Enhancing Clinical Decision Support for Precision Medicine: A Data-Driven Approach
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors of the paper in this article propose a data-driven framework based on the Simon Decision Model. The paper proposes the concept of Clinical Identification (CEid) dataset and clusters 692 rows of data using k-means clustering method (with the elbow method to determine the optimal number of clusters), which results in four clusters and demonstrates the maximum and minimum values of each type of metrics in each cluster. The paper takes the minimum and maximum values of each numerical variable in each cluster as a reference, and maps and assigns each record in the original data to the corresponding cluster separately. Finally, a random forest model is applied for classification prediction. The paper promotes advances in data-driven aspects and can bring a positive impact in terms of efficiency and cost-effectiveness of resource allocation in the field of precision medicine. However, the paper is not yet up to the level of publication in this journal and needs further improvement in the following areas:
1. The explanatory notes for the data sets are not clear. For example, the refined dataset used for clustering in section 3.3 includes 692 rows and 9 variables. It is recommended that the author's article specifically explain the source and type of data.
2. In the prediction section, the mapping of clustering was performed for a dataset containing 1,631,242 records, followed by classification prediction for 258,054 records with 19 features. However, there are multiple records attributed to a single patient in this part of the data, and if this part of the data is categorized, it is possible that different records of a patient may be attributed to different clusters. Existing articles are missing or too generalized for this part of the explanation. It is recommended that the authors explain in detail how this mechanism could help the future of precision personalized medicine.
3. Temporal clustering is mentioned several times, but the study focuses on emphasizing trends in certain variables over time for individual patients, neglecting to explain other types of temporal clustering studies. It is recommended that the authors provide further analysis and interpretation of other types of temporal clustering studies in the article.
4. Incorrect headings. For example: page 18, heading 5.6.1.
5. Spelling errors. For example: page 22, "daya" spelling error
6. The descriptive aspects of the article's language are suggested for further revision and improvement. For example, in the introduction section, the causal relationship between the efficacy of streptomycin in treating tuberculosis and the concept of tailoring treatment to individual patients is unclear. And the first paragraph very abruptly introduces the current status of people with dementia.
7. Reference format. There is a problem with different font sizes in the references.
Comments on the Quality of English LanguageThe overall quality of the English language is fine, but further revision and improvement of the language descriptions is recommended, and attention should also be paid to spelling.
Author Response
- The explanatory notes for the data sets are not clear. For example, the refined dataset used for clustering in section 3.3 includes 692 rows and 9 variables. It is recommended that the author's article specifically explain the source and type of data
Response: Thank you for your feedback regarding the details on "Data Collection and Exploration." In the manuscript, we have already explained the source and categories of the data on page 7. Specifically, we mentioned that "data was collected from an Intensive Care Unit (ICU). As shown in Figure 2, the dataset includes 10 distinct categories, each offering valuable insights into various aspects of patient health and medical interactions."
To further clarify, we have now added a detailed explanation of the key data processing and transformation steps. The revised section on page 8 outlines the data processing tasks applied to prepare the data effectively for clustering.
- In the prediction section, the mapping of clustering was performed for a dataset containing 1,631,242 records, followed by classification prediction for 258,054 records with 19 features. However, there are multiple records attributed to a single patient in this part of the data, and if this part of the data is categorized, it is possible that different records of a patient may be attributed to different clusters. Existing articles are missing or too generalized for this part of the explanation. It is recommended that the authors explain in detail how this mechanism could help the future of precision personalized medicine.
Response:: Thanks for the comment. It is important to highlight that the goal of clustering is not to group patients into homogeneous groups! The goal is to group clinical conditions into homogeneous clusters. Each cluster shows the behaviour of clinical data regardless of patient identity.
In the application level, Those clusters should be interpreted by doctors to identify conditions of interest and define treatment protocols. Whenever a patient's clinical condition is framed into a cluster (using max and min of the most important variables), the protocol can be started/applied.
- Temporal clustering is mentioned several times, but the study focuses on emphasizing trends in certain variables over time for individual patients, neglecting to explain other types of temporal clustering studies. It is recommended that the authors provide further analysis and interpretation of other types of temporal clustering studies in the article.
Response:Temporal analysis was not the primary focus of this study, so we have revised our terminology to prevent any potential confusion. In this concept, the goal was to monitor and visualize patient's clinical status over time, specifically the days of clinical events, to illustrate how a patient’s clinical profile evolves during their hospital stay. Additionally, using the CIEd formulation, our approach allows for the filtering of clinical data based on specific days. This capability aids decision-makers in analyzing and observing a patient's health status over time, thereby supporting more informed decisions. While we touched upon temporal clustering for this purpose, we acknowledge that further analysis of other types of temporal clustering studies could provide additional insights, and we recommend exploring these in future work.
- Incorrect headings. For example page 18, heading 5.6.1.
Response:Thank you for pointing out the incorrect headings. These have been corrected in the revised version of the manuscript.
- Spelling errors. For example page 22, "daya" spelling error:
Response:We appreciate you highlighting the spelling errors, such as "daya" on page 22. These mistakes have been corrected in the updated version of the manuscript.
- The descriptive aspects of the article's language are suggested for further revision and improvement. For example, in the introduction section, the causal relationship between the efficacy of streptomycin in treating tuberculosis and the concept of tailoring treatment to individual patients is unclear. And the first paragraph very abruptly introduces the current status of people with dementia.
Response: Thank you for your valuable feedback on the manuscript. We have carefully reviewed and revised the descriptive aspects of the article's language, particularly in the introduction section. We clarified the causal relationship between the efficacy of streptomycin in treating tuberculosis and the concept of tailoring treatment to individual patients. Additionally, we have restructured the first paragraph to ensure a smoother transition when introducing the current status of people with dementia.
- Reference format. There is a problem with different font sizes in the references.
Response: Thank you for bringing this issue to our attention. The reference format has been corrected to ensure consistent font sizes throughout. We appreciate your careful review and feedback.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe article is very comprehensive and it outlines the need for personalized medicine and the challenges faced by current Clinical decision support system using data-driven framework.
Article looks like a combination of few chapter’s in a thesis, hard to identify the state of art in this project.
Problem statement is clear, but there is no clarity on sub research questions taken for this study. Authors must highlight the research questions taken for investigation.
Literature review for this study must be included which highlights the strength and weakness of each previous stud. This research has numerous existing works available in the literature. Authors must highlight the research gap and brief how the proposed study bridges the gaps in clinical decision support system.
There is missing of details on “ Data collection and exploration” about data quality, quantity, or integration with other sources.
Techniques and metrics have been detailed, but no clarity which technique is suitable for this study.
Author must detail the Simon's model and how it relates to the proposed framework. Whether the framework help to identify "satisficing" solutions or optimize decision-making under bounded rationality? must be included.
Results are not clearly organized as per the proposed study, authors must rephrase in organizing the result section as per the stages in proposed and method for enhancing clinical support decision support system, and conclusion for this study is not clear.
References must be updated.
Comments on the Quality of English LanguageNo issues
Author Response
2.1. Article looks like a combination of few chapter’s in a thesis, hard to identify the state of art in this project.
Response: Thank you for your valuable feedback. We completely agree with your comment. To address this, we have revised pages 4 and 5 to better highlight the gaps in the literature, particularly focusing on data processing, which is the main scope of this study. Additionally, we have inserted Figure 1 to clarify the identified gaps .
2.2. Problem statement is clear, but there is no clarity on sub research questions taken for this study. Authors must highlight the research questions taken for investigation.
Response: Thank you for your insightful feedback. Page 3. The question and objectives are addressed: “Hence, this research seeks to address the fundamental question: "How can data-driven insights shape the framework of Clinical Decision Support Systems and advance optimal clinical decision-making?" To achieve this overarching goal, a set of interconnected objectives will guide the investigation.”
We aimed to investigate the role data data-driven insight to seek precision medicine for improving the clinical decision-making process.
2.3. Literature review for this study must be included which highlights the strength and weakness of each previous stud. This research has numerous existing works available in the literature. Authors must highlight the research gap and brief how the proposed study bridges the gaps in clinical decision support system.
Response: Thank you for your insightful feedback. As this comment is closely related to the earlier one (#1), we have made significant revisions to address these concerns. We have expanded the literature review to provide a more comprehensive analysis of the strengths and weaknesses of previous studies in this field. Pages 4 and 5 now clearly outline the research gaps and detail how our study bridges these gaps, particularly within the context of clinical decision support systems.
2.4. There is missing of details on “ Data collection and exploration” about data quality, quantity, or integration with other sources.
Response: Thank you for your feedback regarding the details on "Data Collection and Exploration." In the manuscript, we have already explained the source and categories of the data on page 7.Specifically, we mentioned that "data was collected from an Intensive Care Unit (ICU). As shown in Figure 2, the dataset includes 10 distinct categories, each offering valuable insights into various aspects of patient health and medical interactions."
To further clarify, we have now added a detailed explanation of the key data processing and transformation steps. The revised section on page 8.outlines the data processing tasks applied to prepare the data effectively for clustering.
2.5 Techniques and metrics have been detailed, but no clarity on which technique is suitable for this study.
Response: In this study, we applied clustering and classification techniques based on our research questions and objectives. The primary goal was not to rank the techniques but to propose a foundational framework that addresses key data processing challenges. As a result, we did not focus on ranking the AI techniques. However, in the evaluation phase, we utilized various metrics to assess the performance of the clustering and classification algorithms employed. Future research could build upon this work by identifying the most effective predictive techniques to achieve the highest possible accuracy in specific applications.
2.6 Author must detail the Simon's model and how it relates to the proposed framework. Whether the framework help to identify "satisficing" solutions or optimize decision-making under bounded rationality? must be included.
Response: In the introduction, we address the concept of Simon's model in decision-making, particularly how it relates to the idea of "satisficing" rather than optimizing decisions under bounded rationality. In the discussion section, we analyze the relationship between Simon's model and our proposed analytics framework.
This paragraph is added for better clarification “While our study primarily focuses on descriptive and predictive analytics, it lays the groundwork for future research exploring optimization. As discussed, further studies are needed to incorporate simulation techniques and prescriptive analytics, which would enable the identification of "satisficing" solutions that align with the rationality constraints of decision-making. Our framework, therefore, sets the stage for developing decision-making strategies that balance the trade-offs between optimal solutions and those that are practically achievable under bounded rationality”
2.7. Results are not clearly organized as per the proposed study, authors must rephrase in organizing the result section as per the stages in proposed and method for enhancing clinical support decision support system, and conclusion for this study is not clear.
Response: We completely agree with your feedback and appreciate your valuable input. To address the issues in organizing the results section, we identified and corrected mistakes in the titling and subtitling, which should now provide greater clarity. The revised results section is structured into three major parts that align with the stages of the proposed method for enhancing the clinical decision support system. We have also worked to clarify the conclusion to better reflect the study's findings.
3.1. Crafting CEid & Interrelating clinical events
3.2. Analytical insight through CEid
3.3. Unveiling patterns throughout temporal clustering analysis
3.4. Insight into the data behavior within clusters
3.5. Predicting the temporal cluster on data behavior
To offer a clearer conclusion, we revised it to emphasize the key findings and insights presented in this study.
2.8 References must be updated.
Response: Thank you for your advice. Out of the 42 references, 12 are from 2020 or later, and we have made an effort to incorporate more recent sources.
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsRevised version looks statisfactory.
