Study Design
In this cross-sectional study, patients of the `Adult Cystic Fibrosis Center, University Hospital Essen - Ruhrlandklinik, Germany´ participated during an outpatient or inpatient stay between January 2019 and March 2020. Each subject answered a personality and HRQoL questionnaire and provided personal demographic information. Clinical data were collected by the attending physicians during the visit/stay.
Subjects
The study included outpatients and inpatients over 18 years of age attending the `Adult Cystic Fibrosis Center, University Hospital Essen – Ruhrlandklinik´, who had a confirmed diagnosis of CF (11). Patients after lung transplantation and patients who could not complete the questionnaires without assistance were excluded.
Measures - Disease Status
The attending physician recorded disease status, including ppFEV1, BMI, genotype, Pseudomonas aeruginosa colonization status, pancreatic function status, and the presence of CF-related diabetes mellitus.
Measures - Personality
Personality was assessed by the revised `Persönlichkeits-Stil- und Störungs-Inventar´ (PSSI), a self-report questionnaire for adolescents 14 years and older and adults (12). The questionnaire contains 140 items related to 14 scales (see Additional file 1 - Table 1). Answers are provided on a four-point Likert scale within the poles `strongly disagree´ to `strongly agree´. The questionnaire conceptualizes personality traits as non-pathological equivalents to personality disorders from DSM-IV and ICD-10. For any clinical disorder category, there is an analogous personality style, which can be described by interindividual differences on a corresponding quantitatively graded dimension. However, a personality trait recorded in its extreme expression represents a potential risk for a specific personality disorder. The questionnaire shows good psychometric properties.
Measures - Health-Related Quality of Life
The revised `Cystic Fibrosis Questionnaire for adolescents and adults over 14 years old (CFQ-R 14+)´ was used to assess HRQoL. This CF-specific, self-report questionnaire consists of 50 items on 12 scales and includes generic and disease-specific dimensions as well as a scale for the evaluation of subjective health. Each domain is standardized on a 0-100 scale with higher scores indicating better HRQoL. The German version of the CFQ-R 14+ shows good psychometric properties (13).
Measures - Questionnaire on CS, Therapy Adherence, and Demographic Variables
To gain information about cough suppression (CS), therapy adherence, and demographic variables, participants completed a self-report questionnaire developed by the research team (5). The survey includes 12 items that measure the participants’ assessment of future health perspectives (one item), cough frequency, and sputum quantity (two items), CS (one item), therapy adherence (three items), and diagnosis disclosure (four items, Additional file 1 - Table 2). In addition, the participants were asked to specify their educational, employment, and marital status.
Statistical Analyses
Data handling and statistical analyses were performed with SPSS 27.0 (IBM Corp., Armonk, NY). All analyses were corrected for multiple testing controlling the two-tailed false-discovery-rate (FDR) at q < .05 (14), except for exploratory analyses comparing PSSI subscale means between personality clusters and those analyses investigating the relationship between personality clusters and CFQ-R subscale scores. Effect size calculations (d, semipartial correlation (sr2), partial 2 (p2),odds ratio (OR)) relied on SPSS. For ease of interpretation, effect size measures were converted to Cohen’s d (small 0.1 ≤ r < 0.3, medium 0.3 ≤ r < 0.5, large r ≥ 0.5 ) using an online calculator (d) (15).
Details on testing the assumptions of all outlined statistical procedures are provided in the Additional Material.
Single Scale Analysis
The central tendency of each PSSI subscale was tested against a mean of 50 by one-sample t-tests as the PSSI subscale scores follow a t-distribution with a mean of 50 and a standard deviation (SD) of 10. In the presence of outliers (values exceeding ± 2.5 the median absolute difference (16)), one-sample Wilcoxon tests were performed to compare subscale scores to a median of 50.
We used detailed information regarding the distribution of PSSI subscale t-scores from the manual to separately compare the frequency of scores below and above the normal range of PSSI scores (40 – 60) between the norming sample (N = 1.943, male = 1.056) and the sample of pwCF by 2-tests of independence. All analyses were checked for sufficient cell size.
The linear relationship between ppFEV1, BMI, and CFQ-R overall score as well as therapy adherence (dependent variables) with the PSSI subscale t-scores (independent variables) was assessed by multiple regression considering an analysis-specific subset of covariates with significant correlations regarding the dependent variable of interest (please see the Additional Material). Considering the scale of measure of CS (dependent variable), its relationship with the PSSI subscales (independent variables) was determined by ordinal logistic regression relying on a logit function, likewise accounting for important covariates. Please note, that prior to analysis, all categorical variables were dummy coded (please see the Additional Material).
Cluster Analysis
To identify certain personality styles among pwCF irrespective of differences compared to the norming sample, the PSSI subscale t-scores were subjected to a two-step cluster analysis using a log-likelihood distance measure for clustering and the Akaike Information Criterion (AIC) to determine the maximum number of clusters with a preset of no more than 10 clusters.
The resulting clusters (independent variable) were compared regarding ppFEV1, BMI, CFQ-R overall score, and therapy adherence (dependent variables) by an analysis of covariance (ANCOVA) considering the same analysis-specific subset of covariates as detailed above. The relationship between personality clusters and CS was evaluated by logistic regression.