Anxiety symptoms in adolescents at risk for psychosis: a comparison among help seekers
Several studies have reported on how anxiety disorders and anxiety symptoms are already present before the onset of psychosis. However, anxiety disorders are typically studied in these studies at diagnosis-level. The aim of present study was to investigate the profile of anxiety symptoms in subjects at risk of developing psychosis and to compare the anxiety profile with those who are not at risk.
Data were collected at Helsinki University Central Hospital (HUCH) by an early detection and intervention team. Of 185 help-seeking respondents, between 12 and 18 years of age, 59 adolescents were classified as being at risk of psychosis and 126 as not being at risk via an interview conducted by a validated at-risk assessment tool (PROD). Anxiety was measured using the Beck Anxiety Inventory (BAI).
The anxiety total sum score was higher in the at-risk group for psychosis (mean 8.33 vs. 13.34, p = .000). Both subfactors of the anxiety scale, cognitive anxiety (p = .000) and somatic anxiety (p = .000), differed significantly by risk status. After using the Bonferroni correction for multiple analysis, items of relax (p = .000), nervous (p = .002), losing control (p = .000) and faint (p = .002) had statistically significant higher mean scores in the group at risk of psychosis. In logistic regression analysis, being female (p = .015) and the subfactor relating to cognitive anxiety (p = .044) significantly explained the at-risk status for psychosis.
Adolescents at risk for psychosis have a higher level of anxiety compared with other help-seeking adolescents. These results should be considered in clinical practice.
Key Practitioner Message
- Anxiety has previously been related to psychosis and an at-risk state for psychosis. In the present study, help-seeking adolescents differ significantly in terms of their level of total anxiety symptoms, because adolescents at risk for psychosis had more anxiety symptoms
- Item-by-item analysis scores in items such as feeling relaxed, feeling nervous, losing control and feeling faint were statistically significantly higher in adolescents at risk for psychosis
- The anxiety subfactor of cognitive anxiety explained significantly at-risk state for psychosis
- In clinical practice, the possibility of persisting anxiety symptoms should be considered when working with help-seeking adolescents at risk state for psychosis
Anxiety disorders have been related to schizophrenia (Achim et al., 2011; Yun et al., 2011), and previous evidence shows that some unique aspects of anxiety are related to schizophrenia symptoms (Norman, Malla, Corteze & Diaz, 1998). However, several studies have even reported how anxiety disorders and anxiety symptoms are present already before the onset of psychosis (Addington et al., 2011; Marshall et al., 2011; Rietdijk et al., 2011; Svirskis et al., 2005).
The period before the onset of the first episode psychosis, usually called the at-risk state for psychosis or Ultra High Risk (UHR), is based on the statistical risk of making the transition to psychosis with specific psychosis-risk symptoms. The risk state itself is not a diagnosable psychiatric condition, even though there has been a recent discussion of it. These symptoms are mainly assessed using a specific instrument. The SIPS/SOPS (Structured Interview of Prodromal Symptoms/Scale Of Prodromal Symptoms) SIPS/SOPS interview has been reported to have a predictive validity of between 27% and 54%; previous studies have found that subjects defined as being in an at-risk state have developed psychosis within 12 months (Lemos et al., 2006; Miller et al., 2002, 2003). More recent studies show a tendency for lower transition rates because of earlier referral, effective treatment strategies and the inclusion of a larger number of false-positive subjects (Yung et al., 2008). When using the CAARMS (Comprehensive Assessment of At Risk of Mental State) method, Yung et al. (2008) found a 16% transition rate within a 2-year follow-up period, whereas Ruhrmann et al. (2010) found a 19% transition rate within an 18-month follow-up period when using a European version of the UHR criteria that combined SIPS/SOPS and basic symptoms of schizophrenia. Basic symptoms include subjectively experienced disturbances in perception, cognition, language, motor function, initiative, stress tolerance and level of energy (Gross, Huber, Klosterkötter & Linz, 1987). Furthermore, Ziermans, Schothorst, Sprong and van Engeland (2011) reported a transition rate of 15.6% based on SIPS/SOPS criteria and the basic symptoms observed within a 2-year follow-up period, whereas Haroun, Dunn, Haroun and Cadenhead (2006) reported a transition rate of 13%, based on SIPS/SOPS criteria, within a 1-year follow-up period. Finally, Skeate, Patterson and Birchwood (2004) reported a transition rate of 6.6%, based on UHR criteria and basic symptoms, within a 6-month follow-up period.
Two previous studies have used the BAI to assess anxiety in patients suffering from schizophrenia. Yun et al. (2011) discovered that the total BAI scores had a mean of 18.50 for the schizophrenia group, a mean of 13.42 for the remitted group and a mean of 5.55 for healthy controls. Norman et al. (1998) reported a mean BAI score of 11.85 in a sample of schizophrenia patients. They found associations between the psychosis factor of reality distortion and the BAI items of numbness, heart pounding, feeling terrified, difficulty breathing and feeling faint. Studies on anxiety among those in an at-risk state for developing psychosis are for the most part lacking. Svirskis et al. (2005) found that anxiety disorders defined by a SCID-I diagnosis were common among subjects vulnerable to psychosis; in their study, 26 of 39 subjects vulnerable to psychosis had at least one lifetime anxiety and/or a somatisation disorder diagnosis. Rietdijk et al. (2011) reported that mood and anxiety disorders were present in 39.1% of those cases treated in secondary mental health care before the onset of psychosis. Finally, in intervention studies of those in an at-risk state for psychosis, there is a tendency for the level of anxiety to reduce during treatment (Addington et al., 2011; Granö et al., 2009; Marshall et al., 2011), suggesting that different anxiety symptoms are present when in an at-risk state for psychosis.
Previous studies have shown a link between anxiety disorders and at-risk status for psychosis. However, anxiety disorders are typically studied in these studies at diagnosis-level. Hence, a more detailed picture of various anxiety symptoms when in an at-risk state for psychosis is lacking. As a hypothesis, we assumed that increasing psychosis risk is characterised by increasing levels of subclinical anxiety symptoms. The aim of the present study was to investigate anxiety symptoms in subjects in an at-risk state for psychosis and to compare the anxiety profile with those who are not at risk for psychosis.
The Jorvi Early psychosis Recognition and Intervention (JERI) project (2009–2011) at Helsinki University Central Hospital (HUCH) was designed to be an early detection, intervention and research initiative for treating adolescents at risk of developing first-episode psychosis. The catchment area consisted of five towns and had a total population of half a million inhabitants. The JERI team was a multiprofessional team of psychiatric nurses, occupational therapists, psychologists and a supervising psychiatrist. The JERI team met adolescents between 12 and 22 years of age in their community settings, e.g. at school or at home, together with their parents and a community co-worker who had originally contacted the JERI team because of unclear and undiagnosed mental health problems. The community co-workers were instructed to contact a JERI team about adolescents who had become their clients because of recently occurred mental health problems. The mental health problems included a wide range of possible at-risk symptoms for psychosis, such as a depressive mood, anxiety symptoms, sleeping difficulties, withdrawing from social contact, staying at home instead of going to school, problems in attention or concentration at school, odd behaviour, hearing or seeing things that others do not see or hear. The community co-workers participated in regular seminars on early mental health problems and how to work in co-operation with the JERI team. Additionally, the community co-workers received flyers about the risk symptoms and how to contact the JERI team if such symptoms should occur. The main exclusion criterion was that subjects could not receive other psychiatric care at the same time as the JERI intervention. Hence, the subjects studied by the JERI-team included a broad range of adolescents who had subthreshold psychiatric symptoms; although the symptoms had been detected, the adolescents had not been remitted to psychiatric care as a first choice. The team had three main tasks: first, to identify the possible risk of developing psychosis among help-seeking adolescents using self-report screening tools and an interview; second, in such cases, to meet frequently with the client and family together with the community co-worker to find a way to reduce stress and support the client in his or her overall functioning at school or at work; and third, if some other severe psychiatric disorder or untreated psychosis should emerge, to guide the subject to adequate psychiatric care. The detection and intervention model was developed originally at Jorvi Hospital in HUCH between the years 2006 and 2008 (Granö et al., 2009).
The team used a Finnish PROD screen (Heinimaa et al., 2003) as an instrument to distinguish adolescents at risk for psychosis from those who were not at risk for psychosis in the same way as suggested by Heinimaa et al. (2003). All of the subjects were interviewed using the screen, which examined their symptoms during the previous 12 months. Each respondent first completed the PROD screen as a questionnaire; after that, each respondent was additionally interviewed by a team worker to assess whether each item was a real positive risk item. The final scores were based on the interviewers' scores, and the cut-off point was 2/12 positive specific-risk items. However, this means in practice that on a continuum of psychosis, two or more specific risk symptoms give a relative reliable risk status to be in a ‘higher risk’ group and people who scored just one specific risk symptom or are without any specific risk symptoms are in a ‘lower risk’ group. The subjects were assessed at baseline between 1 April 2009 and 31 October 2011. The focus was on an adolescent population and therefore subjects were selected only from secondary school age. We excluded from the study adolescents who were between 19 and 22 years of age. Of a total 528 telephone consultations, the JERI team met 185 adolescents who were selected from adolescent age range referrals and asked them to fill in questionnaires. Of those respondents between 12 and 18 years of age, 59 were assigned to an at-risk status psychosis group and 126 respondents were classified as other help seekers. The present study was accepted by the ethics committee of HUCH and voluntary participation in the study was emphasised. All of the participants gave their informed consent prior to the study.
The PROD screen (a 21-item dichotomy screen in which follow-up questions are asked if a positive answer is received for any of the 12, specific, psychosis-risk item questions) is based on the questions in SIPS (Structural Interview for Prodromal Symptoms), IRAOS (Interview for the Retrospective Assessment of the Onset of Schizophrenia) and BSABS (Bonn Scale for the Assessment of Basic Symptoms) screens (Gross et al., 1987; Häfner et al., 1992; McGlashan, 1998). Specific risk items include, for example, difficulties in thinking clearly or in concentrating, interfering thoughts or interrupted thoughts, difficulties in understanding written text or speech and disorders connected with hearing odd sounds or voices without there being an obvious source (Heinimaa et al., 2003). The PROD screen has been standardised by comparing the results with those from the SIPS interview (McGlashan, 1998); the PROD screen had a 75% specificity and an 80% sensitivity with a cut-off point of two or more specific risk items in an epidemiologically mixed sample (Heinimaa et al., 2003). This was the threshold used in the present study. Adolescents who scored 0 or 1 points in the PROD screen were assigned to the ‘not at risk’ group, whereas those who scored 2 points or more were to the ‘at risk’ group. The SIPS interview has been reported to have a predictive validity of between 27% and 54%; SIPS/SOPS-defined, prodromal state subjects developed psychosis within 12 months (Lemos et al., 2006; Miller et al., 2002, 2003). All team members were certified to use the SIPS interview.
The Beck Anxiety Inventory (BAI; Beck & Steer, 1993) was used to assess anxiety. The BAI is a 21-item, self-report instrument that is rated on a four-point scale from 0 (not at all) to 3 (severe). The BAI has good psychometric properties with high test–retest reliability and good internal consistency (Beck & Steer, 1993). The Finnish version of the inventory was used in the present study.
According to hypothesis I, group differences between at-risk for psychosis group and not-at-risk for psychosis group were analysed by analysis of variance in variables of demographics (age and gender), cognitive anxiety, somatic anxiety, total anxiety and pre-psychotic symptoms. In the post-hoc analysis I, sub-items of BAI were analysed by analysis of variances. The Bonferroni correction for multiple analysis was used in the analyses of variance for item-by-item differences (alpha level .05 divided by the number of 21 sub-items on the BAI scale). In the post-hoc analysis II, logistic regression analysis was used to control for gender, cognitive anxiety and somatic anxiety, as those variables differed by risk-status in analysis of variances. All statistical analyses were performed using the PASW 19.0 (PASW Statistics for Windows, SPSS Inc, Chicago, IL) statistical package.
The at-risk group did not differ from the not-at-risk group in age. The at-risk group had significantly more girls than boys (Table 1).
Table 1. Clinical characteristics and analysis of variance in 185 adolescents at risk of developing psychosis
|Age (years)||14.9 (1.74)||15.0 (1.50)||15.0 (1.67)||.074||.787|
|Girls (N %)||60||44||104||12.54||.001|
|Boys (N %)||66||15||81|| || |
|Anxiety sum score (BAI)||8.33 (7.38)||13.34 (7.60)||9.92 (7.79)||18.21||.000|
|Cognitive anxietyb||4.02 (3.72)||6.56 (3.94)||4.83 (3.96)||17.95||.000|
|Somatic anxietyb||4.30 (4.23)||6.78 (4.53)||5.09 (4.46)||13.20||.000|
|Pre-psychotic symptoms (PROD-screen)||.28 (.45)||4.29 (2.03)||1.56 (2.22)||449.04||.000|
Hypothesis 1: Is increasing psychosis risk characterised by higher anxiety levels?
The at-risk group scored more highly on total anxiety and on both component subscores (cognitive and somatic anxiety; Table 1).
Post-hoc analysis I – sub-items of the BAI
After the Bonferroni correction for multiple analysis, the at-risk group scored more highly on items relating to nervousness, difficulty relaxing, fear of losing control, and feeling faint (Table 2).
Table 2. Analysis of variance of item-by-item differences in Beck Anxiety Inventory (BAI) by risk status of developing psychosis (n = 185)
Post-hoc analysis II – anxiety and demographic predictors of at-risk status
At-risk status was significantly predicted by the cognitive component of anxiety after gender was adjusted for (Table 3).
Table 3. Gender-adjusted logistic regression analysis of subscales of the Beck Anxiety Inventory by being at risk of developing psychosis (n = 185)
Increasing psychosis risk is characterised by higher anxiety levels
In the present study, we found that anxiety levels measured by the BAI were significantly higher among those at risk of developing psychosis. This is in accordance with previous literature (Addington et al., 2011; Marshall et al., 2011; Rietdijk et al., 2011; Svirskis et al., 2005; Yun et al., 2011). An item-by-item analysis of the 21 subcategories revealed that feeling relaxed, feeling nervous, losing control and feeling faint were statistically significantly higher in the at-risk group than in the not-at-risk group. Logistic regression analysis revealed that female gender and cognitive anxiety explained the at-risk status for developing psychosis. As a main finding, we found that increasing psychosis risk was characterised by increasing levels of subclinical anxiety symptoms and particularly, cognitive anxiety.
The main objective of the present study was to identify and specify anxiety in a group of adolescents at risk for psychosis. The findings are interesting. To our knowledge, the present results provide new information about the relationships between various anxiety symptoms and being in an at-risk state for developing psychosis. The BAI questionnaire has not been widely used in studies of those at risk of developing psychosis. Two previous studies have used the BAI to assess anxiety in patients suffering from schizophrenia. Yun et al. (2011) discovered that the total BAI scores had a mean of 18.50 for the schizophrenia group, a mean of 13.42 for the remitted group and a mean of 5.55 for healthy controls. We had a mean BAI score of 13.17 for the at-risk group and a mean BAI score of 8.26 for the not-at-risk group; thus, our results are in line with those of Yun et al. Moreover, Norman et al. (1998) reported a mean BAI score of 11.85 in a sample of schizophrenia patients. They found associations between the psychosis factor of reality distortion and the BAI items of numbness, heart pounding, feeling terrified, difficulty breathing, and feeling faint. Interestingly, all of the sub-items that Norman, Malla, Corteze and Diaz reported as being associated with the psychosis factor of reality distortion are also significantly different in the present study before the Bonferroni correction except for the item of breathing. In sum, the present results show a similar tendency as in these previous studies conducted with schizophrenia patients.
Anxiety and demographic predictors of at-risk status
Logistic regression analysis revealed that being female and the subfactor of cognitive anxiety (p = .044) explained the at-risk status for developing psychosis. For the at-risk group, the gender balance was similar to what Salokangas et al. (2007) found in their study (67% women, 33% men), which was a part of the international EPOS (European Prediction of Psychosis Study) study. Salokangas et al. (2007) also used the PROD screen to identify subjects at risk for psychosis before the final sample was screened using the SIPS interview. Hence, it is not surprising that female gender was significantly associated with at-risk status for psychosis in regression analysis. It might be possible that young girls are more skilled at recognising and verbalising subjective feelings and experiences than boys. Moreover, the subfactor of cognitive anxiety also explained at-risk status for developing psychosis. Interestingly, Morrison (2001) has suggested that both anxiety disorders and psychotic symptoms are based on misinterpreting intrusions into awareness; thus, it might be possible that respondents who score high in both psychosis risk symptoms and items that measure cognitive anxiety tend to misinterpret intrusions from bodily sensations or the environment. Finally, respondents in the at-risk group for psychosis had higher scores for the sub-item of losing control than did those in the not-at-risk group. This might partly be explained by the fact that the item related to losing control and that the issue of having difficulties with reality testing overlap somewhat, which typically is present in the at-risk state for various symptoms, for example, in dimensions P1–P5 in the SIPS interview. This might also be a result of a higher anxiety level.
According to the study, there is up to a 25% chance of obtaining false-positive risk subjects when using the PROD screen (Heinimaa et al., 2003). It is possible that our risk group may have included false-positive risk subjects, including those who would never be at true risk of developing psychosis. This false-positive rate may be the result of the items used in the PROD screen, as several risk items in the screen are based on basic symptoms. Basic symptoms are earlier risk symptoms for psychosis and not the symptoms measured using the SIPS scale; in the long run, basic symptoms predict a new psychosis with a great deal of validity and reliability (Klosterkötter, Hellmich, Steinmeyer & Schultze-Lutter, 2001). Furthermore, some true risk-subjects may also exist in the nonrisk group because the PROD screen has an 80% sensitivity compared with the SIPS method (Heinimaa et al., 2003). The PROD instrument defines at-risk status for later psychosis with a cut-off point of two or more. This cut-off point has been standardised against the SIPS-UHR status. This means that on a continuum of psychosis, two or more specific risk symptoms give a relatively reliable risk status compared with SIPS and, at the same time, people not at risk who scored one specific risk symptom may also, in theory, be at greater risk for later psychosis than adolescents without any specific risk symptoms. As an easy and fast screen, the PROD screen seems to be a useful tool for comparing at-risk subjects with nonrisk subjects, even though the true risk status should always be evaluated using a proper instrument, for example, with SIPS or CAARMS.
It is possible that the selection process for the subjects included in the present study affected the sample and, hence, the results. Our subjects were all help seekers who were not receiving psychiatric care at the time. It is possible that subjects with a help-seeking status are currently experiencing stress, which may cause anxiety. This may cause our subjects to have more anxiety than subjects in psychiatric care, who might have sought care because of other mental health problems. Moreover, unfortunately, we were not able to collect data on possible anxiety disorders among our subjects and we were not able to compare the results with those patients with diagnosed anxiety disorders because the team's objective was to focus on subclinical symptoms. This is a clear limitation of the present study and an important aim for future research. The BAI has not been standardised for adolescents or for subjects at risk for psychosis. This may cause problems in the reliability and validity of the results. As the at-risk state for psychosis itself may include developmental delay and social adversity, and as some early cognitive problems may be present in the at-risk state, it may be possible that the self-reported answers to anxiety questions may be biased in light of the problems related to at-risk status for psychosis.
In summary, the present results suggest that help-seeking adolescents had higher anxiety sum scores if they were at risk for developing psychosis. Adolescents differed in several anxiety symptoms depending on the risk. These results should be considered in clinical practice and in the care of adolescents.
The study was conducted with a clinical sample of patients at Helsinki University Central Hospital, and was not subject to external study funding. The authors have declared that they have no competing or potential conflicts of interest.