How current and past anxiety disorders affect daily life in adolescents and young adults from the general population—An epidemiological study with ecological momentary assessment

Prior research indicated, based on retrospective assessments of symptomatology, that 25% of individuals with “remitted” anxiety disorders (AD) experience a relapse. The present study used ecological momentary assessment (EMA) to examine how ADs affect everyday life among community adolescents and young adults with current or remitted AD compared to healthy controls and to each other.


| INTRODUCTION
Anxiety disorders (AD) are the most common mental disorders in childhood and adolescence (Kessler et al., 2012), which are the core risk phases for developing AD (Beesdo et al., 2009;Kessler et al., 2012;Rapee et al., 2009). In addition to the burden of AD per se, frequently developing comorbidity negatively impacts symptomatology and functioning (Hofmeijer-Sevink et al., 2012;Kroenke et al., 2007). Although about two thirds of participants with AD remit after 2 years, recurrence occurs in about one fifth (Penninx et al., 2011;Scholten et al., 2013). After remission, some individuals reported functioning level comparable to healthy individuals, in turn, some reported ongoing impairment similar to chronic AD (Iancu et al., 2014). Therefore, besides investigating present ADs, to improve our understanding of "remitted" ADs is important: are these individuals fully recovered or does symptomatology or other impairments remain, which may put them at increased risk to relapse or develop other forms of psychopathology?
Traditionally, studies on ADs used retrospective information based on interviews and questionnaires (Beesdo et al., 2009).
Although these measures are established with demonstrated validity and reliability, they depict only a snapshot measured once in time or with larger time-intervals between assessments.
Insights into the experiences of individuals with current or past AD during everyday life are largely lacking. Ecological momentary assessment (EMA) is a diary method of gathering real time data on feelings, thoughts, and the context in which they occur (Shiffman et al., 2008). EMA has many advantages, as with its real-time assessment it is less vulnerable to recall bias and its assessment under real-life circumstances improves ecological validity (Myin-Germeys et al., 2009).
Previous work on ADs in children and adolescents in everyday life provides some insights on daily life phenomena. For example, Magallon-Neri et al. (2016) showed that, if a convenience sample of adolescents (n = 101) did report problems associated with mental health symptoms (e.g., feelings of anxiety or lack of concentration), they were closely related to states of sadness and anxiety (Magallon-Neri et al., 2016). Comparing adolescents with and without DSM-IV ADs (each group n = 65), no significant differences in amount or type of momentary affect were found, but anxious adolescents reported greater intensity of nervousness, sadness, and feeling upset than their nonanxious counterparts (Tan et al., 2012). Another EMA-study investigated a convenience sample (n = 1723) of adolescents and defined high-, medium-and low-anxious subgroups (using a tertial split of EMA-anxiety ratings; Henker et al., 2002). They found that highly anxious teenagers reported higher levels of anxiety, stress, anger, sadness, fatigue, as well as lower levels of happiness and well-being than the low-anxious groups (Henker et al., 2002). Overall, empirical studies and findings regarding experience of everyday life in individuals with ADs show a big variety regarding investigated groups, assessment contingency, observations per day, assessment days and constructs (Walz et al., 2014). Prior research focused primarily on (negative) affect. However, other domains may be important. Symptomatology of mental disorders, mental attitudes (e.g., optimism) or reactions to environmental conditions (e.g., stress) have been linked to AD previously, but have not been examined in real life. For example, ADs and depressive disorder show high comorbidity (Kessler et al., 2005) and are both associated with anger (Hawkins & Cougle, 2011). Additional, also excessive mood has been linked to anxiety previously (French et al., 1996). As ADs are marked by inappropriate stress reactions in response to feared but mostly not dangerous stimuli, it is not surprising that participants with AD score higher on a stress scale compared to nonclinical controls (Antony et al., 1998). Besides heightened stress levels, AD has previously been linked to optimism and pessimism (Räikkönen et al., 1999), as well as with low perceived self-efficacy (Bandura, 1988;Muris, 2002). Also, individuals with AD reported impairment in initiating and maintaining sleep (Papadimitriou & Linkowski, 2005), already present during childhood and adolescence (Alfano et al., 2006). Findings of anxious-reporting more negative and less positive thoughts compared to nonanxious children (Hogendoorn et al., 2012) highlight the role of cognition in AD symptomatology. Considering all these findings, it is not surprising that a meta-analysis revealed AD patients, independently of type of AD, to report poorer quality of life compared to healthy controls (Olatunji et al., 2007).
In summary, a range of altered affects and experiences may explain the impairment of individuals with current or past AD, beyond AD symptoms. Examining different daily life domains could therefore help to identify further targets for intervention to improve remission and to prevent relapse. Hence, the present study aims to use an EMA approach in a general population sample of adolescents and young adults considering a range of health-related domains to answer the question, how adolescents and young adults with current or past ADs experience everyday life compared to healthy peers. We assume, that adolescents and young adults with current AD show significant impairment compared to healthy controls [HC], but do they also show more impairment compared to the remitted AD group? Do individuals with remitted AD still show impairment compared to healthy controls, which may make them vulnerable for relapse?
Data came from the baseline investigation of the Behavior and Mind Health (BeMIND) study, an epidemiological cohort study of adolescents and young adults (14-21 years) in Dresden, Germany . The purpose of the BeMIND study is to examine developmental trajectories of mental disorders. A random age and sex stratified sample was drawn from the population registry and 1180 adolescents and young adults were assessed comprehensively between 11/2015 and 12/2016 (T0; response rate 21.7%, cooperation rate 43.4%; AAPOR, 2016). All participants provided written SEIDL ET AL.

| 273
informed consent or assent, and all legal guardians of minors also provided written informed consent.
Besides diagnostic, experimental, and biological assessments at two personal appointments in the study center at Technische Universität Dresden, EMA was a central study part (see details below).
The present analysis sample was based on participants who completed at least 50% (<16 assessments) of the EMA assessments, thus providing sufficient and reliable data. Furthermore, for the purpose of this analysis, participants were categorized in one of three mutually exclusive groups: current AD, remitted AD, and healthy controls (HC; see details below). The flowchart of the analysis sample is shown in Figure 1. The BeMIND study protocol was reviewed by the ethics committee of the TU Dresden (EK381102014). For detailed study information, see Beesdo-Baum et al. (2020).  Table S1); (b) current AD (n = 65) comprising participants who fulfilled criteria of panic disorder, generalized AD, social AD, separation AD or agoraphobia within the last 4 weeks; and (c) remitted AD (n = 52) comprising participants with lifetime panic disorder, generalized AD, social AD, separation AD or agoraphobia, but indicating remission (last occurrence of anxiety or avoidance before the last 6 months). For comorbidity analyses, DSM-5 core diagnoses within the last 4 weeks were included (see Table S1).

| Ecological momentary assessment
EMA was administered via smartphone eight times daily on 4 consecutive days (2 week-days, weekend) after the diagnostic interview (within 2 weeks 98.6% of participants). One time-based morning-, 6 day-, and one evening-assessment were administered, sharing most items. Sleep-quality during the last night was additionally assessed in the morning and quality of life and self-efficacy in the eveningassessment. Variables were assessed using a seek-bar which was transformed into a scale of 0-100 (e.g., 0 = "Never"; 100 = "Always"; see Figure 2) and all items were mandatory. Assessing self-efficacy, the option "Was not necessary" was included for two items (See Table A1 in the Appendices) and if chosen no slider rating appeared.
An individual reminder scheme, considering sleep times and periods during which participants did not want to be disturbed (e.g., school times), was created for each participant. Reminders were distributed symmetrically throughout the day but at unknown times for the F I G U R E 1 Flowchart of analysis sample. AD, anxiety disorder; EMA, ecological momentary assessment F I G U R E 2 Example of ecological momentary assessment question and scale participant. Each survey could be postponed for 5 min three times, or could be omitted. Participants were instructed by trained study staff to enhance motivation and a training day with three assessments was included. After the EMA period, participants returned the smartphones to the study center and data stored on the smartphone were transferred to the study server. For a full documentation see Beesdo-Baum et al. (2020). Constructs used in this current analysis included manic symptomatology (Altman Self-Rating Mania Scale; Bräunig et al., 2009) Table A1 in the Appendices.

| Statistical analysis
Sample weights were applied so that the overall BeMIND sample is representative for the population of the 14-21-year-old residents of Dresden with regard to sex and age. For all EMA outcomes, the individual means over all assessments for each participant were calculated (assuming missing at random). These individual means were then used as dependent variables (DVs) in regression models with group (current AD, past AD, HC) as dummy predictor variable. DVs were originally measured on a scale of 0-100, but divided by 100 to transform values to a scale of 0-1. We treated DVs as individual means of fractions and used fractional response regression models (Papke & Wooldridge, 1996) to compare the different groups. Because regression coefficients of fractional response models are hard to interpret, we calculated group predictive margins of the DVs for each group under study. A group predictive margin represents the mean of all regression-predicted DV values over all individuals within a group.
In that way, it yields a best guess for the mean DV value of a group. To depict differences in margins, contrasts of predictive margins are presented and reflect a potential outcome mean contrast (i.e., the contrast of the prediction in DV if the whole sample were healthy vs. if the whole sample had current AD). This method was chosen instead of simply calculating group mean comparisons and, for example, performing a t test because (1) we can add covariates in a regression model, and more importantly, (2) the DVs are modeled directly as what they are, fractions, leading to better estimates of standard errors and therefore more reliable confidence intervals compared to a model where outcomes are assumed to be normally distributed. Significance tests are used as an exploratory helping device to pinpoint the most interesting differences and not to be interpreted as based on rigorous hypotheses. We accept a 5% Type 1 error rate for each single test as a feature of our study in exchange for a lower Type 2 error rate for each single comparison. This approach without alpha correction favors sensitivity over robustness of findings which we consider a characteristic of an exploratory study. To examine everyday life domains as a function of AD, the groups current AD, remitted AD, and HC were compared while age and sex were included as covariates. Additional consistency and reliability analyses of EMA scales can be found in Appendices A2 (including Tables A2.1 and A2.2).

| Sample characteristics and assessment distribution
Sample characteristics and EMA compliance are shown in Table 1  We note again that measures originally assessed on a slider from 0 to 100 were transformed to a 0-1 scale. In general, participants with current AD differed significantly from HC on almost all measures (except for manic symptomatology) and showed the biggest contrasts compared to the other groups (see Table 2). The smallest significant contrast was found for anger symptomatology (contrast: −0.06; con-

| Remitted AD without comorbidity versus HC
To examine whether differences between remitted AD and HC are due to current comorbidity, remitted AD cases without comorbidities were compared to HC. Comorbidities do not seem to explain the above reported differences in everyday life between remitted AD and HCs, as there were still significant differences in non-comorbid remitted AD versus HCs (significant differences on 12 out of 16 constructs; Table 3).
Contrasts were also comparable in size. In the domain of symptomatology, the biggest significant contrast was shown for anger symptomatol-

| DISCUSSION
The aim of this study was to explore everyday experience in adolescents and young adults with current and remitted AD compared to HCs and to each other. Exploratory findings showed that not only current AD but also remitted AD showed impairments in a range of health related constructs when compared to HCs and this was not explained by comorbidities. Further, remitted AD differed from current AD in some but not all investigated constructs of everyday life experience.
The differences between the current AD and the HC group is in line with previous research using retrospective measures, showing associations between anxiety or an AD diagnosis and different constructs, including symptomatology (French et al., 1996;Hawkins & Cougle, 2011;Kessler et al., 2005), attitudes such as optimism (Räikkönen et al., 1999), self-efficacy (Bandura, 1988;Muris, 2002), stress (Antony et al., 1998), quality of life (Olatunji et al., 2007) and sleep (Papadimitriou & Linkowski, 2005). As impairment due to prior AD is still ongoing in individuals with remitted ADs (Iancu et al., 2014), it is not surprising that the remitted AD group also showed elevated levels of symptomatology, experiential avoidance, as well as worse mood, self-efficacy, quality of life and sleep-quality compared to HCs. In addition, on all three mood-scales, in self-efficacy as well as in quality of life contrasts between remitted AD and HCs were bigger than between the two AD groups. While the current AD group differed significantly from HCs on all valence scales, remitted ADs only differed significantly from HCs on the negative scales (stress, negative thoughts, pessimism). These findings indicate that positive scales (positive thoughts and optimism) might be able to distinguish between remitted and current AD participants. To our knowledge, this difference has not previously been investigated and may be a promising target for further research examining potential risk, or also protective factors for stability versus remission.
Given the evidence for continued symptomatology and impairments in everyday life in remitted AD, we tested whether this effect is a mere consequence of existing other mental disorders. The assumption that comorbidities explain differences was not supported, as the group of remitted AD without comorbidities still differed significantly from the HC group. However, remaining subclinical AD could explain these results.
Our findings indicate daily life experience in remitted AD to be more alike to the experience of current AD than HCs. On 10 con- This is important, as ongoing impairments could be risk factors for relapse, explaining the increased risk for a waxing and waning course of anxiety (Wittchen et al., 2000) and the risk to have symptoms of the same or other mental disorders later in life (Beesdo et al., 2009;Kim-Cohen et al., 2003;Woodward & Fergusson, 2001).
Further research should therefore address the question whether symptoms and impairments in everyday life predict relapse of ADs. If this would be the case, practical implications include to not only assure full remission on a symptomatology basis but also to focus more broadly on other domains of functioning and well-being to possibly prevent relapse.

F I G U R E 3 Margins and confidence intervals of all constructs.
On the x-axis the mean seekbar-position of the respective construct in each diagnostic group is depicted (originally assessed as 0-100) on the transformed 0-1 scale. Higher scores, therefore, represent, for example, more anxiety, better mood, higher levels of stress, more thoughts, and so forth Our findings must be viewed in light of some limitations. First, beside the smaller sample size of the AD-comparisons compared to the HC-comparisons, the number of participants within some of the investigated subgroups is quite small. The smallest group consists of 38 participants (remitted AD without comorbidity), which could be considered too small for appropriate analyses. Therefore, the present results have to be interpreted with caution, with regard to generalizability and are needed to be replicated in larger samples. However, the analyzed data comprises assessments over several days, increasing the validity of the findings. Second one has to keep in mind that the HC group was defined as a super-healthy group without any lifetime diagnosis, which may not be considered a typical population reference.
Although this allows to estimate the effect of current and past AD, it does not provide insights into the specificity of findings with regard to other psychopathology. Third, we considered participants with lifetime AD without anxiety or avoidance behavior regarding the respective AD in the past 6 months as "remitted AD." There is no gold standard for defining remission in ADs. Age at last occurrence reports of participants may also have been biased by recall error. However, time based information of the Composite International Diagnostic Interview revealed good ICCs Kessler & Üstün, 2004;Wittchen et al., 1998). Fourth, the present study did not control for inflation of Type 1 error. As the aim was to exploratively examine daily life in the different diagnostic groups, Type 1 error inflation was accepted in return for higher sensitivity to detect existing differences.
However, future studies will be needed to replicate the present findings using confirmatory analyses. Fifth, another source of potential error could be the self-developed items. However, Cronbach's alpha, correlation coefficients and absolute differences of assessments and assessment days are reported, providing information on sufficient reliability. Last, we did not take into account the advantage of EMA to investigate dynamic changes in affect over hours and days and their role in AD. This is a promising approach that should be used in further research to provide a more detailed picture of everyday experience in affected individuals.
Despite these limitations, the study has its strengths. First, the investigated groups were derived from a general population sample of adolescents and young adults, which is expected to not be biased

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

DATA AVAILABILITY STATEMENTS
The data that support the findings of this study are available from the corresponding author upon reasonable request.