Open Access

Rebecca Schwaiger, Isabel Maurus, Moritz Lembeck, Irina Papazova, David Greska, Susanne Muenz, Eliska Sykorova, Cristina E. Thieme, Bob O. Vogel, Sebastian Mohnke, Charlotte Huppertz, Astrid Röh, Katriona Keller-Varady, Berend Malchow, Henrik Walter, Bernd Wolfarth, Wolfgang Wölwer, Karsten Henkel, Dusan Hirjak, Andrea Schmitt, Alkomiet Hasan, Andreas Meyer-Lindenberg, Peter Falkai, Lukas Roell

• Exercise interventions are nowadays considered as effective add-on treatments in people with schizophrenia but are usually associated with high dropout rates. Therefore, the present study investigated potential predictors of adherence from a large multicenter study, encompassing two types of exercise training, conducted over a 6-month period with individuals with schizophrenia. First, we examined the role of multiple participants’ characteristics, including levels of functioning, symptom severity, cognitive performance, quality of life, and physical fitness. Second, we used K-means clustering to identify clinical subgroups of participants that potentially exhibited superior adherence. Last, we explored if adherence could be predicted on the individual level using Random Forest, Logistic Regression, and Ridge Regression. We found that individuals with higher levels of functioning at baseline were more likely to adhere to the exercise interventions, while other factors such as symptomExercise interventions are nowadays considered as effective add-on treatments in people with schizophrenia but are usually associated with high dropout rates. Therefore, the present study investigated potential predictors of adherence from a large multicenter study, encompassing two types of exercise training, conducted over a 6-month period with individuals with schizophrenia. First, we examined the role of multiple participants’ characteristics, including levels of functioning, symptom severity, cognitive performance, quality of life, and physical fitness. Second, we used K-means clustering to identify clinical subgroups of participants that potentially exhibited superior adherence. Last, we explored if adherence could be predicted on the individual level using Random Forest, Logistic Regression, and Ridge Regression. We found that individuals with higher levels of functioning at baseline were more likely to adhere to the exercise interventions, while other factors such as symptom severity, cognitive performance, quality of life or physical fitness seemed to be less influential. Accordingly, the high-functioning group with low symptoms exhibited a greater likelihood of adhering to the interventions compared to the severely ill group. Despite incorporating various algorithms, it was not possible to predict adherence at the individual level. These findings add to the understanding of the factors that influence adherence to exercise interventions. They underscore the predictive importance of daily life functioning while indicating a lack of association between symptom severity and adherence. Future research should focus on developing targeted strategies to improve adherence, particularly for people with schizophrenia who suffer from impairments in daily functioning.…