Fig. 5.

Physical and mental health, and quality of life in the COVID-19 recovery clusters. Clustering of the survey study participants in respect to symptom-specific recovery times was done by the semi-supervised PAM algorithm (partitioning around medoids, Euclidean distance, training cohort: Austria [AT], test cohort: Italy [IT]). Minimum/maximum scaled readouts of clinical and physical recovery, mental health and quality of life at the time of survey completion in the clusters in the Austria (AT) and Italy (IT) survey study cohorts are presented. Dichotomous items (incomplete convalescence, weight loss, new medication and need for rehabilitation) were binarized (yes: 1, no: 0) prior to visualization. Statistical significance for differences between the clusters was assessed by Kruskal–Wallis with ${\mathrm{\eta }}^2$ effect size statistic (numeric variables) or ${\mathrm{\chi }}^2$ test with Cramer V effect size statistic (categorical variables). P values were corrected for multiple testing with Benjamini–Hochberg method. Lines represent mean values, 2 $\times$ SEM intervals are displayed as tinted regions. Effect sizes and p values are shown in the plots. Numbers of individuals assigned to the recovery clusters are indicated in the plot legends. Incomplete recovery: self-reported incomplete recovery; # persist. symptoms: number of symptoms at 28 days after clinical onset; phys. Performance loss: physical performance loss as compared with the time before COVID-19; QoL impairment score: score of impaired quality of life; OMH impairment score: overall mental health impairment score; ANX score anxiety score, Patient Health Questionnaire, PHQ-4; DPR depression score, Patient Health Questionnaire, PHQ-4; stress score: mental stress score, 7 item PHQ stress module