Overview of "An AI-based Decision Support System for Predicting Mental Health Disorders"

Introduction

Tutun et al. (2023) address the global mental health crisis, noting that approximately one billion people suffer from mental health disorders such as depression, anxiety, bipolar disorder, and schizophrenia. Traditional diagnostic tools like the Symptom Checklist-90-Revised (SCL-90-R) are comprehensive but lengthy, containing 90 questions that require significant time to administer and interpret. This complexity leads to low participation and completion rates, as well as potential inaccuracies due to manual analysis by mental health professionals.

Objective

The study aims to develop an AI-based Decision Support System (DSS) that can efficiently and accurately detect and diagnose multiple mental health disorders using a reduced number of assessment questions. By leveraging advanced analytics and artificial intelligence, the authors seek to create a tool that improves diagnostic accuracy while increasing participation and completion rates.

Methodology

Data Collection and Preprocessing

• Developed a web portal called Psikometrist to digitally administer the SCL-90-R and collect participants' responses.
• Ensured ethical considerations by anonymizing data and removing sensitive demographic information to prevent biases.

Variable Selection and Creation

• Utilized the Network Pattern Recognition (NEPAR) algorithm to analyze the relationships among the 90 questions and identify redundancies.
• Constructed two NEPAR models:
• NEPAR-Q for reducing the number of questions by identifying and removing those with high similarity.
• NEPAR-P for feature engineering by calculating similarities among participants to enhance predictive models.
• Reduced the original 90-question SCL-90-R to a 28-question tool named SCL-28-AI without compromising the ability to diagnose all 10 primary mental disorders.

Model Training

• Trained various machine learning models, including Lasso Regression (L-LR), Ridge Regression (R-LR), Random Forest (RF), and Support Vector Machine (SVM).
• Used participants' responses to the SCL-28-AI and the similarity features from NEPAR-P as inputs.
• Emphasized the use of explainable and transparent algorithms to adhere to ethical AI principles.

Results

• The Lasso Regression model with NEPAR-P features achieved the highest accuracy of 89% in diagnosing mental health disorders.
• The SCL-28-AI demonstrated comparable diagnostic capabilities to the original SCL-90-R but with significantly fewer questions, leading to improved participation and completion rates.
• Incorporating participant similarities via NEPAR-P enhanced the performance of the predictive models.

Implications

Theoretical Implications

• Demonstrated the feasibility of integrating ethical guidelines into AI-based DSS development, emphasizing fairness, transparency, and trustworthiness.
• Provided a methodology for incorporating ethical considerations in AI applications within healthcare.

Practical Implications

• Offered a more efficient assessment tool (SCL-28-AI) that reduces the burden on both patients and mental health professionals.
• Enabled automatic and accurate diagnoses without the need for manual interpretation, allowing professionals to focus on treatment planning.
• Highlighted the potential for AI to improve access to mental healthcare, especially in resource-limited settings.

Ethical Considerations

• Ensured fairness by removing demographic variables that could introduce biases.
• Maintained transparency by using explainable machine learning models and providing insights into the decision-making process.
• Built trustworthiness through collaboration with mental health professionals and adherence to ethical AI design principles.

Limitations and Future Research

• The study utilized data from a specific web platform (Psikometrist), which may limit generalizability.
• Future research should apply the framework to larger and more diverse datasets.
• There is potential to explore more complex models and include additional variables such as genetic data or medication history.

Conclusion

Tutun et al. (2023) successfully developed an AI-based DSS that can accurately diagnose multiple mental health disorders using a streamlined set of 28 questions. By integrating ethical principles into the design and employing advanced analytics, the study offers a promising solution to improve mental health diagnostics. This tool has the potential to enhance clinical decision-making, reduce costs, and increase access to mental healthcare globally.

Reference

Tutun, S., Johnson, M. E., Ahmed, A., Albizri, A., Irgil, S., Yesilkaya, I., Ucar, E. N., Sengun, T., & Harfouche, A. (2023). An AI-based Decision Support System for Predicting Mental Health Disorders. Information Systems Frontiers, 25(3), 1261–1276. https://doi.org/10.1007/s10796-022-10282-5