PONE-D-24-32699Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical NotesPLOS ONE

Dear Abuhmed,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Nov 08 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Menglin Yang

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf.

2. Please note that PLOS ONE has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“The authors extend their appreciation to the King Salman Center for Disability 687 Research for funding this work through Research Group no KSRG-2023-101.”

We note that you have provided additional information within the Acknowledgements Section that is not currently declared in your Funding Statement. Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“King Salman Center for Disability Research for funding this work through Research Group no KSRG-2023-101”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

4. We note that you have indicated that there are restrictions to data sharing for this study. For studies involving human research participant data or other sensitive data, we encourage authors to share de-identified or anonymized data. However, when data cannot be publicly shared for ethical reasons, we allow authors to make their data sets available upon request. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

Before we proceed with your manuscript, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., a Research Ethics Committee or Institutional Review Board, etc.). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of recommended repositories, please see https://journals.plos.org/plosone/s/recommended-repositories. You also have the option of uploading the data as Supporting Information files, but we would recommend depositing data directly to a data repository if possible.

Please update your Data Availability statement in the submission form accordingly.

5. Please note that your Data Availability Statement is currently missing [the repository name]. If your manuscript is accepted for publication, you will be asked to provide these details on a very short timeline. We therefore suggest that you provide this information now, though we will not hold up the peer review process if you are unable.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: I Don't Know

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This paper proposes an ensemble framework using large language models (LLMs) and language models (LMs) to predict mortality risk for ICU patients with mental disorders, based on unstructured clinical notes. The study explores different model architectures, input modalities, and ensemble strategies to improve prediction accuracy.

Statistical analyses:

1- Standard classification metrics (accuracy, precision, recall, F1) are used.

2- Results are presented comprehensively across different models and configurations.

3- Statistical significance tests are not reported, which would strengthen comparisons.

4- Confidence intervals or error bars on results would help assess reliability.

Data and Analysis Evaluation:

Use of MIMIC-IV provides a large, diverse dataset

Appropriate filtering to focus on patients with mental disorders

Balanced dataset creation helps avoid bias

Inclusion of both discharge summaries and radiology reports is valuable

The sample size of 11,914 patients is substantial. Data preprocessing steps are described, though more details on handling of missing data would be helpful.

Writing Quality Check:

The paper is generally well-written and clearly organized. The methodology and results are presented in a logical flow.

# Minor comments that should be resolved or included:

1- Abstract could be more concise and highlight key findings more clearly

2- Some sentences are overly long and complex, impacting readability

3- A few grammatical errors and typos need correction

4- Figures could be improved for clarity, especially Fig. 1

5- Lack of comparison to non-LLM baselines or clinical risk scores, should be included

6- No ablation studies to isolate impact of different components, should be included

7- Lack of complexity analysis of Fine-tuning LLMs, should be included

8- Limited exploration of early fusion or other fusion strategies, should be included

9- Missing statistical significance tests for model comparisons

10- Inadequate discussion of limitations and potential biases

Reviewer #2: This study is an interesting study on predicting mortality rates in ICU patients with mental disorders from unstructured medical notes using a LLM ensemble method. While the idea for the study is very relevant in the contemporary times, there are some improvements that this study needs to make it suitable for publication:

1. The paper would benefit from condensing the points of focus into specific Research Questions (RQ), which can then be referred to in the discussions later in the paper to connect the relevant findings with specific RQs for a more complete discussion of the results.

2. F1 score is the metric for evaluation used in this paper, which might be a common method in this usecase but the authors still needed to substantiate with appropriate citations for their reasoning to use it.

3. The paragraphs need to have a 'justified' alignment.

4. The tables need to be more legible.

5. The results could be better presented with some form of visualisation.

Reviewer #3: Title: Mortality Prediction for ICU Patients With Mental Disorders Using Large Language Models Ensemble and Unstructured Medical Notes

Summary

The article presents the comparison of various LM/LLM model configurations and tuning for predicting in-hospital (ICU) deaths of patients determined to be suffering from mental health disorder. The study uses four LLM and four LM with textual data derived from the MIMIC IV dataset (discharge summaries and radiology reports). Results are presented for a balanced (equal number of patients with/without mental health disorder)

Strengths

The article is well written and easy to read.

The authors do a good job in motivating the paper, i.e. mental health as a risk factor for mortality in ICU admissions.

The literature review is thorough and clearly introduces past efforts and state-of-the-art ML/DM techniques for utilising text-based input to predict in-hospital mortality.

The LM/LLMs selected for the study are good.

The evaluation of the various forms of the selected LM/LLM - out of the box, tuned, ensemble and different modalities for tuning is interesting and informative.

The paper's method and findings do address the research questions raised in the Introduction.

Major Weaknesses

It is not clear how the patients were assigned to "mental disorder". The authors mention ICD-10 code beginning with "F" relating to mental health disorders. However, it is not clear how the "F" ICD-10 code is used. Is a single "F" code in the patient's entire history sufficient to set this flag in the input feature? Is the feature binary (has/has not mental health disorder)? Is the specific ICD-10 code included as a feature.

The authors did not make clear the patient distribution in MIMIC IV such as how many of the patients suffered from mental disorders and how many did not, or how many died in ICU and how many survived (for at least one year). This is particularly important in choosing the data - were all patients who died in ICU included and a random selection of the remaining patients included to balance the study dataset then chosen? The study could have been made more robust by choosing multiple partitions of died/not died patients.

I would like to have seen a more detailed description of the dataset used for the study. As far as I can tell, the input consists of Discharge Summary, Radiology, a flag for mental disorder and a target variable (died in ICU).

Description of the training/validation/testing regime is limited. The authors mention the dataset was partitioned 90% training/validation and 10% testing. However, there is no description of the number or repetitions of the experiments, whether different partitioning was used in different experiment runs, etc.

The results reported are the "best" for each model and experiment. It would be useful to know the range of results and whether the "best" is significantly different from the usual or the worst result.

The paper addresses the question of how various LM/LLM models compare in predicting death in ICU. There are already clinical multi-morbidity models in use that generate mortality risk predictions (over various time horizons - 90 day, 1 year, 5 year). These models are already accepted and in clinical use. I think the paper would be strengthened by comparing with one or more of these models.

Concerns

The authors have relied heavily on pre-prints or unpublished archive references (12 out of 67 references). I think this is not a good idea.

To make it easier for the reader to pick out the best performing models, the table rows containing these should be bolded/highlighted.

As far as I can see, the datsets used for training are not made available to the reader.

Corrections

Introduction

anorexia Nervosa -> anorexia nervosa

Section 3.2

(1st sentence) particularly those that involves -> particularly those that involve

Section 3.3

Figure 1 illustr2ates -> illustrates

References

There are many malformed (missing key information such as publisher, etc). These include

There are many malformed references for URLs (author attribution is incorrect). These include [2], [3], [4], [5], [6]

Some references use full journal name, others abbreviations. Referencing should be consistent - suggest using teh full journal name throughout - see [19], [56], [58]

There is inconsistent capitalisation of journal names. Some capitalise each word, others only the first word. Again, the authors are advised to be consistent and to adopt the standard of the journal to which the article has been submitted.

Similarly, there is inconsistent capitalisation in article titles. Some Some capitalise each word, others only the first word. In some, the first word after a colon is capitalised, while in others it is not. Also, there are instances where capitals should be used to refer to an artefact - e.g MIMIC IV rather than Mimic iv (ref [54]) and Mental-LLM rather than Mental llm (ref [26]). Again, the authors are advised to be consistent and to adopt the standard of the journal to which the article has been submitted.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes: Robert Andrews

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PONE-D-24-32699R1Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical NotesPLOS ONE

Dear Authors:

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Feb 16 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Menglin Yang

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #3: (No Response)

Reviewer #4: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #3: Partly

Reviewer #4: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #3: Yes

Reviewer #4: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #3: No

Reviewer #4: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #3: Yes

Reviewer #4: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #3: Title: Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical Notes

Comments:

I thank the authors for addressing my comments. I am happy with the responses for all but three of my comments. All have to do with data.

Firstly, I am still unsure about the inclusion criteria for patients. The authors clearly state the set of ICD codes that identify mental disorders. What I am missing is if inclusion was based on the presence of one or more of these codes in the latest admission, or whether the inclusion is based on the occurrence of one of these codes in the entire patient history. For instance, if a patient has more than one admittance in the database and an early admittance includes an F50-F59 code, but the latest admittance does not show this code, would the patient be included in the study?

Also, is the mental disorder required to be the primary diagnosis or just included as one of the diagnosis codes? In my experience, hospital records may include up to 20 diagnosis codes.

Next, I am still unsure about the structure feature vectors. My understanding is that the only attributes presented include either/both discharge/radiology report and a target, with no other demographic or clinical history information being included. I think this needs clarifying.

Lastly, I feel that not taking into account comorbidities may introduce some bias into the data. What fraction of the patient cohort had potentially life threatening comorbidities (as well as mental disorder)? That is, are mental disorders independent of other conditions? Were comorbidities equally distributed across the [died |did not die] partitions?

I understand that this is not a clinical paper, it is about experimenting with the capabilities of medically trained LLM/LM models. Nevertheless, it is important to understand and acknowledge the characteristics of the data as they impact (or do not impact) on prediction.

I commend the authors on the modifications they have made to the paper. It is indeed improved and will be a significant contribution.

Minor

Spelling mistake: Section 4.4, 1st paragraph - archives -> achieves

Reviewer #4: The article has successfully met all the necessary requirements for publication, adhering to the stipulated guidelines and standards. After thoroughly reviewing its content, structure, and alignment with the publication's objectives, I am pleased to confirm that there are no outstanding issues or points of concern. As such, I have no further questions or comments to raise at this stage.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #3: Yes: Robert Andrews

Reviewer #4: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PONE-D-24-32699R2

Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical Notes

PLOS ONE

Dear Dr. Abuhmed,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we have decided that your manuscript does not meet our criteria for publication and must therefore be rejected.

Specifically:

==============================

ACADEMIC EDITOR: Please insert comments here and delete this placeholder text when finished. Be sure to:

• Clearly explain how the manuscript fails to meet the PLOS ONE publication criteria. Please ensure your decision is not justified on novelty or impact.
• Provide specific feedback from your evaluation of the manuscript

For Lab, Study and Registered Report Protocols: These article types are not expected to include results but may include pilot data. Please do not reject for lack of results.

==============================

I am sorry that we cannot be more positive on this occasion, but hope that you appreciate the reasons for this decision. Please try to solve reviewer's concerns and resubmit.

Kind regards,

Academic Editor

PLOS ONE

PONE-D-24-32699R3Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical NotesPLOS ONE

Dear Dr. Abhumed,

Thank you for submitting your manuscript to PLOS ONE. After consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a minor revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by 20 June 2025. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Minh Huu Nhat Le, MD

Academic Editor

PLOS ONE

Journal Requirements:

1. Please note that your Data Availability Statement is currently missing [the repository name]. If your manuscript is accepted for publication, you will be asked to provide these details on a very short timeline. We therefore suggest that you provide this information now, though we will not hold up the peer review process if you are unable.

2. We note that you have indicated that there are restrictions to data sharing for this study. For studies involving human research participant data or other sensitive data, we encourage authors to share de-identified or anonymized data. However, when data cannot be publicly shared for ethical reasons, we allow authors to make their data sets available upon request. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

Before we proceed with your manuscript, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., a Research Ethics Committee or Institutional Review Board, etc.). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of recommended repositories, please see https://journals.plos.org/plosone/s/recommended-repositories. You also have the option of uploading the data as Supporting Information files, but we would recommend depositing data directly to a data repository if possible.

Please update your Data Availability statement in the submission form accordingly."

3. Please upload a copy of Figure 1 & 3 which you refer to in your text on your paper. Or if the figure is no longer to be included as part of the submission please remove all reference to it within the text.

4. Please ensure that you refer to Table 1 in your text as, if accepted, production will need this reference to link the reader to the Table.

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Additional Editor Comments (if provided):

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #5: (No Response)

Reviewer #6: All comments have been addressed

Reviewer #7: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #5: Yes

Reviewer #6: Partly

Reviewer #7: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #5: Yes

Reviewer #6: N/A

Reviewer #7: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #5: Yes

Reviewer #6: No

Reviewer #7: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #5: Yes

Reviewer #6: Yes

Reviewer #7: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #5: Paper is well-written, clear and easy-to-follow. The technical soundness and novelties are strong. The experiments are extensive that show all aspects to assess the effectiveness of methods. Fig 2 quality should be increased. I believe this work is sufficient to be published.

Reviewer #6: The idea is okay, but it mostly just applies existing tools (LLMs, ensembling) without doing anything really new.

The paper doesn’t compare the models to basic baselines like logistic regression or ICU scoring systems (e.g., SOFA, SAPS), which makes it hard to judge if these LLMs are actually better.

They use majority voting for the ensemble, but don’t explain why that method was picked or if other fusion methods were tested.

It’s unclear how long clinical notes (which can be huge) were handled. Were they cut off? Summarized? This could affect the results a lot.

Only one dataset (MIMIC-IV) is used, and it’s just for patients with mental disorders. It’s hard to know if the results would work for other types of patients or in different hospitals.

The results don’t include confidence intervals or show if the improvements are statistically solid. Some scores are very close — we don’t know if the gains are real.

The paper doesn’t explain what the models are learning or how decisions are made — there’s no interpretability, which is important in clinical work.

Writing is a bit wordy in places and technical details (like preprocessing or how inputs were built) could be clearer.

Reviewer #7: Overall: This manuscript presents a novel approach to predicting ICU mortality in patients with mental disorders using large language models and unstructured clinical notes, demonstrating considerable contributions with impressive performance results. Although there are several grammatical and typographical errors, they are not significant. However, there are some parts in both methods and experiments that need to be revised.

Strength:

- The introduction provides a comprehensive background on mental health disorders with current statistics from reputable sources (WHO, NIH). It also thoroughly reviews existing approaches for ICU mortality prediction. Based on these bases, this study effectively establishes the novelty of using LLMs for mortality prediction in mentally disordered ICU patients.

- Research questions (RQ1-RQ7) are specific and directly address identified gaps in the literature. The contributions are clearly articulated and aligned with the research objectives.

- The dataset description is presented well, providing clear information about the MIMIC-IV clinical notes and inclusion criteria for mental disorder patients.

- Although the methodology section lacks some foundations to make it more rigorous, it is generally good and can be considered one of the most important contributions of the paper due to its novelty.

- The experimental setup is fair, and the experimental results are impressive.

Weakness:

- The introduction is excessively detailed in certain areas, along with problems in paragraph division (particularly in paragraphs 3 & 4, lines 67-101), creating a disjointed narrative. As a results, research gaps are scattered throughout rather than cohesively presented.

- Despite the adequate general motivation, it is not as compelling as it could be due to the fragmented presentation of research gaps. Another reason is that the connection between mental disorders and ICU mortality risk could be more explicitly linked to the proposed technical approach.

- There are many research questions (RQ1-RQ7). It would be better if these questions can be clustered into logical categories (e.g., "Model Architecture Questions," "Data Modality Questions").

- The statement "The ensemble of heterogeneous language models (LMs) and LLMs effectively addresses these limitations" in the abstract is imprecise, as LLMs are inherently a subset of language models.

- The Methods section lacks a sufficient explanation for using Low-Rank Adaptation (LoRA) over alternative fine-tuning approaches. Figure 2 appears to be directly reused from the original LoRA paper rather than being redrawn specifically for this study's context.

- Data preprocessing steps are minimally described, with limited details on tokenization and handling of medical terminology.

- The claim that "the utilized dataset is balanced with respect to the presence of severe comorbidities" (lines 328-332) lacks supporting evidence such as statistical tests or distribution visualizations.

- Several instances of redundant terminology appear throughout the manuscript (e.g., "LLM models" where"LLM" already means "Large Language Model"), and Section 3.2 contains a duplicated colon at the end of paragraph 2 (line 368).

- Table 2 is too small, making it difficult to verify its contents.

- Some sentences have awkward phrasing or grammatical issues (e.g., lines 512-513, 601, 652-654).

Concern:

- The manuscript lacks visualizations of data distribution and experiments such as mental disorder categories, length distributions of clinical notes, model performance across different patient subgroups, or confusion matrices for the best-performing models.

- The preprocessing for datasets is not described to ensure reproducibility and justification.

- Is the 1B parameter threshold for distinguishing between LMs and LLMs arbitrary or based on evidence?

- Currently, the number of datasets which are employed is not enough, making it difficult to confirm the robustness and generalizability of the proposed method's effectiveness. There is a need to consider expanding the experimental scope in terms of data.

Final decision: Major Revision

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #5: No

Reviewer #6: No

Reviewer #7: Yes: Minh-Toan Dinh

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PONE-D-24-32699R4Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical NotesPLOS ONE

Dear Dr. Abuhmed,

Thank you for submitting your manuscript to PLOS ONE. I am pleased to inform you that your manuscript has been provisionally accepted, pending minor revisions. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please revise your manuscript carefully to address all minor remaining issues. In particular, we request that you conduct a thorough check for grammar, spelling, and clarity to ensure the manuscript meets the standards of written academic English. Additionally, please confirm that your manuscript has been screened for any unintentional plagiarism and that all sources are properly cited. While the reviewers have recommended acceptance, we ask that you review and incorporate all minor comments raised throughout the review process. Once these revisions are completed, we will proceed with final acceptance and move the manuscript into production.

Please submit your revised manuscript by Sep 18 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Minh Le, MD, PhD candidate

Academic Editor

PLOS ONE

Journal Requirements:

If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. 

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #5: All comments have been addressed

Reviewer #8: All comments have been addressed

Reviewer #9: All comments have been addressed

Reviewer #10: All comments have been addressed

Reviewer #11: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #5: Partly

Reviewer #8: Yes

Reviewer #9: Yes

Reviewer #10: Yes

Reviewer #11: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #5: Yes

Reviewer #8: Yes

Reviewer #9: Yes

Reviewer #10: Yes

Reviewer #11: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #5: Yes

Reviewer #8: Yes

Reviewer #9: Yes

Reviewer #10: Yes

Reviewer #11: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #5: Yes

Reviewer #8: Yes

Reviewer #9: No

Reviewer #10: Yes

Reviewer #11: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #5: Latest revision is well-written. Idea is clear. Experimental quantitative results are sufficient. I believe this work is ready for publication.

Reviewer #8: The paper can be accepted for publication.

The paper can be accepted for publication.

The paper can be accepted for publication.

Reviewer #9: I appreciate the authors’ thoughtful and well-structured responses to the initial comments. Most concerns have been addressed with appropriate revisions and clarifications.

Reviewer #10: Authors have addressed all the raised concerns by the reviewers. The quality of the paper is now in a standard scale and reay to be published.

Reviewer #11: In summary, the authors have rigorously addressed previous reviewer concerns and the work represents the solid scientific contribution with clear clinical relevance. They had conducted extensive ablation studies and justified theories to demonstrate their hypotheses. Therefore, I recommend an acceptance for this manuscript.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #5: No

Reviewer #8: No

Reviewer #9: No

Reviewer #10: No

Reviewer #11: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Mortality Prediction for ICU Patients with Mental Disorders Using Large Language Models Ensemble and Unstructured Medical Notes

PONE-D-24-32699R5

Dear Dr. Abuhmed,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Minh Le, MD, PhD candidate

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments: