No Health without Mental Health: European Clinical Psychology Takes Responsibility
To search for a specific ID please enter the hash sign followed by the ID number (e.g. #123).

Predictions/dynamics of mental health symptoms

Session chair: Beierl, Esther, Dr.
Shortcut: S15
Date: Friday, 1 November, 2019, 11:45 a.m.
Room: E03
Session type: Symposia


11:45 a.m. S15-01

Assessing the impact of affective dynamics on the reporting of somatic symptoms using a smartphone app (#77)

D. Gosar1, L. Kopač2, B. Kovač2, S. Seršen2

1 University Medical Center Ljubljana, Department of Child, Adolescent and Developmental Neurology, Ljubljana, Slovenia
2 University of Ljubljana, Faculty of Arts, Department of Psychology, Ljubljana, Slovenia

Structured Abstract

Introduction: Studies have long indicated that negative affect influences the reporting of somatic symptoms in health care settings and the normal population. Using a smartphone app we set out to gather intensive longitudinal data to examine the dynamics behind this relationship. 

Methods: Our study included 34 participants (age range = 19 to 25 years; 85% female) who reported on their mood and somatic symptoms twice a day for three weeks. They repeatedly filled out a smartphone questionnaire, which included 10 items from the Positive and Negative Affect Schedule (PANAS) and 8 items from the Somatic Symptom Scales (SSS-8). At the onset of the study they also completed the Personality Assessment Inventory (PAI) to assess their symptoms of conversion and somatoform disorders, anxiety and other forms of psychopathology that might impact the dynamics between affect and somatic symptoms. By using continuous time structural equation modeling we calculated measures of autocorrelation (AR) and cross-correlation (CR) among the constructs of positive affect (PA), negative affect (NA) and somatic symptoms.

Results: Our general structural model indicated that the main driver of somatic symptom reporting after 12 hours was the inertia of previous somatic symptoms (AR=.31) and previous expression of NA (CR=.23). The impact of NA on somatic symptom reporting lasted up to a period of 36 hours (CR=.10). A reverse effect was also present, with somatic symptoms prospectively predicting NA, but to a lesser degree (CR=.09). These dynamics were impacted by symptoms of conversion disorder, anxiety and depression. PA prospectively predicted less somatic symptom reporting in participants with greater physiological symptoms of anxiety, while NA predicted less somatic symptoms in participants with symptoms of conversion disorder.

Conclusions: The dynamics between NA and somatic symptoms revealed by our study highlight the potential smartphones have for research in clinical psychology.

Keywords: Smartphones, Ecological Momentary Assessment, Somatic Symptoms, Affect
12:00 p.m. S15-02

Where machine learning meets mental health: development of an efficient tool for early assessment of individuals at risk for PTSD (#69)

E. T. Beierl1, I. Böllinghaus2, D. M. Clark1, 2, 3, E. Glucksman4, A. Ehlers1, 2, 3

1 University of Oxford, Oxford, United Kingdom
2 King's College London, London, United Kingdom
3 Oxford Health NHS Foundation Trust, Oxford, United Kingdom
4 King's College Hospital NHS Foundation Trust, London, United Kingdom

Structured Abstract

Introduction. The identification of risk factors for the development of posttraumatic stress disorder (PTSD) allows targeted early treatment and prevention. Previous research has identified three broad categories of psychological risk factors: a) pre-trauma factors, such as personal or family history of psychopathology, b) factors operating during the trauma, such as dissociation and emotional responses, and c) post-trauma factors, such as lack of social support and coping. The large number and variety of risk factors demonstrate the need for an efficient early assessment of individuals at risk for PTSD. Methods. A large prospective study was conducted, for which participants were recruited from the Emergency Department of a metropolitan hospital following treatment for injuries caused by an assault or a road traffic accident. Of 1291 people who were interested and suitable for the study, 828 provided data, and for 740 participants PTSD diagnoses at 1 month after the traumatic event were available. Supervised machine learning algorithms for the prediction of PTSD diagnosis at 1 month after the traumatic event by 60 well-established risk factors were developed, trained, and validated. Results. Our algorithms performed very well regarding the prediction of PTSD at 1 month and a subset of the most predictive risk factors for PTSD at 1 month could be identified. Conclusions. We propose an efficient tool for the early assessment of individuals at risk for the development of PTSD after a traumatic event consisting of the most predictive risk factors identified by our algorithms. Implications include applying our algorithms for the prediction of a diagnosis of depression after a traumatic event. Future studies could apply our algorithms for the prediction of other diagnoses by risk factors for other disorders.

Keywords: posttraumatic stress disorder, machine learning, risk factors, early assessment
12:15 p.m. S15-03

Using computational models and smartphones to describe mood dynamics and mood impact on decisions (#85)

B. Blain1

1 UCL -Max Planck Centre, Institute of Neurology, London, United Kingdom

Unstructured Abstract

The happiness or mood of individuals is an important metric for societies, but we know little about how the cumulative influence of daily life events are aggregated into mood or how mood impacts decision-making. Previous work using computational modelling combined with smartphone data shows that momentary mood in a risky decision-making task reflects the combined influence of past expectations and reward prediction errors. Here, I first show in a learning task that mood fluctuations do not adapt flexibly to environmental statistics unlike behaviour: mood parameters are linked to stable traits unrelated to environmental statistics. Then, I show that a computational mood parameter in a risky decision-making task correlates with depression scores. Eventually, I show that bad mood results in short-sighted decisions. Computational models combined with data collection with smartphones provides a way to describe how mood fluctuates in the general population and in mood disorders, and how mood affects choices.

Keywords: Decision-making, Computational modelling, major depression, Mood
12:30 p.m. S15-04

Which Features of Emotion Dynamics Contribute to Paranoid Thoughts? An Experience-Sampling Study (#94)

U. Nowak1, T. Lincoln1

1 Universität Hamburg, Clinical Psychology and Psychotherapy, Hamburg, Germany

Structured Poster Abstract

Introduction: After decades of psychological research, we can now say with some certainty that negative affect is implicated in the aetiology of psychotic experiences. Whilst most studies focused on the intensity of negative affect either at specific moments or averaged across time periods, recent investigations suggest that psychotic symptoms are also influenced by the dynamics with which negative affect fluctuates over time. To extend this line of research, this study investigated whether and which features of emotion dynamics contribute to paranoid thoughts.

Methods: A population sample of N=141 participants followed a 6-day experience-sampling procedure and provided momentary reports on negative affect and paranoid thoughts 9 times per day. From negative affect scores we calculated their variability (SD), their instability (MSSD), and their inertia (person-specific autoregressive coefficients) to capture different aspects of how negative affect dynamically fluctuated during the experience-sampling period.

Results: Nonparametric correlations indicated that variability of negative affect (ρ = .47, p < .001) and instability (ρ = .48, p < .001) but not inertia (ρ = .14, p = .089) were associated with the number of paranoid thoughts. We then conducted Poisson regression analyses to account for overlap between dynamic markers. Variability and instability remained significant predictors of paranoid thoughts after controlling for inertia. However, when instability and variability were entered as simultaneous predictors, only variability remained significant.

Conclusions: Future research needs to examine whether similar differential associations between emotion dynamics and paranoia emerge in clinical samples and whether these are specific to paranoia. Emotion regulation strategies should then be identified that protect individuals against risk patterns of emotion dynamics and the resulting insights should be incorporated into emotion-focused interventions for psychosis.

Keywords: Paranoia, Emotion dynamics, Experience-sampling