Predicting OCD severity from religiosity and personality: A machine learning and neural network approach

Area of Science:

• Psychiatry
• Computational Psychology
• Data Science

Background:

• Obsessive-compulsive disorder (OCD) impacts a substantial segment of the US population, necessitating advanced analytical approaches for comprehensive understanding.
• Traditional methods for assessing OCD severity often rely on aggregate scores, potentially overlooking nuanced individual-level factors.

Purpose of the Study:

• To investigate the predictive power of personality traits, religiosity, and spirituality on OCD severity using advanced machine learning techniques.
• To compare the efficacy of machine learning models against traditional linear regression in predicting OCD severity.
• To explore the influence of item-level features versus aggregate scores in understanding OCD.

Main Methods:

• Employed machine learning and deep learning techniques on a dataset of 229 participants.
• Utilized item-level features and aggregate scores from assessments of OCD severity, personality traits, religiosity, and spirituality.
• Compared predictive accuracy and explanatory power of neural networks and linear regression models, incorporating demographic factors.

Main Results:

• Item-level features demonstrated greater influence on predicting OCD severity than aggregate scores.
• A neural network model, while not exceeding linear regression in predictive accuracy, offered a richer understanding of OCD heterogeneity and nonlinear relationships.
• Demographic factors significantly contributed to the explanatory power for predicting OCD severity.

Conclusions:

• Machine learning models can achieve predictive power comparable to linear models while preserving essential nonlinear relationships in psychological data.
• The study advocates for sophisticated predictive modeling in psychological research to better capture the complexity of disorders like OCD.
• Findings challenge conventional analytical approaches, emphasizing the value of item-level data and advanced modeling for understanding psychological phenomena.