Collaborative outcomes study on health and functioning during infection times (COH-FIT): Insights on modifiable and non-modifiable risk and protective factors for wellbeing and mental health during the COVID-19 pandemic from multivariable and network anal

There is no multi-country/multi-language study testing a-priori multivariable associations between non-modifiable/modifiable factors and validated wellbeing/multidimensional mental health outcomes before/during the COVID-19 pandemic. Moreover, studies during COVID-19 pandemic generally do not report on representative/weighted non-probability samples. The Collaborative Outcomes study on Health and Functioning during Infection Times (COH-FIT) is a multi-country/multi-language survey conducting multivariable/LASSO-regularized regression models and network analyses to identify modifiable/non-modifiable factors associated with wellbeing (WHO-5)/composite psychopathology (P-score) change. It enrolled general population-representative/weighted-non-probability samples (26/04/2020-19/06/2022). Participants included 121,066 adults (age=42±15.9 years, females=64 %, representative sample=29 %) WHO-5/P-score worsened (SMD=0.53/SMD=0.74), especially initially during the pandemic. We identified 15 modifiable/nine non-modifiable risk and 13 modifiable/three non-modifiable protective factors for WHO-5, 16 modifiable/11 non-modifiable risk and 10 modifiable/six non-modifiable protective factors for P-score. The 12 shared risk/protective factors with highest centrality (network-analysis) were, for non-modifiable factors, country income, ethnicity, age, gender, education, mental disorder history, COVID-19-related restrictions, urbanicity, physical disorder history, household room numbers and green space, and socioeconomic status. For modifiable factors, we identified medications, learning, internet, pet-ownership, working and religion as coping strategies, plus pre-pandemic levels of stress, fear, TV, social media or reading time, and COVID-19 information. In multivariable models, for WHO-5, additional non-modifiable factors with |B|>1 were income loss, COVID-19 deaths. For modifiable factors we identified pre-pandemic levels of social functioning, hobbies, frustration and loneliness, and social interactions as coping strategy. For P-scores, additional non-modifiable/modifiable factors were income loss, pre-pandemic infection fear, and social interactions as coping strategy. COH-FIT identified vulnerable sub-populations and actionable individual/environmental factors to protect well-being/mental health during crisis times. Results inform public health policies, and clinical practice.

被引量：- 发表：1970

The challenges of using machine learning models in psychiatric research and clinical practice.

To understand the complex nature of heterogeneous psychiatric disorders, scientists and clinicians are required to employ a wide range of clinical, endophenotypic, neuroimaging, genomic, and environmental data to understand the biological mechanisms of psychiatric illness before this knowledge is applied into clinical setting. Machine learning (ML) is an automated process that can detect patterns from large multidimensional datasets and can supersede conventional statistical methods as it can detect both linear and non-linear relationships. Due to this advantage, ML has potential to enhance our understanding, improve diagnosis, prognosis and treatment of psychiatric disorders. The current review provides an in-depth examination of, and offers practical guidance for, the challenges encountered in the application of ML models in psychiatric research and clinical practice. These challenges include the curse of dimensionality, data quality, the 'black box' problem, hyperparameter tuning, external validation, class imbalance, and data representativeness. These challenges are particularly critical in the context of psychiatry as it is expected that researchers will encounter them during the stages of ML model development and deployment. We detail practical solutions and best practices to effectively mitigate the outlined challenges. These recommendations have the potential to improve reliability and interpretability of ML models in psychiatry.

被引量：- 发表：1970

Regarding the meta-analyses of pharmacogenomics-guided antidepressant treatment: A response to Milosavljevic et al.

被引量：- 发表：1970

Response to ensuring safety in cannabinoid prescriptions: A call for critical assessment.

被引量：- 发表：1970

Clinical research in pediatric autism.

被引量：- 发表：1970