Do We Need Another CT Scanner?—The Pilot Study of the Adoption of an Evolutionary Algorithm to Investment Decision Making in Healthcare

Round 1

Reviewer 1 Report

This study, in my opinion, makes an important contribution to the field of healthcare resource allocation using machine learning (ML). The study proposes a feasible approach for using ML algorithms to support investment decisions in the healthcare sector, particularly for high-cost medical devices.

The study's methods are appropriate and well-described, beginning with a review of the literature to identify relevant epidemiological and clinical need predictors. The authors developed an evolutionary algorithm model for predicting the need for CT scanners across local counties in Poland using data from reliable sources such as The Central Statistical Office and The National Health Fund.

The study's findings are clear, indicating that over 4 million CT scan procedures were performed across 130 counties in Poland in 2019. The study found 39 instances of agreement between historical data and the EA model's hypothetical scenario based on epidemiological and clinical need predictors. The EA model predicted a lower need for CT scanners than historical data in 58 cases, while a higher number of CT procedures were required in 22 counties. The remaining 11 cases were unresolved.

 

 

The study's conclusion emphasises the potential benefits of using machine learning techniques for optimal healthcare resource allocation. It emphasises the automation of health policymaking as well as the flexibility and transparency brought about by the use of ML in investment decisions in the healthcare sector.

Overall, this study provides valuable insights into the potential use of ML in healthcare resource allocation, and the authors have presented their findings in a clear and concise manner. I recommend minor revisions to clarify some of the language used. Besides, the presentation of the data and result can be more attractive: informative takeaway message in table title is needed, also, please highlight the important results that are matters to the research. The reference format should be improved. 

 

1.     What is the main question addressed by the research?

The authors would like to know if adoption of evolutionary algorithms to the investment decision making in healthcare is feasible or not in this study.

 

2. Do you consider the topic original or relevant in the field, and if
so, why?

It is hard to decide if it is original or not. The problem they are addressing seems to be a broad topic. Of course, the dataset might be unique and that the specific method is highly probable to be original to the best of my knowledge.

 

3. What does it add to the subject area compared with other published
material?

This is the key weakness of the paper. The authors did not really conduct an in-depth literature review and that there is no comparison of the other published material.

 

4. What specific improvements could the authors consider regarding the
methodology?

To be frank, I can not observe any specific improvement. To me, it is of no surprise as this is a pilot study as indicated in the title.

5. Are the conclusions consistent with the evidence and arguments
presented and do they address the main question posed?

They indicated that it is possible to adopt ML to the decision making in the healthcare sector. The results somehow align with the conclusions made.

 

6. Are the references appropriate?

The reference format is not appropriate and chaotic, the references chosen are relevant though.

7. Please include any additional comments on the tables and figures.

Tables are difficult to read, a lot of redundant results, no highlight of result to readers make it difficult to grasp the gist or key takeaways for each table. In short, the presentation of results is not attractive and not captivating.

 

Author Response

Dear Sirs

thank you very much for your kind in depth review of our publication. Here comes responses to the reviewers' suggestions with indications what kind of changes we have introduced to address all comments.

Response 1 

 

1. What is the main question addressed by the research?

The authors would like to know if adoption of evolutionary algorithms to the investment decision making in healthcare is feasible or not in this study.

 

2. 
   Do you consider the topic original or relevant in the field, and if
   so, why?

It is hard to decide if it is original or not. The problem they are addressing seems to be a broad topic. Of course, the dataset might be unique and that the specific method is highly probable to be original to the best of my knowledge.

 

3. 
   What does it add to the subject area compared with other published
   material?

This is the key weakness of the paper. The authors did not really conduct an in-depth literature review and that there is no comparison of the other published material. RESPONSE: ADDITIONAL LITERATURE SEARCH WAS CONDUCTED AND ADDITIONAL REFERENCES WERE INSERTED (FOUR references -3,33,44,45). IT HAS TO BE UNDERLINED HOWEVER THAT THERE ARE LIMITED RESEARCH IN THIS FIELD HENCE LIMITED COMPARISON DESPITE EXTENSIVE LITERATURE REVIEW WE HAVE CONDUCTED.

 

4. 
   What specific improvements could the authors consider regarding the
   methodology?

To be frank, I can not observe any specific improvement. To me, it is of no surprise as this is a pilot study as indicated in the title.

5. 
   Are the conclusions consistent with the evidence and arguments
   presented and do they address the main question posed?

They indicated that it is possible to adopt ML to the decision making in the healthcare sector. The results somehow align with the conclusions made.

 

6. Are the references appropriate?

The reference format is not appropriate and chaotic, the references chosen are relevant though.

RESPONSE: IT HAS BEEN IMPROVED

7. 
   Please include any additional comments on the tables and figures.

Tables are difficult to read, a lot of redundant results, no highlight of result to readers make it difficult to grasp the gist or key takeaways for each table. In short, the presentation of results is not attractive and not captivating.

 

RESPONSE: IT HAS BEEN IMPROVED

 

 

Author Response File: Author Response.docx

Reviewer 2 Report

Well done on compiling an interesting manuscript which takes a novel approach to investigating resourcing of CT services based on machine learning. The manuscript is largely well written although minor grammatical edits are needed throughout. 

On what basis was the formula for fitness function derived - solely by the authors or based on previous work? Is this the works principal limitation as the scientific rationale for just including these  specific weighting factors (ratios of females, old, young, patients, scanners etc). Was any consideration given to other factors such as the national cancer registry statistics per region etc or the location of major trauma centres / tertiary care centres which are more likely to experience more referrals and need for CT than smaller local hospitals etc? Also need to discuss whether the % of females etc impacted results esp as the distribution of females appears consistent throughout the tables - could a regression analysis be applied to investigate the strength of inlfuence of each demographic indicator?

Discussion could benefit from more analysis of the results here and the ilkely practical impact on local service provision and health service decisions - much of the content of the conclusion should be moved into the discussion as the conclusion is to summarise the primary finding(s) of the work only

Any consideration given to the optimal use of CT services - CT scanners are large investments but may not be utilised constantly throughout the week (weekends, nights etc)

It would have been useful to present data for inappropriate CT scans here as well for an overview on the populatoin level as to how many were estimated not to be appropriate. 

Formatting of tables needs to be improved

 

Author Response

Dear Sirs,

thank you for your comments. here comes the response to the Reviewer 2

On what basis was the formula for fitness function derived - solely by the authors or based on previous work? Is this the works principal limitation as the scientific rationale for just including these  specific weighting factors (ratios of females, old, young, patients, scanners etc). Was any consideration given to other factors such as the national cancer registry statistics per region etc or the location of major trauma centres / tertiary care centres which are more likely to experience more referrals and need for CT than smaller local hospitals etc? Also need to discuss whether the % of females etc impacted results esp as the distribution of females appears consistent throughout the tables - could a regression analysis be applied to investigate the strength of inlfuence of each demographic indicator?

RESPONSE: ADDITIONAL DESCRIPTION TO THE RATIONALE FOR PREDICTORS FOR THE FUNCTION WAS DESCRIBED. DUE TO LIMITED DATA ABOUT GENDER IN THE DATABASE, WE COULD NOT INVESTIGATE THE IMPACT OF FEMALE ON RESULTS (GENDER ALONG OTHER NEED PREDICTORS WERE DERIVED FROM DIFFERENT DATA SOURCE)

Discussion could benefit from more analysis of the results here and the ilkely practical impact on local service provision and health service decisions - much of the content of the conclusion should be moved into the discussion as the conclusion is to summarise the primary finding(s) of the work only

RESPONSE: ADDITIONAL PARAGRAPH WAS ADDED

Any consideration given to the optimal use of CT services - CT scanners are large investments but may not be utilised constantly throughout the week (weekends, nights etc)

RESPONSE: TRUE! SUCH RECOMMENDATIONS CAN BE DERIVED. WE DID FOCUS ON THE ACCESS TO PUBLICLLY FUNDED FACILITIES WHICH ARE AVAILABLE 24/7 IN POLAND

It would have been useful to present data for inappropriate CT scans here as well for an overview on the populatoin level as to how many were estimated not to be appropriate. 

RESPONSE: ADDITIONAL INFORMATION WAS PROVIDED IN THE SUPPLEMENTARY TABLE

Formatting of tables needs to be improved

RESPONSE: DONE

 

Thank you very much for your suggestions

Author Response File: Author Response.docx