An Interpretable Machine Learning Approach to Predict Sensory Processing Sensitivity Trait in Nursing Students

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 ===Summary

Authors explored Sensory Processing Sensitivity (SPS), a personality trait associated with heightened sensitivity to stimuli, identifying individuals as Highly Sensitive People (HSP). Focused on nursing students, they developed a predictive model utilizing emotional intelligence, communication skills, and conflict styles. Findings revealed a 33% prevalence of HSP, more pronounced in women and those with prior health training, emphasizing the significance of sex and emotional intelligence dimensions in early detection and suggesting the utility of individualized prediction models for intervention strategies in nursing education.

 

===Points in favor

+ The study uses a novel approach to address an interesting issue.

+ The discussion section is detailed.

 

===Points against

- Related work needs to be organized in tables for easy understanding. 

- Motivation to select ML algorithm is not clear. We know ML algorithm has been useful in many problems. Authors can explain the current issues with traditional approach (possibly at end of - Section 1 or new section related work) and how ML will benefit the SPS problem(Section 4). 

- Section 2.4 can be detailed with some flow diagrams or design diagrams.

- The threats to validity section is missing.

Comments on the Quality of English Language

None

Author Response

EJIHPE-2878626

 

Title: An interpretable machine learning approach to predict sensory processing sensitivity trait in nursing students

 

Authors: Alicia Ponce-Valencia , Diana Jiménez-Rodríguez, Juan José Hernández Morante *, Carlos Martínez Cortés , Horacio Perez-Sanchez, Paloma Echevarría- Pérez 

 

 

REVIEWER #1

Comments and Suggestions for Authors

Summary

Authors explored Sensory Processing Sensitivity (SPS), a personality trait associated with heightened sensitivity to stimuli, identifying individuals as Highly Sensitive People (HSP). Focused on nursing students, they developed a predictive model utilizing emotional intelligence, communication skills, and conflict styles. Findings revealed a 33% prevalence of HSP, more pronounced in women and those with prior health training, emphasizing the significance of sex and emotional intelligence dimensions in early detection and suggesting the utility of individualized prediction models for intervention strategies in nursing education.

Points in favor

+ The study uses a novel approach to address an interesting issue.

+ The discussion section is detailed.

 

We would like to thank the reviewers’ comments, especially those regarding the points in favor. In the present document, the responses to the reviewer’s comments are in blue characters. Moreover, we have revised the comments of the reviewer and tried to follow the indications as indicated, point-by-point, below.

 

Points against

- Related work needs to be organized in tables for easy understanding.

We apologize but we do not really understand this reviewer comments.

- Motivation to select ML algorithm is not clear. We know ML algorithm has been useful in many problems. Authors can explain the current issues with traditional approach (possibly at end of - Section 1 or new section related work) and how ML will benefit the SPS problem (Section 4).

The point of the reviewer is quite interesting. Certainly, here's a more concise response to the reviewer's comments:

 We appreciate your valuable feedback and the opportunity to clarify our choice of ML for predicting the SPS trait. Our motivation stems from the limitations of traditional diagnostic tools in early identification and the need for personalized intervention strategies. Traditional methods often lack the adaptability and precision that ML models offer, especially in handling complex, non-linear relationships within data. ML's capability to learn from and make predictions based on large datasets allows for a nuanced understanding of SPS, offering a promising alternative for early detection. This approach not only enhances diagnostic accuracy but also tailors interventions to individual needs, critical in nursing education where SPS prevalence is significant. We've revised Sections 1 and 4 to elaborate on these points, showcasing ML's advantages over conventional approaches and its potential to improve outcomes for nursing students with SPS.

 

 

- Section 2.4 can be detailed with some flow diagrams or design diagrams.

Thank you again for your valuable suggestion. As per your recommendation, we have now included a flow diagram in Section 2.4 to provide a detailed overview of the machine learning workflow employed in our study. This diagram visually represents the key steps involved in data preprocessing, model training, and interpretability analysis. We believe that the addition of this diagram will enhance the comprehensibility of our methodology for the readers.

 Supplementary Figure 1: Schematic representation of the machine learning workflow employed in the study. The collected data underwent preprocessing steps including outlier removal, variance analysis, one-hot encoding of categorical variables, and correlation analysis. The preprocessed data was then used to train multiple machine learning models, namely Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), Random Forest (RF), RuleFit (RLF), and Support Vector Machine (SVM), using the SIBILA tool. The trained models were evaluated based on their performance metrics, and the best-performing model was selected for further analysis. Interpretability techniques such as Feature Importance, and Shapley Additive Explanations (SHAP) were applied to the selected model to gain insights into the predictive features and their contributions to the model's predictions.

 

- The threats to validity section is missing.

Thank you for highlighting the absence of a "Threats to Validity" section in our manuscript. We agree that discussing the potential limitations and factors that could influence the validity of our findings is crucial for providing a comprehensive and transparent report of our research. To address this, we have now included a new subsection titled "4.1. Threats to Validity" in the Discussion section of the revised manuscript. This subsection discusses the potential limitations of our study, such as the cross-sectional design, sample characteristics, and the need for further validation in diverse populations.

 

Comments on the Quality of English Language

None

 

 

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

This study aims to develop a predictive model for identifying Sensory Processing Sensitivity (SPS) in nursing students, using machine learning techniques. The research involved 672 nursing students, focusing on relationships between SPS and factors like emotional intelligence, communication skills, and conflict resolution styles. The study finds a higher prevalence of SPS in women and those with previous health training. The study also observes specific characteristics associated with SPS, such as greater emotional repair and communication skills.

Strengths:

- Novelty: The study addresses an innovative and significant topic, blending machine learning with psychological traits in a nursing education context.

- Comprehensive Methodology: The approach is thorough, combining traditional statistical methods with advanced machine learning techniques to predict SPS.

- Relevant Instruments: Utilization of well-validated instruments like the Reduced Scale for Highly Sensitive People (r-HSP) and Trait Meta Mood Scale (TMMS-24) ensures reliability in data collection.

Weaknesses:

- Generalizability: The study focuses solely on nursing students from a single university, which might limit the generalizability of the findings to other populations or educational settings.

- Model Interpretability: While machine learning models are used, there is less emphasis on the interpretability of these models. For practical applications, understanding the model's decision-making process is crucial.

- Longitudinal Perspective: The cross-sectional design limits the ability to infer causality or the stability of SPS traits over time.

Suggestions for Improvement:

- Expand the Sample: Future research could include a more diverse sample from multiple universities or include professionals in the nursing field.

- Enhance Model Transparency: Further explanation or use of more interpretable machine learning models would be beneficial, particularly for practical application in educational settings.

- Longitudinal Study: Conducting a longitudinal study would provide insights into how SPS traits develop or change over time in response to educational or environmental factors.

- Additional References: Consider including more literature on the intersection of machine learning and psychological trait prediction to bolster the theoretical framework. For instance, studies on machine learning applications in personality trait prediction could provide additional context and validation for the methodology used.

Comments on the Quality of English Language

The manuscript's English language quality is generally good, with clear and coherent presentation of ideas. The structure follows conventional academic standards, and terminology specific to the fields of machine learning and psychology is appropriately utilized. However, there are some areas where the language could be refined for enhanced clarity and readability:

1. Consistency in Terminology: Ensure consistent use of terms throughout the manuscript. For instance, the term 'SPS' (Sensory Processing Sensitivity) should be consistently used after its initial introduction.

2. Sentence Structure: Some sentences are overly complex, which could potentially confuse the reader. Shorter, more concise sentences may improve readability without losing the necessary technical detail.

3. Use of Passive Voice: While the passive voice is common in scientific writing, occasional use of active voice can make the text more engaging.

4. Minor Grammatical Adjustments: A few grammatical inconsistencies and typographical errors need correction. For instance, careful proofreading to correct minor typographical errors and to ensure subject-verb agreement would enhance the manuscript's professionalism.

5. Clarity in Descriptions: Some sections, especially those describing the machine learning methodology, could benefit from simpler explanations or a brief introduction to certain concepts for readers who might be less familiar with these topics.

Author Response

EJIHPE-2878626

Title: An interpretable machine learning approach to predict sensory processing sensitivity trait in nursing students

Authors: Alicia Ponce-Valencia , Diana Jiménez-Rodríguez, Juan José Hernández Morante *, Carlos Martínez Cortés , Horacio Perez-Sanchez, Paloma Echevarría- Pérez 

 

REVIEWER #2

Comments and Suggestions for Authors

This study aims to develop a predictive model for identifying Sensory Processing Sensitivity (SPS) in nursing students, using machine learning techniques. The research involved 672 nursing students, focusing on relationships between SPS and factors like emotional intelligence, communication skills, and conflict resolution styles. The study finds a higher prevalence of SPS in women and those with previous health training. The study also observes specific characteristics associated with SPS, such as greater emotional repair and communication skills.

Strengths:

- Novelty: The study addresses an innovative and significant topic, blending machine learning with psychological traits in a nursing education context.

- Comprehensive Methodology: The approach is thorough, combining traditional statistical methods with advanced machine learning techniques to predict SPS.

- Relevant Instruments: Utilization of well-validated instruments like the Reduced Scale for Highly Sensitive People (r-HSP) and Trait Meta Mood Scale (TMMS-24) ensures reliability in data collection.

We would also like to thank the reviewers’ comments, for his/her extensive and interesting research. In the present document, the responses to the reviewer’s comments are in blue characters. Moreover, we have revised the comments of the reviewer and tried to follow the indications as indicated, point-by-point, below.

Weaknesses:

- Generalizability: The study focuses solely on nursing students from a single university, which might limit the generalizability of the findings to other populations or educational settings.

We fully agree with the reviewer’s comment. We have included these issues in the limitation section (end of section 4) of the revised paper.

- Model Interpretability: While machine learning models are used, there is less emphasis on the interpretability of these models. For practical applications, understanding the model's decision-making process is crucial.

Thank you for raising an important point regarding the interpretability of the machine learning models used in our study. We agree that understanding the decision-making process of the models is crucial for practical applications and for gaining insights into the factors contributing to the prediction of the SPS trait. To address this, we have now expanded our discussion on model interpretability in the revised manuscript. Specifically, we have added a new paragraph in Section 3.4 "Prediction models of the presence of SPS trait" to highlight the importance of model interpretability and the techniques we employed to gain insights into the predictive features and their contributions to the model's predictions. We have also provided additional details on the interpretability analysis in the Methods section (Section 2.4) to clarify the specific techniques used, such as Feature Importance, Local Interpretable Model-Agnostic Explanations (LIME), and Shapley Additive Explanations (SHAP).

 We believe that these additions will enhance the transparency and interpretability of our machine learning approach, enabling readers to better understand the factors driving the model's predictions and their potential implications for practical applications.

 

- Longitudinal Perspective: The cross-sectional design limits the ability to infer causality or the stability of SPS traits over time.

We fully agree with the reviewer’s comment. We have included these issues in the limitation section (end of section 4) of the revised paper.

Suggestions for Improvement:

- Expand the Sample: Future research could include a more diverse sample from multiple universities or include professionals in the nursing field.

Again, we have included this information as a limitation of the present work.

- Enhance Model Transparency: Further explanation or use of more interpretable machine learning models would be beneficial, particularly for practical application in educational settings.

Thank you for your suggestion to enhance the transparency of our machine learning models, particularly for practical applications in educational settings. We appreciate your insight into the importance of interpretability and explainability in the context of our study.

To address your comment, we have made several additions to the manuscript to emphasize the steps taken to ensure model transparency and interpretability. In Section 2.4, we have now provided more details on the specific interpretability techniques employed, such as Feature Importance, Local Interpretable Model-Agnostic Explanations (LIME), and Shapley Additive Explanations (SHAP). These techniques allow us to identify the key predictive features and their contributions to the model's decision-making process, enhancing the transparency of our approach.

 Furthermore, in Section 4.2 (previously Section 4), we have expanded our discussion on the practical implications of our findings for educational settings. We highlight how the interpretability of our models can inform targeted interventions and support strategies for HSP students, enabling educators to understand the factors influencing the prediction of the SPS trait and tailor their approaches accordingly.

 We believe that these additions will enhance the transparency and applicability of our machine learning approach, making it more accessible and actionable for educators and other stakeholders involved in supporting HSP students.

- Longitudinal Study: Conducting a longitudinal study would provide insights into how SPS traits develop or change over time in response to educational or environmental factors.

As commented previously, we have included these issues in the limitation section (end of section 4) of the revised paper

- Additional References: Consider including more literature on the intersection of machine learning and psychological trait prediction to bolster the theoretical framework. For instance, studies on machine learning applications in personality trait prediction could provide additional context and validation for the methodology used.

The comment of the reviewer is fairly interesting. Following his/her indications, we have included a new paragraph with the following references in the introduction section of the revised paper to expand the theoretical framework of this work: [1–4].

1. Yarkoni, T.; Westfall, J. Choosing Prediction Over Explanation in Psychology: Lessons From Machine Learning. https://doi.org/10.1177/1745691617693393 2017, 12, 1100–1122, doi:10.1177/1745691617693393.
2. Dwyer, D.B.; Falkai, P.; Koutsouleris, N. Machine Learning Approaches for Clinical Psychology and Psychiatry. Annu Rev Clin Psychol 2018, 14, 91–118, doi:10.1146/ANNUREV-CLINPSY-032816-045037/CITE/REFWORKS.
3. Christ, N.M.; Elhai, J.D.; Forbes, C.N.; Gratz, K.L.; Tull, M.T. A Machine Learning Approach to Modeling PTSD and Difficulties in Emotion Regulation. Psychiatry Res 2021, 297, 113712, doi:10.1016/J.PSYCHRES.2021.113712.
4. Zhou, Y.; Han, W.; Yao, X.; Xue, J.J.; Li, Z.; Li, Y. Developing a Machine Learning Model for Detecting Depression, Anxiety, and Apathy in Older Adults with Mild Cognitive Impairment Using Speech and Facial Expressions: A Cross-Sectional Observational Study. Int J Nurs Stud 2023, 146, 104562, doi:10.1016/J.IJNURSTU.2023.104562.

 

 

Comments on the Quality of English Language

The manuscript's English language quality is generally good, with clear and coherent presentation of ideas. The structure follows conventional academic standards, and terminology specific to the fields of machine learning and psychology is appropriately utilized. However, there are some areas where the language could be refined for enhanced clarity and readability:

1. Consistency in Terminology: Ensure consistent use of terms throughout the manuscript. For instance, the term 'SPS' (Sensory Processing Sensitivity) should be consistently used after its initial introduction.
2. Sentence Structure: Some sentences are overly complex, which could potentially confuse the reader. Shorter, more concise sentences may improve readability without losing the necessary technical detail.
3. Use of Passive Voice: While the passive voice is common in scientific writing, occasional use of active voice can make the text more engaging.
4. Minor Grammatical Adjustments: A few grammatical inconsistencies and typographical errors need correction. For instance, careful proofreading to correct minor typographical errors and to ensure subject-verb agreement would enhance the manuscript's professionalism.

We regret these grammatical mistakes and we have revised the paper following your indications. In addition, we have made some modifications to unify terms and occasionally use an active voice.

5. Clarity in Descriptions: Some sections, especially those describing the machine learning methodology, could benefit from simpler explanations or a brief introduction to certain concepts for readers who might be less familiar with these topics.

We appreciate your suggestion to provide simpler explanations and brief introductions to certain concepts to make our work more accessible to readers who may be less familiar with these topics.  To address your comment, we have made several revisions to improve the clarity and comprehensibility of our machine learning methodology:

• In Section 2.4, we have added a brief introduction to machine learning and its relevance to our study. This introduction provides a high-level overview of the key concepts and terminology used in our methodology, making it easier for readers to follow along.
• We have simplified the explanations of the specific machine learning algorithms used in our study (ANN, KNN, RF, RLF, and SVM) by providing concise descriptions of their underlying principles and their application in our context.
• The interpretability techniques (Feature Importance, LIME, and SHAP) are now explained in more accessible terms, highlighting their purpose and how they contribute to understanding the model's decision-making process.

Throughout the manuscript, we have made an effort to use clearer and more concise language when discussing machine learning concepts, avoiding unnecessary jargon and technical terms wherever possible. We believe that these revisions will significantly improve the clarity of our descriptions and make our work more accessible to a broader audience, including those who may not have extensive background knowledge in machine learning.

 

Overall, we wanted to extend our gratitude to the reviewers for their insightful feedback on our manuscript. Your constructive comments have significantly improved our study's clarity and depth. Thank you for your invaluable guidance.

 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have done a good job revising the paper. I'm happy with the new version.

Comments on the Quality of English Language

The manuscript needs only minor editing.