Depression and Quality of Life in People with Autism Spectrum Disorder -
A Network Approach
Lena Christ University of Amsterdam Student number: 10204741 Word count: 4671Table of Contents
Abstract………
Introduction………..
Method……….
Data Source and Participants………...
Materials………... Procedure……….. Data-Analysis………... Results……….. Participant Characteristics……… Network Comparison………...
Measures of Node Centrality………...
Discussion……… References……… Appendix……….. 3 4 7 7 8 9 9 11 11 12 15 16 20 23
Abstract
Autism spectrum disorder (ASD) is a neurodevelopmental disorder with high rates of
comorbidity. Depression is especially common among people with ASD. Depression, in turn,
is linked to poor quality of life (QoL). This study used the network approach to analyse how
the network of QoL domains differs for people with ASD and people who are prone to
depression. Eight hundred and forty people with ASD completed a QoL questionnaire. One
hundred and eighteen people without ASD completed the same questionnaire and an
additional questionnaire that measures proneness to depression (PtD). The analysis showed
that the ASD group’s QoL network was no more similar to the PtD group’s network than to
the network of healthy controls. Nevertheless, some interesting differences and similarities
Depression and Quality of Life in People with Autism Spectrum Disorder Autism spectrum disorder (ASD) is a life-long neurodevelopmental condition
characterised by deficits in the ability to relate to and communicate with others. ASD is
widespread and the symptomatology of people with ASD is highly diverse: while some are
able to live an independent life, others will need constant care. This is partly due to the high
rates of comorbidity between ASD and axis one disorders (Matson & Nebel-Schwalm, 2007;
Simonoff et al., 2008), especially with anxiety and mood disorders (Ghaziuddin et al., 2002;
Hofvander et al., 2009). Unfortunately, depression is often not recognized in people with
ASD because of symptom overlap (Stewart, Barnard, Pearson, Hasan, & O’Brien, 2006).
Additionally, people with ASD often find it difficult to identify and communicate their
emotions (Ghaziuddin et al., 2002; Stewart et al., 2006). As a result, numerous people with
ASD suffer from untreated depression (Stewart et al., 2006). People with depression are more
frequently sick, are often unable to work and need more help with their daily chores
compared to healthy individuals. This produces high costs for society such as health care
related costs and unemployment payments (Luppa, Heinrich, Angermeyer, König, &
Riedel-Heller, 2007). In addition, ASD symptoms are often aggravated in people who suffer from
both ASD and depression (Stewart et al., 2006). Because of these high rates of undetected
depression in people with ASD it is important to find a way to detect proneness to depression
(PtD) and prevent the onset of a depressive episode in the ASD population.
One way of detecting whether someone is likely to develop depression could be to
look at their quality of life (QoL). QoL refers to someone’s subjective experience of
well-being. It is a psychological construct that consists of a set of interrelated domains referring to
a person’s psychological, physical and social functioning (Trivedi et al., 2010). It can be
example of a QoL domain would be someone's social life or someone’s living situation. Prior
research has shown that depression is strongly tied to poor QoL (Cramer, Torgersen, &
Kringlen, 2010; Trivedi et al., 2010). Although poor QoL is generally treated as a symptom
of depression rather than a cause (Cramer, Torgersen, & Kringlen, 2010; Rapaport, Clary,
Fayyad, & Endicott, 2005), the causal relationship between the two has yet to be determined.
It is likely that poor QoL and depression influence each other and that improving one could
possibly improve the other. Therefore, I suggest that QoL should not only be viewed as an
outcome instrument but also as a multidimensional tool that can help detect PtD and may
even offer opportunities to prevent the onset of depression. This study aims to identify the
domains of QoL which are related to PtD in people with ASD and show how these domains
interact with each other. As stated above, multiple studies have shown that people with ASD
are more prone to depression than healthy controls (Ghaziuddin et al., 2002; Hofvander et al.,
2009). There is also a vast amount of evidence suggesting that QoL is significantly lower for
people with ASD than for healthy controls (Saldana et al., 2009; van Heijst & Geurts, 2014).
Therefore, it is assumed that the finding that depression is strongly tied to poor QoL (Cramer,
Torgersen, & Kringlen, 2010; Trivedi et al., 2010) extends to people with ASD, even though
this connection has not yet been studied in the ASD population.
Traditionally in research QoL is viewed as a latent, unidimensional variable that is
formed by a number of independent predictors, the domains of QoL (Trivedi et al., 2010; van
Heijst et al., 2014). Accordingly, the majority of research concerning QoL in people with
ASD has focused on identifying specific predictors of QoL. For example, it was found that
social support, academic functioning and/or satisfactory employment, family life and
self-determination are good predictors of QoL in the ASD population (Burgess & Gutstein, 2007).
Those findings are supported by the results of two studies that showed that people with ASD
support had a bigger impact on QoL in people with ASD than disability characteristics had
(Billstedt, Gillberg, & Gillberg, 2010; Renty & Roeyers, 2010). The latter suggests that poor
QoL in people with ASD is not the direct result of the disorder but rather that QoL is
influenced by one’s environment, which - at least in most cases - can be changed. A problem
with the latent variable approach, however, is that it does not take into account if and how
these identified domains of QoL interact with each other. For example, someone with severe
disabilities might not have as many opportunities to meet people and build friendships as
someone with less severe disabilities. The number and quality of friendships could, in turn,
influence a person’s social support. Consequently, disability characteristics influence social
support, which shows that the latent variable approach, in which the predictors are
independent from each other, is not suitable for conceptualising QoL. Another thing to
consider is what kind of inferences for the clinical practice can be made from each approach.
If the domains of QoL are independent from each other, as proposed by traditional research,
we would need to approach them each separately in order to improve QoL as a whole. In a
multidimensional approach, however, we can focus on the domain that has the greatest
impact on all other domains, and through improving this one, high-impact domain, we can
achieve improvement in other domains simultaneously. Thus, a multidimensional view of
QoL is not only more realistic but also much more efficient in regards to practical inference.
Therefore, I suggest a multidimensional approach to QoL in which the domains are not
independent but interconnected.
Some researchers argue that multidimensional systems are represented best as a
network of interconnected variables (Borsboom & Cramer, 2013; Schmittmann et al., 2013).
In such a network model, variables, in this case the domains of QoL, are not viewed as
predictors of a latent variable but as components of an interactive network (Borsboom,
network analysis to a wide range of psychological constructs such as depression (Cramer,
Borsboom, Aggen, & Kendler, 2011; Cramer, Waldorp, van der Maas, & Borsboom, 2010;
Van de Leemput et al., 2013), generalized anxiety disorder (Cramer, Waldorp, et al., 2010),
posttraumatic stress disorder (McNally et al., 2014) and even QoL (Kossakowski et al., in
prep.). In conclusion, network analysis has been shown to be a fruitful alternative to the latent
variable approach (Borsboom & Cramer, 2013; Schmittmann et al., 2013).
In this exploratory study QoL is viewed as a multidimensional system of interrelated
domains. We are interested in how the network of QoL domains differs for people with ASD
and people with PtD. No specific hypotheses were tested. Generally, it would be expected
that the QoL networks of people with ASD and people with PtD will be more similar to each
other than to the QoL network of healthy controls. If we can identify domains of QoL that
fulfil an important role in the network of people with ASD (e.g. closely related to other
domains, or central in the network), this could provide a roadmap for clinicians and
practitioners regarding treatment options. In addition, if we can identify domains or structures
that are similar in the QoL networks of people with PtD and people with ASD, we can
assume that those domains or structures play a role in developing depression in both
populations.
Method Data Source and Participants
Two datasets were included in the analysis. The data for the ASD group were
originally gathered by researchers of the Dr. Leo Kannerhuis and consists of data from 1032
people diagnosed with ASD. The Dr. Leo Kannerhuis is a Dutch non-profit organisation
The data for the two non-ASD groups, the PtD and the control group, were gathered
by our own research team. One hundred and seventy eight participants were recruited via
social media. No compensation was offered due to the minimal time burden of completing
the questionnaires. Participation was not restricted by age, gender or occupation. Participants
were assigned to one of the two non-ASD groups according to their score on the PtD
questionnaire.
Materials
QoL was measured with the Dutch Quality of Life and Care (QoLC) questionnaire
(Wennink & Van Wijngaarden, 2004), which was used by the Dr. Leo Kannerhuis and
therefore allowed comparison of the two datasets. Each of the 11 items refers to a specific
domain of QoL. The domains measured are: bodily functioning, autonomy, psychological
functioning, living situation, occupation, financial situation, social contacts, personal
relationships, sex life, leisure time and life in general. Participants were asked to rate how
satisfied they are with those 11 domains on a 10-point Likert scale where 1 indicates “not
satisfied at all” and 10 indicates “completely satisfied”. A high score on this questionnaire
indicates a better QoL than a lower score. The minimum score is 11 points and the maximum
score is 110 points.
Participants in the non-ASD groups were asked to fill in an additional PtD
questionnaire. The nine items on this questionnaire were adapted from the Dutch version of
the PHQ-9 (Kroenke, Spitzer, & Williams, 2001). Each item refers to a common symptom of
depression: having little interest or pleasure in doing things; feeling down, depressed or
hopeless; trouble falling or staying asleep, or sleeping too much; feeling tired or having little
energy; poor appetite or overeating; feeling bad about oneself or feeling like a
speech and movement or being agitated; having suicidal thoughts or thoughts about hurting
oneself. Participants were asked to rate how frequently they experienced those symptoms in
the last two months compared to others on a 5-point Likert scale ranging from “a lot less
frequent than others” to “a lot more frequent than others”. The questionnaire handles a
minimum score of 0 points and a maximum score of 36 points. The higher someone scores
the more likely they are to develop depression. For this study a cut-off score of 27 points was
selected, which is equivalent to answering “a little more frequent than others” on all 9 items.
This means that participants who scored 27 points or higher were assigned to the PtD group.
Procedure
The questionnaires were filled in online. Participants were redirected to the
questionnaires via a link on a social media website where they were asked to agree to an
informed consent which provided information about the aim and procedure of the study. It
also stated that participation was completely voluntarily. All participants in the non-ASD
groups filled in the QoLC questionnaire first and the PtD questionnaire second. There was no
time limit.
Data-Analysis
The data-analysis was adapted from Costantini et al. (2014) and McNally et al.
(2014).
Using the R package qgraph by Epskamp, Cramer, Waldorp, Schmittmann and
Borsboom (2012), three networks for each group were computed. The domains of QoL are
represented by the network nodes. Nodes are connected to each other through edges. Each
edge has a weight that indicates how strongly two nodes are associated with each other. The
correlation networks which show all correlations between nodes without controlling for
confounding variables were computed. Correlation networks are useful to get a first
impression about associations but they also show spurious correlations, correlations that are
better explained by another node in the network. For example, ice cream sales and drowning
rates at a public pool would probably strongly correlate in a correlation network, implying
that ice cream sales cause drownings or vice versa. In reality, an increase in visitor counts is
probably the reason for both occurrences. If more people visit the pool, the chances of
someone buying ice cream or drowning will increase, too. To avoid drawing wrong
conclusions based on spurious correlations, it is best to control for them. To do so, partial
correlation networks were computed. In a partial correlation network only direct correlations
are displayed. Thick edges in a partial correlation network indicate a strong correlation
between two nodes that cannot be explained by another node in the network. Both the
correlation network and the partial correlation network are completely connected. This means
that the partial correlation network might still display some spurious associations that are
based on measurement errors. Therefore, even a partial correlation network could be
misleading because it might display a weak association between two nodes when in reality
there is no association at all. This is why a third kind of network was computed, the adaptive
LASSO network. In an adaptive LASSO network edges with a very low weight are removed
from the partial correlation network based on the assumption that the associations displayed
by low weight edges are caused by measurement errors and do not actually exist. The
advantage of the adaptive LASSO network is that it only shows associations that are really
there, the disadvantage is that weak associations might be missed since they are not displayed
in the network. Therefore, it is always best to compare the partial correlation network to the
Finally, the measures of node centrality for the partial correlation and the adaptive
LASSO networks were computed: betweenness, closeness and strength. These measures
indicate the importance of a node to a network. A higher score indicates greater centrality, or
importance. The betweenness score describes how often a node is lying on the shortest path
between two other nodes. Thus, the higher the betweenness score, the more spurious
correlations this node explains. The closeness score is calculated by dividing the mean
distance from one node to all other nodes through 1. A node that is directly linked to all other
nodes in a network would have a closeness score of 1. The strength of a node is the sum of all
edge weights of a node. The measures of centrality are displayed in multiple plots, three
separate plots for each group and one additional plot that compares the measures of node
centrality for the partial correlation networks between the groups. Node centrality in the
adaptive LASSO networks was not compared between groups because adaptive LASSO
networks are dependent on sample size and therefore not comparable between groups with
different participant counts.
Results Participant Characteristics
A total of 120 participants without ASD completed the QoLC and PtD questionnaires.
Two participants were excluded from the analysis because they did not answer all questions.
Of the remaining 118 participants 51 scored 27 points or higher on the PtD questionnaire
(M=30.88, SD=3.20) and were therefore assigned to the PtD group. The mean age in the PtD
group was 22.2 years old (SD=2.91) with 74% women. The mean QoLC score in the PtD
group was 77.88 points (SD=11.25). The remaining 67 participants scored lower than 27
control group. The mean age in the control group was 26.57 years old (SD=10.23) with 70%
women. The mean QoLC score in the control group was 85.34 points (SD=9.03). The Dr. Leo
Kannerhuis provided the dataset for the ASD group. This dataset contained a total of 1032
people with ASD who completed the QoLC questionnaire. One hundred and ninety one
participants were excluded from the analysis because they did not answer all questions and
one participant was excluded because he did not fulfil the minimum age requirement of 16
years. This age requirement was chosen retrospectively because the youngest participant in
the non-ASD groups was 16 years old. The remaining 840 participants in the ASD group
were on average 31.06 years old (SD=12.94) with 27,5% women. The mean QoLC score in
the ASD group was 70.94 points (SD=13.15).
Network Comparison
Similarities between the ASD and the PtD group. Based on the correlation networks displayed in Figures A1, A2 and A3 (in the appendix) the ASD, PtD and control
group seem to be fairly similar to each other. The domains of QoL are densely interconnected
with life in general being the most central node in all three correlation networks. However,
judging by the partial correlation and adaptive LASSO networks displayed in Figures 1, 2, 3,
A4, A5, A6, A7, A8 and A9 there are some distinctive differences between the groups.
Among all three groups there is a strong positive association between psychological
functioning and life in general which is most prominent in the ASD group’s networks (Fig. 1,
A4 and A7) and least prominent in the control group’s network (Fig. 3, A6 and A9). Social
contacts and social relationships show a high enhancing association in the networks of the
ASD (Fig. 1, A4 and A7) and PtD group (Fig. 2, A5 and A8) but only a moderate association
Characteristics unique to the ASD group. Occupation and leisure time show a high enhancing association in the ASD group’s network (Fig. 1, A4 and A7) but only a weak
association in the non-ASD groups’ networks (Fig. 2, 3, A4, A5, A6 and A7). The association
between autonomy and financial situation is quite prominent in both non-ASD groups’
networks (Fig. 2, 3, A4, A5, A6 and A7) but not in the ASD group’s network (Fig. 1, A4 and
A7).
Characteristics unique to the PtD group. Psychological functioning and leisure time show a high enhancing association in the PtD group’s network (Fig. 2, A5 and A7) but
neither in the ASD group’s network (Fig. 1, A4 and A7) nor in the control group’s network
(Fig. 3, A6 and A9).
Next to the positive associations mentioned above, there are also some weaker
negative associations displayed in the partial correlation networks of all three groups.
However, those seem to disappear in the adaptive LASSO network of the ASD (Fig. 1, A4
and A7) and the control group (Fig. 3, A6 and A9). The negative associations seen in the
partial correlation network of the PtD group are stronger, as is confirmed by the adaptive
LASSO network (Fig. 2, A5 and A7). Those negative associations are most prominent
between bodily functioning and social contacts, between bodily functioning and financial
Figure 1. Partial correlation and adaptive LASSO network of the ASD group. Positive
associations are displayed in green and negative associations are displayed in purple.
Figure 2. Partial correlation and adaptive LASSO network of the PtD group. Positive
Figure 3. Partial correlation and adaptive LASSO network of the control group. Positive
associations are displayed in green and negative associations are displayed in purple.
Measures of Node Centrality
The plots of node centrality for the individual groups are shown in the appendix in
Figures A10, A11 and A12. When comparing node centrality in the partial correlation
networks among groups (Fig. 4), it is notable that nodes in the non-ASD groups score higher
on closeness and strength than nodes in the ASD group, while nodes in the PtD group score
higher on closeness and strength than nodes in the control group (PtD>control>ASD). This
implies that the PtD group’ partial correlation network is the densest of the three partial
correlation networks.
Looking at the centrality scores of the individual nodes, most of the above mentioned
differences and similarities between groups are supported. For example, autonomy is less
central in the ASD group’s network than in the non-ASD groups’ networks, financial
situation scores high on closeness and strength in the PtD group’s network only,
and PtD group compared to the control group, and life in general scores high on closeness
and strength throughout groups.
Figure 4. Measures of node centrality for the partial correlation networks of the ASD, PtD
and the control group displayed in red, blue and green, respectively.
Discussion
This study used the network approach to analyse how the network of QoL domains
differs for people with ASD and people with PtD. Our expectation, that the QoL networks of
people with ASD and people with PtD will be more similar to each other than to the QoL
network of healthy controls, could not be confirmed. Nevertheless, one similarity between the
ASD and the PtD group that was not present in the control group was identified: The domains
social contacts and personal relationships had a stronger association in the ASD and PtD
group than in the control group. While those two domains overlap quite a bit, social contacts
are more superficial than personal relationships which refer to the bond we have with the
ASD and the PtD group than in the control group. These findings could indicate that social
contacts in people with ASD and people with PtD are often limited to family and a few close
friends. This is in line with prior research and highlighting that building and maintaining
relationships is often difficult for people with ASD (White & Roberson-Nay, 2009).
Therefore, expanding the social environment of people with ASD could have a positive effect
on QoL in general.
Although it was not found that people with ASD have more in common with people
that are prone to depression than with healthy controls, some structures that are unique in the
networks of people with ASD could be identified: Occupation and leisure time are associated
strongly in people with ASD but not in people without ASD. Occupation is also more
important to QoL in people with ASD. This can easily be explained if we take a closer look at
the QoLC questionnaire which defines occupation as how someone spends their days.
Considering that people with ASD are often unemployed, it makes sense that occupation and
leisure time are often equal in people with ASD. In addition, prior research showed that
people with ASD are especially dissatisfied with their occupation and leisure time (Billstedt
et al., 2010). This could suggest that the day-to-day activities available to people with ASD
are in need of improvement. An idea would be to include people with ASD who are unable to
work in volunteer programs where they can spend their time in a more meaningful way. This
could also help with improving the social aspects mentioned above if, for example, we would
offer to engage individuals with ASD in group projects that aim to create social interaction.
There are also some unique structures in the QoL network of people with ASD that are less
obvious. Financial situation and autonomy seem to be less important for people with ASD
than for people without ASD. This might be because people with ASD are often dependent
generally need more help with their daily chores, which makes them less autonomous than
people who do not need as much help.
However, there were also quite a few limitations to this study which might explain
why there are not more similarities between the QoL networks of people with ASD and
people with PtD. Firstly, participants in the PtD group did not score as low on the QoLC
questionnaire as expected. Even though QoL was lower in people with PtD than in healthy
controls, it was not as low as in people with ASD. One possible explanation for this could be
that the PtD cut-off score for assigning participants to the PtD group was chosen too low.
This could mean that not everyone in the PtD group was actually prone to depression. One
suggestion for further research would be to either choose a higher cut-off score or to include a
control question in the PtD questionnaire such as: “Would you consider yourself prone to
depression?” or “Did you ever go through a depressive episode?”. Another thing to consider
is that the PtD questionnaire was designed specifically for this study and was not validated in
clinical trials beforehand. Therefore, it is possible that the PtD questionnaire did not reliably
measure PtD. It is also notable that participants in the PtD group were on average about four
years younger than participants in the control group. Even though four years is not a big age
difference, it could be a possible explanation for the difference in PtD scores considering that
most participants in the PtD group were university students in their early twenties, which is
arguably a very stressful time for young adults. The age difference between the control group
and the ASD group was even bigger with people in the ASD group being 5,5 years older on
average. It is not known how age influences QoL but it is plausible that people in their thirties
have a higher QoL than people in their twenties. This could explain why there was not a
bigger difference in QoLC scores between the ASD and the control group.
Secondly, our study is based on the assumption that people with ASD are prone to
for the ASD group was provided by the Dr. Leo Kannerhuis, I did not have the opportunity to
check that assumption and measure PtD in the ASD group. It is possible that people in the
ASD group would have scored lower than expected and would therefore be more similar to
the control group in terms of PtD. This could possibly account for the incongruent results of
the QoL measures.
Thirdly, sample sizes varied greatly among the ASD and the non-ASD groups. It is
possible that the PtD and the control group did not count sufficient participants in order to get
reliable results. The fact that node centrality differed quite a lot between network types in the
ASD groups but not in the ASD group supports this notion and indicates that the
non-ASD groups’ networks are subject to more error. This made it also difficult to compare
groups with each other. I could have chosen to include only a small sample of the ASD group
to match the non-ASD groups. However, prior analysis had shown that a random sample of
67 people with ASD did not result in a reliable network. Therefore, I suggest that this study is
replicated with a larger non-ASD sample.
Lastly, sex was distributed unevenly among groups. In the ASD group participants
were predominantly male and participants in the non-ASD groups were predominantly
female. It is possible that QoL is structured differently in men and in women. If this is the
case, the ASD group and the non-ASD groups might not be comparable.
In conclusion, QoL in people with ASD is likely structured differently than QoL in
people who are prone to depression. Then again, due to the many shortcomings of the
research design, these results may not be valid for the general population and are to be taken
with caution. What can be said with certainty is that people with ASD can only benefit from
programs that engage them in more meaningful ways to spend their time and encourage them
References
Billstedt, E., Gillberg, I. C., & Gillberg, C. (2010). Aspects of quality of life in adults diagnosed with autism in childhood: A population-based study. Autism, 1362361309346066.
Borsboom, D., Cramer, A. O., Schmittmann, V. D., Epskamp, S., & Waldorp, L. J. (2011). The small world of psychopathology. PloS one, 6(11), e27407.
Borsboom, D., & Cramer, A.O. (2013). Network analysis: an integrative approach to the structure of psychopathology. Annual review of clinical psychology, 9, 91-121.
Burgess, A. F., & Gutstein, S. E. (2007). Quality of life for people with autism: Raising the standard for evaluating successful outcomes. Child and Adolescent Mental Health,
12(2), 80-86.
Cramer, A.O.J., Borsboom, D., Aggen, S.H., & Kendler, K.S. (2011). The pathoplasticity of dysphoric episodes: Differential impact of stressful life events on the pattern of depressive symptom inter-correlations. Psychological Medicine.
doi:10.1017/S003329171100211X
Cramer, V., Torgersen, S., & Kringlen, E. (2010). Mood disorders and quality of life. A community study. Nordic journal of psychiatry, 64(1), 58-62.
Cramer, A. O., Waldorp, L. J., van der Maas, H. L., & Borsboom, D. (2010). Comorbidity: A network perspective. Behavioral and Brain Sciences, 33(2-3), 137-150.
Costantini, G., Epskamp, S., Borsboom, D., Perugini, M., Mõttus, R., Waldorp, L.J., Cramer, A.O.J., State of the aRt personality research: A tutorial on network analysis of
personality data in R, Journal of Research in Personality (2014), doi: http://dx.doi.org/10.1016/j.jrp.2014.07.003
Epskamp, S., Cramer, A. O., Waldorp, L. J., Schmittmann, V. D., & Borsboom, D. (2012). Qgraph: Network visualizations of relationships in psychometric data. Journal of
Ghaziuddin, M., Ghaziuddin, N., & Greden, J. (2002). Depression in persons with autism: Implications for research and clinical care. Journal of autism and developmental
disorders, 32(4), 299-306.
Hofvander, B., Delorme, R., Chaste, P., Nydén, A., Wentz, E., Ståhlberg, O., Herbrecht, E., Stopin, A., Anckarsäter, H., Gillberg, C., Råstam, M., & Leboyer, M. (2009). Psychiatric and psychosocial problems in adults with normal-intelligence autism spectrum disorders. BMC psychiatry, 9(1), 35.
Kossakowski et al. (in prep.). The Application of a Network Approach to Health-Related Quality of Life (HRQoL): Introducing a New Method for Assessing HRQoL in Healthy Adults and Cancer Patient (not published yet).
Kroenke, K., Spitzer, R. L., & Williams, J. B. (2001). The Phq ‐9. Journal of general internal medicine, 16(9), 606-613.
Luppa, M., Heinrich, S., Angermeyer, M. C., König, H. H., & Riedel-Heller, S. G. (2007). Cost-of-illness studies of depression: a systematic review. Journal of affective
disorders, 98(1), 29-43.
Matson, J. L., & Nebel-Schwalm, M. S. (2007). Comorbid psychopathology with autism spectrum disorder in children: An overview. Research in developmental disabilities,
28(4), 341-352.
McNally, R. J., Robinaugh, D. J., Wu, G. W., Wang, L., Deserno, M. K., & Borsboom, D. (2014). Mental Disorders as Causal Systems A Network Approach to Posttraumatic Stress Disorder. Clinical Psychological Science, 2167702614553230.
Saldaña, D., Álvarez, R. M., Lobatón, S., Lopez, A. M., Moreno, M., & Rojano, M. (2009). Objective and subjective quality of life in adults with autism spectrum disorders in southern Spain. Autism, 13(3), 303-316.
Schmittmann, V. D., Cramer, A. O., Waldorp, L. J., Epskamp, S., Kievit, R. A., & Borsboom, D. (2013). Deconstructing the construct: A network perspective on psychological phenomena. New ideas in psychology, 31(1), 43-53.
Stewart, M. E., Barnard, L., Pearson, J., Hasan, R., & O’Brien, G. (2006). Presentation of depression in autism and Asperger syndrome A review. Autism,10(1), 103-116.
Trivedi, M. H., Wisniewski, S. R., Nierenberg, A. A., Gaynes, B. N., Warden, D., Morris, M. D. W., ... & Rush, A. J. (2010). Health-related quality of life in depression: a STAR* D report. Annals of Clinical Psychiatry, 22(1), 43-55.
Rapaport, M. H., Clary, C., Fayyad, R., & Endicott, J. (2005). Quality-of-life impairment in depressive and anxiety disorders. American Journal of Psychiatry, 162(6), 1171-1178.
Renty, J. O., & Roeyers, H. (2006). Quality of life in high-functioning adults with autism spectrum disorder The predictive value of disability and support characteristics.
Autism, 10(5), 511-524.
Simonoff, E., Pickles, A., Charman, T., Chandler, S., Loucas, T., & Baird, G. (2008). Psychiatric disorders in children with autism spectrum disorders: prevalence, comorbidity, and associated factors in a population-derived sample. Journal of the
American Academy of Child & Adolescent Psychiatry, 47(8), 921-929.
Van de Leemput, I. A., Wichers, M., Cramer, A. O. J., Borsboom, D., Tuerlinckx, F.,
Kuppens, P., Van Nes, E. H., Viechtbauer, W., Giltay, E. J., Aggen S. H., Derom, C., Jacobs, N., Kendler, K. S., Van der Maas H. L. J., Neale, M. C., Peeters, F., Thiery, E., Zachar, P., & Scheffer, M. (2013). Critical slowing down as early warning for the onset and termination of depression. Proceedings of the National Academy of
Sciences, doi: 10.1073/pnas.131211411
Van Heijst, B. F., & Geurts, H. M. (2014). Quality of life in autism across the lifespan: A meta-analysis. Autism, 1362361313517053.
Wennink, J., & Wijngaarden, B. van (2004). Quality of Life and Care (QoLC). Kwaliteit van leven en vervulling zorgwensen. Utrecht: Trimbos-instituut.
White, S. W., & Roberson-Nay, R. (2009). Anxiety, social deficits, and loneliness in youth with autism spectrum disorders. Journal of autism and developmental disorders,
Appendix
Figure A1. Correlation network of the ASD group. Positive associations are displayed in
Figure A2. Correlation network of the PtD group. Positive associations are displayed in green
Figure A3. Correlation network of the control group. Positive associations are displayed in
Figure A4. Partial correlation network of the ASD group. Positive associations are displayed
Figure A5. Partial correlation network of the PtD group. Positive associations are displayed
Figure A6. Partial correlation network of the control group. Positive associations are
Figure A7. Adaptive LASSO network of the ASD group. Positive associations are displayed
Figure A8. Adaptive LASSO network of the PtD group. Positive associations are displayed in
Figure A9. Adaptive LASSO network of the control group. Positive associations are
Figure A10. Measures of node centrality for the partial correlation (displayed in blue) and
Figure A11. Measures of node centrality for the partial correlation (displayed in blue) and
Figure A12. Measures of node centrality for the partial correlation (displayed in blue) and
