taken from the German psychiatric system of diagnosis-related groups, called PEPP (Entgeltsystem Psychiatrie, Psychotherapie und Psychosomatik) [27], costs of office-based physicians have been calculated on the basis of the Doctors' Fee Schedule within the German Statutory Health Insurance Scheme (Einheitlicher Bewertungsmaßstab, EBM) [28]. Costs of services provided by the child welfare system have been acquired via telephone survey of authorities providing costs for child and youth social services (Table 2). Defined daily dose (DDD) prices for drugs and medication were determined based on active ingredient with the German report for pharmaceutical products [29].

Statistical analyses

Standard errors and 95% confidence intervals for cost data have been estimated by means of nonparametrical bootstrap** with 1000 replications taking into account the clustering of children into families.

Regression based imputations have been performed to take into account missing values. Due to the high number of missing values, each of the imputed variables was imputed individually using costs, group and children’s diagnosis as explaining variables for the imputation. Imputed values were used for cost functions. Cost functions have been estimated by means of a two-part regression using a logistic model for the first part and a linear model with robust standard errors for the second part [30]. Both models took into account within-family clustering of the children [31]. Total annual costs were used as dependent variable (DV), age, gender, as well as baseline measures of children’s diagnosis, children’s SGKJ, parent’s mental health condition (diagnosis within the affective spectrum, CGI and GAF) and family functioning (GARF) as independent variables. A joint test was applied to confirm that all measures of parents are relevant for explaining the variance in total costs. Marginal effects (Delta-method) were calculated stepwise for all independent variables by stepwise addition of the variables into the two-part model. All analyses were performed using Stata 16.

Results

In total, 215 families with 332 CA gave their consent to participate in the study. Parents were on average 42 years old, the participating parent was mostly the mother (N = 156, 75%) and about 52% (N = 111) of the parents reported diagnoses of affective disorders (ICD-10, F32 and F33).

On average, the participating CA were about 12 years old and 172 (52%) were female. Fifty-five percent (N = 186) were diagnosed as having a mental disorder at study baseline (Table 1). Details on services, frequency, and unit costs of service utilisation are presented in Table 2. The most-reported health-related services were: CA psychiatry (n = 28), CA psychiatrist (n = 44), psychotherapist (n = 54), paediatrician and general practitioner (GP, n = 62), and occupational therapist (n = 24). The most-used outpatient child and youth services were socio-pedagogical family assistance (n = 36) and parent–child counselling centres (n = 22). In the cases of 154 persons, no services were used. Details on different drug ingredients, DDD prices and total costs for each drug are displayed in Table 3 [29].

Table 1 Sample characteristics
Full size table
Table 2 Unit costs of service utilisation
Full size table
Table 3 Overview about costs of taken drugs [29]
Full size table

The total mean costs for 12 months for the total sample amount to € 3736.35 (95% CI: € 2816.84–4813.83) per person. CA with diagnosis generated a total of € 5691.93 (95% CI: € 4146.27–7451.38) of costs per person, compared to € 1245.01 (95% CI: € 657.44–1871.49) for children without psychiatric diagnosis (Table 4). Figure 1 shows that the distribution of total costs is positively skewed, common for healthcare cost data (see Fig. 1). Mean inpatient costs amount to € 1549.70 (95% CI: € 897.57–2369.93), outpatient costs to € 383.20 (95% CI: € 283.52–489.67), inpatient child and youth services to € 442.13 (95% CI: € 133.44–809.13), outpatient child and youth services to € 258.25 (95% CI: € 170.53–355.90), school services to € 1063.89 (95% CI: € 641.75–1576.14) and medication to € 39.19 (95% CI: € 15.96–70.67). Children with diagnosis generated significantly higher costs in psychiatric inpatient (p = 0.007) and outpatient services (p < 0.001), as well as in inpatient youth services (p = 0.053), medication (p = 0.043) and total costs (p < 0.001), compared to children without psychiatric diagnosis.

Table 4 Total costs and 95% confidence intervals (CI) for 12 months
Full size table
Fig. 1
figure 1

Total cost for 12 months

Full size image

Six variables were imputed with the number of imputed observations in brackets: age (3), gender (3), SGKJ (48), GAF (40), CGI (32) and GARF (29). The logit part of the two-part regression model indicates significant odds ratios (OR) for individual functioning and diagnosis of the child as well as for family functioning (GARF, see Table 5, first part). The linear part of the two-part regression model reveals that increasing the age of the child is related to increasing costs (bage = 618.51; p = 0.037) while increasing individual functioning in the child is related to decreasing costs (bSGKJ = -368.39; p = 0.002). The logit part of the model explains about 17% of the probability for using any type of service, whereas the linear part explains about 23% of total cost variance for all cases with costs > 0. A joint test of the common effect of parental functioning (GARF, GAF and CGI) showed a chi2(6) value of 14.96 (p = 0.021). Average marginal effects for age and functioning of the child differ only slightly between the linear model and the linear part of the two-part regression model and between the imputed and the not-imputed models. Type A error levels did not differ with regard to the 5% significance criterion. The average marginal effect of age amounts to a 325.42 Euro increase in costs per year of increased age. With each one-unit increase in the SGKJ (functioning of the child), there is a 213.94 Euro decrease in costs (Table 6, the marginal effects of the linear regression model and the not-imputed models can be found in Additional file 1).

Table 5 Model 1–Imputed two-part regression model for all participants with robust estimates
Full size table
Table 6 Marginal effects of the imputed two-part model
Full size table

Discussion

To our knowledge, this is the first study investigating the use and the costs of health and psychosocial services used by COPMI in Germany.

Our results reveal that 43% of the participating CA reported having used at least one health or social service unit. As indicated by the comparison of service categories, about 50% of the total costs were incurred by psychiatric inpatient services while about 30% were incurred by non-medical services provided by child and youth welfare authorities and schools. Although costs for all service categories were significantly higher for participants diagnosed as having a mental illness, one third of the participants without a current diagnosis reported the use of at least one service unit including psychiatric inpatient treatment. These results underline that comprehensive estimation of costs associated with having a PMI should include the whole spectrum of services provided for emotional and behavioural problems in CA. Furthermore, the fact that the use of treatment and support is not limited to those CA who have been diagnosed as having a current mental disorder may either indicate that the diagnostic procedure applied in our study was not sensitive enough to identify all cases with a mental disorder, or that there is a substantial need for services below the threshold of a diagnosis in our target group. Especially the subgroup of children without diagnosis but using services (26%, N = 37) generating mean costs of € 1134.73 in the psychiatric inpatient sector, suggests that these children are not diagnosed correctly or not treated adequately or in the adequate system. To answer the question of to what extent these explanations are appropriate, representative samples of families with PMI would be needed.

The difference in results between children with and without diagnosis can be explained by the fact that even if behavioural problems already occurred in educational or welfare settings, mental disorders are in most cases only diagnosed for the first time by psychologists or psychiatrists in mental health care facilities. Accordingly, the fact that school-based support services are most widely used by children without mental health diagnosis indicates that staff providing these services may detect behavioural problems at a lower threshold [32]. Results of our cost regression model reveal that the probability of using any service is associated with the mental health-related characteristics of the children as well as those of the PMI, while the intensity and the costs of service use is associated with the age and functional capacity of the child.

These results may reflect the fact that parents’ knowledge and appraisal of mental health problems of their children determine their help-seeking behaviour [33, 34]. This implies parents’ awareness of their children’s mental health care needs, but in case of the presence of the parents’ own mental illness, this awareness might be lacking, resulting in non or delayed help-seeking [33]. CA have a mean delay in help-seeking of about four years [35]. Due to the lack of awareness, this time might be even longer in case of COMPI, resulting in an externalisation of healthcare costs to the educational or the child welfare system. In addition to lacking awareness of CA mental health needs, PMI might delay in help-seeking for their children due to shame or fear of stigmatisation [36, 37], lacking mental health literacy [38], or due to their reluctance to reveal their own mental illness to their children, teachers or educational staff [39]. However, the results may also reflect the fact that, in Germany, support for mental health problems in children or adolescents is usually provided by different facilities and also differently financed than support for adults with mental health problems. As a consequence, services for the support of families with PMIs rarely exist and the problems resulting from parenthood and mental disorder are only considered by the health and social care system if they become obvious due to significant behavioural problems of the children or adolescents.

Literature indicates that there is a relationship between caregivers’ mental health and caregiving skills [40], as well as between caregivers’ mental health and low social support [41]. Caregiving skills were not measured in this trial but we found a significant effect upon family functioning, indicating that an increased level of functioning is related to a higher probability of using any type of health or social service and of incurring costs in the first part of the model. Family functioning is known to be correlated with the mental health of COPMI [42], which is also true for our sample (r = 0.344, p < 0.001), indicating a good resilience in participating children. Further investigations about resilience in COPMI and the influence of individual and family functioning on costs are needed. Still, preventive interventions targeting family functioning are shown to be effective [43] and might be cost-saving in the long run.

In our sample, 56% of COPMI have a psychiatric diagnosis themselves. This is consistent with previous findings in the literature. Mattejat et al. [44] for example showed that about 50% of children and adolescents showing up in mental health services live with a PMI. Campbell et al. [45] even report a prevalence of mental illness of up to 79% in parents of children receiving mental health treatment. Van Santvoort et al. [46] confirmed in their review the message of Cicchetti et al. [47, 48] that COPMI are at risk of develo** mental illness—either the same as their parents or another disorder—with a strong tendency for the same disorder as their parents.

Special attention is needed for children who do have a diagnosis but who reported no costs (N = 79), indicating a lack of treatment. Therefore, there is a need to offer early help for the children of PMI as well as to raise awareness in other family members, caregivers, or GPs for noticeably different functioning and behaviour in the child.

Strengths and limitations

This study is the first study presenting primary data on comprehensive health and social care service use and costs of COPMI in Germany. This paper presents a unit cost list for health and social care services for CA with mental health problems in the German healthcare system, therefore adding significant information about youth and social service costs to recently published healthcare costs [11]. In contrast to previous studies we included the full range of school-based and child welfare services.

Limitations of the study need to be considered. First, since the participating families have been recruited in mental health service facilities, the study sample is not representative for COPMI, which limits the generalisability of our results. Second, participating parents or children might not recall all used services or drugs which can possibly lead to an underestimation of real costs. Third, assessment of the children’s psychological status via the report of parents may furthermore result in an underreporting of psychological problems in children below the age of ten. Fourth, the influence of parental diagnoses apart from depression might be underestimated, as other diagnoses are less frequent in the spectrum of mental disorders. Fifth, the use of simple regression-based imputation may underestimate the variance of the imputed variables. Sixth, since we did not measure service needs directly we can estimate the proportion of unmet service needs only indirectly. Seventh, the extrapolation of service use to 12 months might overestimate the frequency of service use and the average costs among those who use services, while the proportion of persons with any service use might be underestimated.

Conclusions

While our results in general reveal that mental and social care services are provided to those children who need support, we also identified 79 children (24%) with a diagnosis of a mental disorder who did not report any use of mental or social care services. This indicates that a significant proportion of COPMI might be disregarded by the current system of mental and social care. On the other hand, the fact that 37 of the children in our sample (11%) reported the use of mental or social care services indicates that the need for support may already exist below the threshold of a clinical diagnosis. Given the fact that the risk of being disregarded by the mental and social care system is higher for those without a diagnosis than for those who have already been diagnosed, we would expect that the proportion of children with unmet needs for support might be considerable.