2.1 Study design
This is a cross-sectional observational study including a quasi-experimental part (figure 1). The whole study was performed online. The first part of the study contained questionnaires, while the second part recorded emotional reaction to visual stimuli. It was not possible to skip one question or a questionnaire. The current data is part of a larger study examining the effect of physical activity in an alpine environment on mental health, part of which has been published (Ower et al. 2018). Innsbruck is one of few urban spaces located directly within the Alps and thus allows for easy access to the alpine environment. The ethics commission of the Medical University of Innsbruck reviewed and approved the study protocol. After being informed in detail about the study aims and procedures, participants provided informed consent prior to study participation. Study recruitment was conducted over a four-month period in 2016.
Participants
Participants and recruiting are described in Ower et al. 2018, participant numbers vary slightly compared to the previous publication due to missing data in individual participants. In brief, a total of 1029 individuals participated in an open online-only survey. They were recruited via email (mailing lists), social media and classified websites or whilst treated at the Department of Psychiatry, Psychotherapy and Psychosomatics (Division of Psychiatry II/Psychosomatic Medicine) at Innsbruck Medical University at the inpatient or outpatient clinic. We included mainly patients with the diagnosis of somatoform, depressive and anxiety disorders. For the present analysis participants who terminated the questionnaire early i.e. prior to the Self-Assessment Manikin (SAM) ratings (missing data n = 436, Figure 2) were excluded from the study. This high drop-out rate was mainly due to the fact that SAM ratings of emotional analytics were performed as the final phase of the questionnaire and it was not possible to skip questions. Comparison of participants terminating early with those included in the data analysis showed that the former were significantly older (mean age ± standard deviation, 33.5 ± 12.1 years vs 29.7 ± 10.1 years, p<0.001, Mann-Whitney U-Test) and that a larger proportion of them was female (68.4 % vs 61.2%, p=0.017, Chi-square test). Despite statistical significance, the differences in age (effect size d= 0.34) and sex distribution (odds ratio = 1.37) were comparatively small. Furthermore participants that reported implausible values (n =8), screened positively for alcohol abuse only (n =54) or for an eating disorder only (Anorexia nervosa and Bulimia nervosa; n=33) were excluded from the present analysis (Figure 2). In Anorexia nervosa or Bulimia nervosa it is known that high levels of PA are used as tool for losing weight and therefore are an expression of disease. Therefore, these patients were excluded (Bezzina et al. 2019). There were 4% to 13% missing values for individual SAM ratings. The 498 participants included in the present analysis consisted of two groups. Patients screened positively for mental health disorder on the Patient Health Questionnaire (PHQ, n =183). Participants without positive PHQ screening (n =315) formed the control group (=HC).
2.2 Stimuli
Stimuli were alternating 5 neutral pictures (re-staged to official International Affective Picture System (IAPS) pictures (slide no. 6150, 7009, 5661, 5500, 7150)) and 5 alpine stimuli (Figure 3). Neutral pictures displayed figural subjects of daily life (e.g. mug, wall, umbrella). Alpine stimuli displayed alpine environments with individuals performing some sort of physical activity therein (e.g. hiking, biking, skiing). The pictures were presented to all participants in the same order. Two picture stimuli had to be excluded due considerations related to the displayed content (canyon wall in the neutral stimuli and paraglider in the mountains in the alpine stimuli) and their mean ratings for at least one of the analyzed dimensions ranging two standard deviations outside the mean of the other stimuli in the group. Pictures were displayed for 5 seconds before the page with the emotional analytic ratings appeared. Each stimulus could only be observed once (Figure 1).
2.3 Measures
Socio-demographic parameters included information on age, sex, education and marital status. Mental health was assessed using the German version of Patient Health Questionnaire (Gräfe et al. 2004). Additionally, open text fields were provided for entering psychiatric diagnoses. Resilience was measured using the Brief Resilience Score (BRS) (Smith et al. 2008), self-perceived stress using the Perceived Stress Scale (PSS) (Cohen et al. 1983) and Physical activity (PA) using the Global Physical Activity Questionnaire (GPAQ-2) (Bull et al. 2009). PA is calculated using metabolic equivalents of task (METs) as a unit for energy expense. As determined of the World Health Organization we classified PA in moderate and vigorous intensity. We adapted the standard questionnaire to measure PA performed in the alpine environment.
To measure emotional response we used the Self-Assessment Manikin (SAM) 9-point Likert-scale. This scale measures emotional analytics in the three dimensions valence, arousal and dominance (J Lang et al. 2008). The valence scale ranges from a frowning, unhappy (adjectives used in the SAM manual: unhappy, annoyed, unsatisfied, melancholic, despaired, bored; lower values) to a smiling, happy figure (happy, pleased, satisfied, contented, hopeful). The arousal scale displays the lowest value with a calm, eyes-closed figure (relaxed, calm, sluggish, dull, sleepy, unaroused), whilst the highest value is represented by an excited figure (stimulated, excited, frenzied, jittery, wide-awake, aroused). The lowest values in the dominance scale are symbolized by a controlled small figure (controlled, influenced, cared-for, awed, submissive, guided.) whilst highest values are represented by a dominant and oversized figure (controlling, influential, in control, important, dominant, autonomous). After presenting a picture for five seconds participants were asked to rate their emotional reaction in the three dimensions. For alpine stimuli, we added a fourth dimension asking about ones attraction to the situation, labelled motivational direction. The 9 point Likert-scale ranged from “I don’t want to be in this situation” to “I want to be in the situation”.
2.4 Statistical methods
Metric variables were analyzed for normal distribution prior to applying further statistical tests by assessing their skewness and their kurtosis, considering skewness values > 0.5 or < -0.5 (Lehman 1991) and kurtosis values > 1 or <-1 (https://brownmath.com/stat/shape.htm) as deviations from a normal distribution requiring non-parametric testing. To compare emotional reactions between overall neutral and alpine pictures we created a mean score for each category. In each category one picture was excluded due to statistical outliers (paraglide in alpine pictures; red wall in neutral pictures). Mean scores were calculated for each emotional dimension per person if at least three scores were completed. Group comparisons (patients vs. HC) were performed using t-test, Mann-Whitney U-test and Chi-square test, depending on the variable type and distribution. As the two groups differed significantly in their age; education, marital status, and work situation, we also performed analyses of covariance with adjustment for these potential confounders. As the emotional analytic ratings displayed missing values (4% to 13%), we performed an additional analysis where missing ratings were replaced by imputed values. The SPSS Missing Value Analysis procedure with Little’s test for missingness completely at random (MCAR) and imputation by expectation-maximization (EM) was used for this purpose (IBM SPSS manual). The relationship between resilience, self-perceived stress, PA and emotional analytics was investigated on a descriptive level by means of correlation analysis. Spearman rank correlation coefficients were used as most the variables involved showed deviations from a normal distribution.