Health Services Use and Costs in Individuals with Autism Spectrum Disorder in Germany: Results from a Survey in ASD Outpatient Clinics

Study data was collected in four German ASD outpatient clinics from patients with a confirmed diagnosis of a pervasive developmental disorder according to ICD-10 (F84.0, F84.1, F84.5, F84.8, and F84.9). All participants had been diagnosed by experienced clinicians using the current diagnostic gold standard in ASD, including the Autism Diagnostic Observation Schedule (ADOS), and the Autism Diagnostic Interview-Revised (ADI-R) (Bölte et al., 2006; Hayes et al., 2018; Rühl et al., 2004). Participants’ IQ had been assessed using the German versions of the following instruments: Wechsler Intelligence Scale for Children (WISC-R (Tewes, 1983), WISC-III (Tewes et al., 1999), WISC-IV (Petermann & Petermann, 2011)), Wechsler Adult Intelligence Scale (WAIS-R (Tewes, 1991), WAIS-III (von Aster et al., 2006), WAIS-IV (Petermann, 2012)), Wechsler Preschool and Primary Scale of Intelligence (WPPSI-III (Petermann et al., 2009)), Kaufman Assessment Battery for Children (Melchers & Preuß, 2009), Wortschatztest (Schmidt & Metzler, 1992), Raven’s Standard Progressive Matrices (Horn, 2009), and Raven’s Coloured Progressive Matrices (Bulheller & Häcker, 2002). While intellectual disability is defined as an IQ < 70, some studies on ASD have included “borderline intellectual functioning” (IQ = 71–84) (Baker & Blacher, 2020). Yet, according to ICD-10 (Dilling & Freyberger, 2012), an IQ between 71 and 84 is defined as a learning disability. Referring to this definition, in the current study IQ levels were dichotomised into learning disability/ intellectual disability (IQ < 85) vs. no learning disability/ intellectual disability (IQ ≥ 85).

Sociodemographic data (including age, sex, and educational attainment) and health service utilisation was assessed through a questionnaire which included a modified version of the Client Service Receipt Inventory (CSRI). The CSRI is a well-established semi-structured interview, in which participants are asked about health services used over the preceding 12 months (Beecham & Knapp, 2001). The instrument can be administered to the service user face to face or by post for self-completion (Patel et al., 2005).

All people who had received any services in one of the four ASD clinics were invited to participate. 1419 survey documents were sent to patients with ASD or their parents/legal guardians (of these, 125 documents could not be delivered due to a wrong address). Patients and caregivers, respectively, received the questionnaire via postal mail and were asked to consent both to the study, and to the pseudonymised linkage of study data with their clinical data (e.g. clinical diagnosis, level of intellectual functioning, ADOS comparison score). Specific data on socioeconomic status and race/ethnicity was not recorded. For participants under 18 years of age, parents/legal guardians filled in the questionnaire. The questionnaire for adults with ASD was directly addressed to these. We assumed that the majority of adult participants with an IQ < 85 were assisted in filling in the questionnaire. 418 persons returned the questionnaire including a signed written consent form (response: 32.3%), of which 385 questionnaires could be evaluated. Characteristics of survey responders vs non-responders are reported in the results section.

Within the German health insurance system (Busse & Blumel, 2014), the majority of health services is largely free of charge for patients; including inpatient and outpatient health care, prescription medication, rehabilitation, non-physician practitioner services (e.g. speech therapy, occupational therapy), and—rather exceptional within Europe—psychotherapy (Signorini et al., 2017). Regarding patients with ASD, most health services (e.g. diagnostic procedures, follow-up examinations, medication, interventions for co-occurring disorders) are reimbursed by statutory health insurance funds. Nevertheless, interventions or support concerning the core symptoms of ASD are covered by integration support aid (“Eingliederungshilfe”) which is funded by social services. As these interventions are provided by different public and private institutions, comprise very diverse approaches (from facilitated communication to behavioural interventions and others), and are not part of the statutory health insurance system, we did not include them in our study.

CSRI items related to contacts with all types of healthcare professionals, and duration of all inpatient and rehabilitation treatment in the last 6 or 12 months, respectively. Outpatient drug prescriptions were assessed for the previous 4 weeks. All figures were extrapolated to a 12-month time period, under the assumption that the data were representative of the entire year.

Questionnaire data was entered by one person in an electronic Case Report Form (eCRF) created in OpenClinica® (OpenClinica Enterprise Version: 3.3), and was checked by another person.

Based on the CSRI data, annual service use costs for inpatient and outpatient medical care were calculated, and stratified by sex, age group, diagnostic category, ADOS symptom score, intellectual functioning (IQ < 85 vs. IQ ≥ 85), and educational level. The level of education was defined in accordance with the International Standard Classification of Education (ISCED) (UNESCO, 1997, 2012), and was classified into three categories: low (ISCED level 0–2B), medium (level 2A) and high education (level 3A). Referring to the German school system, low educational level complies with 9 years of schooling or leaving school without having acquired any school-leaving qualification. Medium educational level is equivalent to 10 years of schooling, and high educational level complies with 12 or 13 years of schooling and a school-leaving qualification, which opens access to higher educational institutions (Schneider, 2008; Schroedter et al., 2011).

Direct costs were calculated by multiplying each resource use item by an appropriate unit cost (macro-costing). All costs were assessed from the perspective of the social security system and were calculated in Euros based on prices of the year 2016, obtained from a number of sources (Table S1 in Additional File 1): Where possible, medical services, including consultations of medical practitioners and non-physician practitioners (physical therapy, occupational therapy, speech therapy), as well as hospital stays and rehabilitation were valued by unit costs according to the calculation of Bock et al. (2015). Costs for outpatient hospital care as well as consultations of child and adolescent psychiatrists and psychotherapists for children were estimated by experts, using EBM-Codes (Doctors’ Fee Scale within the German Statutory Health Insurance Scheme) and/or quarterly fixed amounts. In cases of fixed amounts, a maximum of four were charged for participants who stated more than three contacts per year. For adults, the cost of medication was estimated by means of defined daily doses (DDD) using the Drug Prescription Report 2017 (Arzneiverordnungs-Report, 2017) or previous issues for example if drugs were withdrawn from the market. For children, the DDD and thereby the costs were estimated using either the Drug Prescription Report or were calculated based on the information given in the summary of product characteristics of each drug. In the latter case, the DDD were assessed based on the assumption of a twelve years old child with a weight of 41.5 kg (50th age-adjusted percentile, www.​pedz.​de). Finally, daily drug costs for long-time therapy drugs were multiplied by 365.25.

Baseline data were analysed using descriptive statistics. Mean values and standard deviations for healthcare service use, stratified by type of service (inpatient services, other hospital-based services, outpatient services and non-physician practitioner services), were calculated separately for children (0–11 years), adolescents (12–17 years) and adults (≥ 18 years). To account for non-normality of cost data, average annual costs per individual and by type of service were estimated with 95% confidence intervals (95% CI), using the bias-corrected accelerated bootstrap method (Barker, 2005) based on 5000 replications. Additionally, in order to detect associations between potential predictors (sex, age, ASD diagnoses, intellectual functioning and ADOS comparison score) and total costs, a generalised linear model assuming a gamma distribution with a logarithmical link function was fitted to calculate cost ratios. All statistical analyses were performed with SAS, version 9.4 (SAS Institute, Cary, US). The underlying research data are available from the corresponding author upon reasonable request.

The number of existing questionnaire-based studies on health costs in individuals with ASD is very scarce, Table 6 provides an overview of the four relevant studies. As comparability with older studies may be hampered by differing diagnostic criteria (e.g. ICD-9 vs. ICD-10), divergent therapeutic guidelines, different reimbursement policies, or limited availability of autism-specific health services, the table contains only studies from the last ten years. The study of Xiong et al. (2011), while matching the general comparability criteria, was excluded because not enough details for reasonable comparison were reported.

When comparing the cost figures from Table 6, it has to be kept in mind that there are considerable differences between the health systems of the respective countries. For example, while in Germany psychotherapy both for children and adults is broadly available and reimbursed by statutory health insurance funds, in the UK psychological therapies are less accessible (NHS England, 2018), with patients often paying out-of-pocket for private therapists. Also, the mental health system in the UK is designed with a much stronger focus on outpatient and/ or day-care treatment; in contrast to Germany, which is top of the table in Europe in terms of inpatient beds per 100,000 inhabitants (Signorini et al., 2017).

Moreover, study designs differ in terms of sample size, age range, diagnostic quality, included diagnostic categories, and cost composition.

In the study of Barrett et al. (2012), costs for healthcare, medication, and community social and voluntary services in UK toddlers with ASD were £1383/6 months, which is very close to our findings. As their sample only included 2- to 5-year old children, who usually do not use health services extensively, the similarity in cost sizes is probably due to substantial costs for nursery placement for most children in the Barrett et al. study.

In contrast, in the study of Barrett et al. (2015), costs (£500/6 months) were much lower than in our current study. This is somewhat surprising, as participants were on average 15.7 years old—an age, in which patients with ASD usually show a number of co-occurring conditions which need therapy. While costs for hospital services are roughly comparable to our findings, costs for outpatient (“community”) services are much lower than in our study, which may indicate a lower availability of these services in the UK health system.

In relation to the results of Roddy and O’Neill (2019) from Ireland, our cost figures are more than 2.4-fold higher, which may be due to the stronger focus on cheaper outpatient services within the Irish health system, and the restricted access to medical resources in this country (Thomas et al., 2014).

Finally, the “true” costs for children with ASD in the study by Lavelle et al. (2014) are most likely higher than our data, as their study only reported the additional costs for children with ASD in the US, i.e. the difference to health costs for children without an ASD diagnosis. In their sample, patients with an ASD diagnosis had significantly more outpatient visits, home healthcare visits, and prescription medication.

Remarkably, the average individual costs found in our questionnaire-based study come very close to the data from a cost-of-illness model for individuals from Germany with ASD and an IQ > 70 (0–18 years: 2473–3252 EUR, > 18 years: 3664 EUR) (Bachmann et al., 2013).

In our study, costs for inpatient services were the largest cost component (over all age groups: 19.8% of total cost), accounting for 36.0% of total costs in adults with ASD. At least for psychiatric healthcare services in children with ASD, the preponderance of inpatient service cost has also been shown by other researchers (Croteau et al., 2019). The afore-mentioned cost structure is similar to the health cost distribution in schizophrenia, where inpatient costs also are the greatest component of health service costs in most countries (Kovács et al., 2020), including Germany (Konnopka et al., 2009). Yet, unlike in schizophrenia, inpatient costs in ASD are potentially mainly driven by the treatment of co-occurring psychiatric conditions. The rather comparable dimension of inpatient treatment costs in ASD and schizophrenia stands in contrast to the differences of these conditions in terms of stigma (much lower in ASD) and quality of life (much higher in ASD) levels (Bachmann et al., 2019; Kamp-Becker et al., 2010). In opposition to this, the composition of health costs for individuals with ADHD shows a higher share of outpatient costs compared to inpatient costs (Daley et al., 2020).

The fact that the overall cost shares of pharmacotherapy (11.1%) and occupational therapy (11.1%) are similar is somewhat surprising, given the shaky evidence for occupational therapy for ASD-related symptoms (usually sensory features) (Watling & Hauer, 2015). Even more surprising is the proportion of occupational therapy in children, namely 26.0%, despite this treatment modality not being recommended in this age group for the treatment of ASD core symptoms (Freitag et al., 2020; National Institute for Health & Care Excellence, 2013). Given the high popularity of occupational therapy in Germany also for other neurodevelopmental disorders in children and adolescents, e.g. ADHD (Braun et al., 2013), our findings seem to be a German peculiarity, which probably reflects the good availability of these services (in contrast to e.g. early intensive behavioural interventions) and prescribers’ lacking knowledge of evidence-based interventions for ASD. Nevertheless, in our opinion, occupational therapy constitutes a rather unfounded cost driver (Haverkamp et al., 2017).

In this study, lower IQ, female sex, and Asperger syndrome were associated with higher health costs in individuals with ASD, while age or symptom severity (as indicated by the ADOS comparison score) were not. Regarding the association of lower IQ with higher costs, this result is in line with previous studies. The study of Buescher et al. (2014), for example, reported 1.6- to 1.7-fold greater lifetime costs for individuals with an ASD and intellectual disability (ID) vs. individuals with an ASD and no ID, which fits well with the cost ratio of 1.9 from our study.

The association of female sex with higher ASD health costs is a novel finding. One possible explanation are statistical effects—often, ASD studies suffer from low numbers of females (Reinders et al., 2019), therefore some gender-related effects may be difficult to detect. Another potential cause is higher psychiatric comorbidity in females with ASD, which may lead to higher healthcare costs through higher hospitalisation rates: In a German sample of adults with ASD, females had higher comorbidity rates (42%; vs. 30% in males) (Strunz et al., 2014). For children and adolescents, a Swedish study yielded similar results regarding symptom severity for comorbid ADHD, learning disabilities, and oppositional defiant disorder (Lundström et al., 2019). Yet, the existing literature regarding psychiatric comorbidity rates is not unequivocal: In a Swedish study of adults with ASD and normal intelligence (including a high portion of pervasive developmental disorder not otherwise specified), no difference between men and women in terms of lifetime psychiatric axis I diagnoses were found (Hofvander et al., 2009). Finally, the higher costs for females with ASD may be partly due to lengthy and cost-intensive diagnostic procedures, because making a correct diagnosis of ASD in females can be more complex and challenging than in males (Lundström et al., 2019; Ratto et al., 2018; Young et al., 2018). Of course, our finding needs to be replicated in a well-composed and reasonably sized sample of individuals with ASD.

The association of Asperger syndrome with higher costs is novel, too. A possible explanation for this finding is that mental health needs of other individuals with ASD, e.g. those with childhood autism and intellectual disability may be overlooked or misdiagnosed (example: signs of depression being interpreted by carers as ‘difficult behaviour’), thus leading to lower service use.

Interestingly, in our study age was not associated with costs. This stands in clear contrast to other studies which reported higher costs in older individuals (Barrett et al., 2015; Rogge & Janssen, 2019). A possible explanation are the high costs for occupational therapy and speech therapy for children in our sample. Finally, the lacking influence of ASD symptom severity (as reflected by the ADOS comparison score) is fully in line with the available literature so far (Barrett et al., 2012, 2015; Roddy & O'Neill, 2019).

This study constitutes the first survey-based health cost study in adults with ASD, and is one of the few international studies that are based on real-world service use data. Moreover, the sample included a broad age range (spanning 4–67 years), and a significant proportion of females. To ensure high quality diagnoses, the diagnostic assessment of participants employed the current gold standard; which is key given the significant rates of over- and misdiagnoses in ASD (Bachmann et al., 2018).

A clear limitation of this study is the sample composition: This study is not a population-based study, but a survey of patients of four large German ASD outpatient clinics. While the patient population of these clinics is representative for a typical clinical population of individuals with ASD in Germany in terms of geography (the patients seen in the four study centres stem from all German federal states, with the exception of Mecklenburg-Vorpommern), age, male/female ratio, IQ, and level of ASD symptom severity), it only contains those service users with a formal ASD diagnosis.

Moreover, the above-average portion of individuals with Asperger syndrome in our sample may have led to a bias in health costs distribution.

Another limitation is the lack of information on co-occurring health conditions, on adaptive functioning (Szatmari et al., 2015), and on participants’ socio-economic status, which probably are important service use (Drahota et al., 2020) and cost predictors in addition to the predictors studied in this work. Also, as the CSRI time frame included service use in the previous 12 months, a recall bias is possible.

While health costs are an important cost component in ASD-related (especially regarding the potentially higher morbidity and mortality), other cost types, e.g. educational costs, can be more important cost drivers in ASD-related societal costs (Bachmann et al., 2013; Rogge & Janssen, 2019). Finally, due to differences in health systems between countries, generalisability of our results is limited.