Explainable AI-Based Identification of Contributing Factors to the Mood State Change in Children and Adolescents with Pre-Existing Psychiatric Disorders in the Context of COVID-19-Related Lockdowns in Greece

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Title: "Explainable AI-based identification of contributing factors to the mood state change of children and adolescents with pre-existing psychiatric disorders in the context of COVID-19 related lockdowns in Greece"

Summary:

The article "Explainable AI-based identification of contributing factors to the mood state change of children and adolescents with pre-existing psychiatric disorders in the context of COVID-19 related lockdowns in Greece" provides a valuable exploration of the impact of COVID-19 lockdowns on the mood states of children and adolescents with pre-existing psychiatric disorders in Greece. The use of Explainable AI (Artificial Intelligence) techniques to identify contributing factors is an innovative approach that adds to the existing body of knowledge. The goal of this study is to use explainable machine learning to pinpoint the variables that account for the worsening or improvement of mood states in clinical samples of young people. The goal of this work is to use a transparent machine learning pipeline to detect and evaluate the influence of the most important aspects of mood state change on the prediction output. With a 76% Best AUC-ROC Score and features, the Random Forest model outperformed all other machine learning classifiers in terms of accuracy. An explainability study revealed that the COVID-19 epidemic and the imposition of limitations caused stress or led to favorable effects.

The article presents relevant findings and implications. However, there are areas that require revision and further improvement to enhance the clarity and quality of the paper.

The following suggestions and comments are provided to guide the revision process:

Introduction:

a. The introduction should provide a clear and concise overview of the research topic, highlighting the significance of studying the mood state changes in children and adolescents with pre-existing psychiatric disorders during COVID-19 lockdowns in Greece. Consider providing more context on the specific psychiatric disorders under investigation.

b. The introduction should clearly state the research objectives and research questions addressed in the study. These should be explicitly mentioned to guide readers throughout the article.

Methodology:

a. Provide a detailed description of the data collection process, including the sources and methods used to collect mood state data from children and adolescents with pre-existing psychiatric disorders. Specify the timeframe and sample size to enhance the replicability of the study.

b. Explain the AI techniques utilized for the analysis and identification of contributing factors. Clarify the explainable AI methods used and provide references to support their validity and effectiveness in this context.

c. Describe the ethical considerations taken into account when working with vulnerable populations and handling sensitive data. Discuss any measures implemented to protect privacy and ensure informed consent.

Results:

a. Clearly present and discuss the findings of the study, focusing on the identified contributing factors to mood state changes. Provide statistical or qualitative evidence to support the conclusions drawn. Utilize visual aids (e.g., tables, graphs) to enhance the clarity and interpretation of the results.

b. Consider including a discussion on the limitations of the study, such as any potential biases in the data collection or analysis process. Addressing these limitations will strengthen the credibility and generalizability of the findings.

Discussion:

a. The discussion section should analyze the results in the context of existing literature on the effects of lockdowns on the mood states of children and adolescents with psychiatric disorders. Highlight the novel contributions of this study and discuss any discrepancies or similarities with previous research.

b. Emphasize the potential implications of the findings for clinical practice, policy-making, and future research. Provide actionable recommendations for supporting the mental well-being of this vulnerable population during similar crises.

References:

a. Review and update the reference list to include the most relevant and recent sources related to the topic. Ensure that all in-text citations are correctly listed, and vice versa. Follow the preferred citation style consistently throughout the article.

Overall Structure:

a. Review the overall structure of the article, ensuring logical flow and coherence between sections. Consider the use of appropriate headings and subheadings to enhance the organization and readability of the paper.

Data Availability and Reproducibility:

a. If possible, provide information regarding the availability of the dataset used in the study to promote transparency and reproducibility. Indicate any restrictions or conditions associated with accessing the data.

By addressing the suggestions and comments provided in this revision report, the article can be enhanced to offer a comprehensive and valuable contribution to the understanding of mood state changes in children and adolescents with pre-existing psychiatric disorders during COVID-19-related lockdowns in Greece.

Comments on the Quality of English Language

Writing Style and Clarity:

a. Ensure that the article is written in clear, concise, and coherent language. Pay attention to sentence structure, grammar, and punctuation. Proofread the manuscript thoroughly to eliminate typographical errors and improve overall readability.

b. Use appropriate subheadings to organize the content effectively, facilitating ease of navigation for readers.

c. Consider explaining technical terms, acronyms, and abbreviations to improve comprehension, especially for readers who may not be familiar with the specific field of study.

Author Response

we would like to thank the reviewer for the valuable comments. attached you can find the responses

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The paper titled "Exploring Factors Affecting Mood State in Youth with Preexisting Psychiatric Disorders using Explainable Machine Learning" by Ntakolia et al. investigates factors contributing to mood changes in youth with preexisting psychiatric disorders using an explainable machine learning approach. The authors conducted a comparative evaluation of seven machine learning algorithms and identified 13 important features, including stress and time spent on social media, using SHAP values from a random forest model. The topic is novel, the methodology appears sound, and the findings are interesting. The authors have also demonstrated a thorough understanding of the limitations of their study, and their efforts in acknowledging these limitations are commendable. However, there are a few concerns that need to be addressed before considering the paper for publication.

 

Major Comments:

 

1. The paper lacks information on the programming language and machine learning framework used in the study. Additionally, details about the hyperparameters of the models are missing. These aspects are crucial for reproducibility and understanding the model's performance. The authors should provide this information in the Methods section.

 

2. In Figure 6, while sorting features by the absolute values of SHAP values is useful, the authors should also consider displaying the SHAP values along with their signs. This will help readers understand the direction of influence each variable has on the response.

 

3. Although the random forest model exhibits the largest AUC-ROC, it would be valuable to compare the features and SHAP values used in each model. Additionally, a comparison of the most contributing factors identified in this study with those from other relevant studies would enhance the discussion's depth.

 

4. The authors converted mood changes into binary categories (negative and positive/stable) for developing classifiers. It might be insightful to explore using the raw mood change scores as a continuous response variable, even if the results are not as effective as binary classifiers. This approach could provide further insights into the relationships between mood changes and contributing factors.

 

Minor Comments:

 

1. In the abstract (line 44) and the conclusion section (line 329), the authors referred to the random forest model achieving "highest accuracy of 76%." It appears that the authors meant "largest ROC-AUC" instead of accuracy, which are two distinct metrics.

 

2. Lines 117 – 120 contain an awkward sentence. The authors should consider rephrasing it for clarity and readability.

 

3. Lines 124 – 126 state that "The ML approach as compared to traditional statistical models are not constrained by assumptions." This statement is inaccurate and could be rephrased to more accurately represent the advantages of the ML approach over traditional statistical models.

 

4. Figures 4 and 5 appear to be redundant. The authors should consider consolidating or clarifying the content to avoid duplication.

 

5. The funding section (Lines 365 – 368) should be updated to reflect the current funding sources, if applicable.

 

Overall, this study addresses an important and novel topic, and the methodology appears robust. However, the authors should address the major concerns related to missing details about the programming language, machine learning framework, and hyperparameters used. Additionally, they should consider incorporating the minor suggestions to enhance clarity and accuracy throughout the paper.

Comments on the Quality of English Language

The authors' English is generally proficient, but there are a few instances of awkward sentence structures that could be improved for better clarity and readability.

Author Response

we would like to thank the reviewer for the comments. attached you can find the responses

Author Response File: Author Response.docx