Artificial Intelligence in Urology—A Survey of Urology Healthcare Providers

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I read the authors' work with interest. They have indeed done a large survey of opinions, limitations, experiences, and application possibilities of AI by urology specialists from 47 different countries. I congratulate them on this.
Here are some suggestions for corrections and clarifications:

 

1. In the abstract, it may be appropriate to briefly reference the main limitations to frame the representativeness of the sample (as explained in the lines 262-266).

 

2. The sentence "many believe AI adoption will not replace clinical practice" (line 43) is vague; it would be better to replace it with the percentages reported in the Results.
   
   
   
3. In the list of the numerous fields of application of AI in urology, there is also the bladder tumors detection, as well as its role in surgical education/training and AI-assisted surgery. It might be interesting to refer to these aspects.

 

4. The manuscript does not clarify how many specialists received the invitation to the questionnaire nor the response rate to it. The authors only report the final number of participants without providing an overview of the representativeness of the sample and any potential selection biases (lines 104-120).
   
   
   
5. In lines 86-87 reference is made to the creation of a questionnaire from modified Delphi method but no mention is made of tests/validation procedures.

 

6. In lines 89-90, reference is made to the creation of the questionnaire through a panel of urologists and urology surgical trainees. What criteria were used to select them (experience, publications, use of AI)?

 

7. It would be preferable to specify in the figure captions that the data shown refer to the sample. In this way, the graphs gain reading autonomy without ambiguity (line 192, 194, 196, 198).

 

8. In lines 114-117, statistical analyses are discussed. It would be interesting to better clarify the analyses performed and all the variables for which the Mann Whitney U test was used.

 

9. Lines 231-233/ 253-255: The statement is valid; it would be useful to support it better with bibliographic references. For example, the same criticisms have been highlighted by prof Cacciamani in the CANGARU project.

 

10. There are numerous studies in the literature that explore the opinions of physicians and students on AI and its role in clinical practice and education. It might be helpful to highlight what makes this study unique, such as its large sample size and the inclusion of participants from multiple countries, even though the representativeness is not entirely homogeneous.

Author Response

Thank you for taking the time and effort in providing the below responses and feedbacks. We have taken all these feedbacks on board and endeavour to address as much as we can. The changes in the manuscript are highlighted (yellow) as requested by the editor. Additionally, we have made short succinct point by point responses to each feedback and also provided line numbers to assist with the review. We hope the updated manuscript provides more robust results and explanation. We look forward to your review and response.

Comments 1. In the abstract, it may be appropriate to briefly reference the main limitations to frame the representativeness of the sample (as explained in the lines 262-266).

• Response: Changes made, see lines 48-49.

Comments 2. The sentence "many believe AI adoption will not replace clinical practice" (line 43) is vague; it would be better to replace it with the percentages reported in the Results.

• Response: Changes made, see lines 39-42, we have added specific numbers and ordinal regression analysis findings as well.

Comments 3. In the list of the numerous fields of application of AI in urology, there is also the bladder tumors detection, as well as its role in surgical education/training and AI-assisted surgery. It might be interesting to refer to these aspects.

• Response: Added additional references to reflect AI usage in the above areas; reference 6 & 9, by Pintar and Knudsen respectively).

Comments 4. The manuscript does not clarify how many specialists received the invitation to the questionnaire nor the response rate to it. The authors only report the final number of participants without providing an overview of the representativeness of the sample and any potential selection biases (lines 104-120).

• Response: Changes made, See lines 118-119, we have added details on the estimated number of individuals who have received email invitation to the survey. Unfortunately, we may only provide information on who has completed the survey. We are unable to provide information on how many specialist or trainees have received the invitation.

Comments 5. In lines 86-87 reference is made to the creation of a questionnaire from modified Delphi method but no mention is made of tests/validation procedures.

• Response: Changes made in lines 87-93. Additional details regarding validity and reliability assessment are seen in line 95-97. The findings of the reliability are addressed in line 140-142.

Comments 6. In lines 89-90, reference is made to the creation of the questionnaire through a panel of urologists and urology surgical trainees. What criteria were used to select them (experience, publications, use of AI)?

• Response: Details added in lines 91-95 to provide some clarify on who these members are on the panel.

Comment 7. It would be preferable to specify in the figure captions that the data shown refer to the sample. In this way, the graphs gain reading autonomy without ambiguity (line 192, 194, 196, 198).

• response: Modifications made to captions to avoid ambiguity.

Comments 8. In lines 114-117, statistical analyses are discussed. It would be interesting to better clarify the analyses performed and all the variables for which the Mann Whitney U test was used.

• Response See lines 126-133, we have changed to use ordinal logistic regression analysis as it is much more robust to assess our data set.

Comments 9. Lines 231-233/ 253-255: The statement is valid; it would be useful to support it better with bibliographic references. For example, the same criticisms have been highlighted by prof Cacciamani in the CANGARU project.

• response: Added reference 21 and 30, by Yu and Cacci respectively.

Comments 10. There are numerous studies in the literature that explore the opinions of physicians and students on AI and its role in clinical practice and education. It might be helpful to highlight what makes this study unique, such as its large sample size and the inclusion of participants from multiple countries, even though the representativeness is not entirely homogeneous.

• Response: Added lines 240-242 to help address why this study is unique.

Reviewer 2 Report

Comments and Suggestions for Authors

A very well written paper that evaluates the usefulness of AI in Urology as percieved by practitioners în  this speciality. 

Introduction is concise and is a very good induction for the readers in the theme that the paper approaches.

Material and Methods js very explanatory. The authors present how the questionnaire was create, how the finzl 25 questions were selected,  and the way it was disseminated to a large number of participants in multiple countries, with the assistance of the Urologic Asia Association. Also this chapter included the means used for data analysis.

Results chapter is extensive and presents a wide range of data acquired using the questionnaire. This included demographic data of the 464 participants to the survey, data about the actual and percieved usage of AI in Urology practice, attitude and belief of AI usage in urological care, and the percieved enabler and barriers of AI use. All data was doubled by easy to understand tables and schemes.

Discussions chapter is clear and in line with the theme of the paper. 

Conclusions are concise and they are supported by the results.

This is an interesting paper as it shows that urologists are open to the use of AI in their practice, and this paper shows that the common consensus is that AI will bring improvements in all levels of Urology care. We are now in full AI boom and as the technology progresses, we need to find ways to integrate it in practice in order to obtain better results for our patients, but also to regulate and better train specialists in this field. I recommend for publication.

Author Response

Thank you for taking the time and effort in providing the below response and feedback, much is appreciated. 

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript presents a relevant and timely survey on the attitudes and beliefs of urology healthcare providers regarding artificial intelligence (AI) in clinical and research settings. The topic is highly significant given AI’s growing integration into urology, and the study provides valuable insights from a large and diverse respondent group. The manuscript is generally well-structured with a comprehensive discussion, but several areas require clarification and improvement.

1. One major concern is the survey methodology and potential sampling bias. Since the survey was primarily distributed through the Urological Association of Asia (UAA), most responses come from Asia, which limits the generalizability of the findings. The authors should explicitly acknowledge this limitation in the discussion and suggest strategies for broader recruitment in future studies.
2. Additionally, the inclusion and exclusion criteria for respondents are not clearly defined. It is unclear whether only urology specialists were included or if other healthcare professionals (e.g., nurses, medical students) also participated.
3. Another concern lies in the interpretation of data and statistical analysis. The finding that 86.2% of participants are willing to use AI in future practice is significant, but there is no breakdown of whether those with prior AI exposure are more accepting of its use. Conducting subgroup analyses would provide greater insight into whether familiarity with AI influences its acceptance. Furthermore, the authors utilize the Mann-Whitney U test to compare age groups, but no justification is provided for why this test was chosen over other potential statistical methods. A clearer rationale for the statistical tests used, along with the consideration of alternative methods such as logistic regression, would enhance the robustness of the analysis.
4. Regarding AI applications in urology, the survey results reveal discrepancies between perceived and actual AI usage. For instance, 58% of respondents believe AI should be used for administrative tasks, yet only 27% report actually using it for this purpose. While the authors suggest that lack of training is a major barrier, this claim is not fully substantiated. It would be beneficial to discuss additional barriers such as institutional resistance, healthcare infrastructure limitations, and economic constraints that might also contribute to these discrepancies. 

 

Author Response

Thank you for taking the time and effort in providing the below response and feedback. We have taken all these feedbacks on board and endeavour to address as much as we can. The changes in the manuscript are highlighted (yellow) as requested by the editor. Additionally, we have made short succinct point by point responses to each feedback and also provided line numbers to assist with the review. We hope the updated manuscript provides more robust results and explanation. We look forward to your review and response.

 

comments 1. One major concern is the survey methodology and potential sampling bias. Since the survey was primarily distributed through the Urological Association of Asia (UAA), most responses come from Asia, which limits the generalizability of the findings. The authors should explicitly acknowledge this limitation in the discussion and suggest strategies for broader recruitment in future studies.

• Response: Agree, these limitations are discussed in lines 316-322. We have added a sentence on strategies for future recruitment, in lines 322-323.

Comments 2: Additionally, the inclusion and exclusion criteria for respondents are not clearly defined. It is unclear whether only urology specialists were included or if other healthcare professionals (e.g., nurses, medical students) also participated.

• response: Agree, clarification has made through lines 114-119. We have also added response rate, reliability analysis and further clarification in the methods sections.

Comments 3: Another concern lies in the interpretation of data and statistical analysis. The finding that 86.2% of participants are willing to use AI in future practice is significant, but there is no breakdown of whether those with prior AI exposure are more accepting of its use. Conducting subgroup analyses would provide greater insight into whether familiarity with AI influences its acceptance. Furthermore, the authors utilize the Mann-Whitney U test to compare age groups, but no justification is provided for why this test was chosen over other potential statistical methods. A clearer rationale for the statistical tests used, along with the consideration of alternative methods such as logistic regression, would enhance the robustness of the analysis.

• Response: Agree, we have decided to use the ordinal logistic regression analysis to provide a more robust analysis. The explanation is seen in lines 129-134. Using the regression analysis, we were able to account for demographics variables such as gender, age, occupation, location of practise, institution and previous AI training history and whether it could influence questionnaire response.

Comments 4: Regarding AI applications in urology, the survey results reveal discrepancies between perceived and actual AI usage. For instance, 58% of respondents believe AI should be used for administrative tasks, yet only 27% report actually using it for this purpose. While the authors suggest that lack of training is a major barrier, this claim is not fully substantiated. It would be beneficial to discuss additional barriers such as institutional resistance, healthcare infrastructure limitations, and economic constraints that might also contribute to these discrepancies. 

• Response: Agree, we have mentioned these additional barriers in the discussion please see lines 276-283, we have made slight changes. Lines 284-286 also addresses additional barriers. 

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript addresses a relevant and important topic concerning the application of artificial intelligence in urology. However, in order to enhance its scientific rigor and clarity, it is recommended to consider the following comments, which cover methodological, structural, and content-related aspects. These suggestions aim to strengthen the overall argumentation, improve the justification of conclusions, and ensure greater reliability and practical applicability of the findings.

 

1. Although the study focuses on the application of artificial intelligence in urology, the abstract lacks even a minimal specification of AI methods. It is not indicated what type of AI solutions are being evaluated.
2. The phrase “AI has many uses in medicine including but not limited to diagnostics…” is overly generic. At least one specific example should be provided to avoid a vague impression.
3. The formulation “participants see value in improving urological care” is too general. It would be useful to briefly indicate in which areas this assessment is particularly evident or which challenges are considered most relevant.
4. The introduction mentions terms such as “machine learning” and “natural language processing,” but they are not explained further or linked to specific examples in urology. It is also not indicated which machine learning algorithms are most commonly used and why they are appropriate for specific situations.
5. Instead of presenting scattered examples, the overview should be structured according to the areas of AI application.
6. Although many references are provided, they mostly illustrate the use of AI rather than analyze methodological advantages, challenges, or limitations. A more critical evaluation should be included—for example, which algorithms have proven to be the most effective, what problems arise with data bias or model explainability.
7. Phrases about market value growth or a “paradigm shift” resemble business or popular media style rather than scientific language. Such statements should either be scientifically substantiated or phrased more cautiously.
8. It is stated that attitudes toward AI in urology have not been studied so far, but there is no explanation why this aspect is important from a scientific or clinical perspective. The manuscript lacks reasoning on how these attitudes may affect AI adoption, clinical practice, or health policy.
9. The use of a “modified Delphi method” is mentioned, but it is not explained in detail how exactly this method was applied: what were the criteria for question selection, how was feedback implemented, was expert consensus sought, and how was the reliability of the questions assessed.
10. There is no qualitative or quantitative assessment of the questionnaire’s validity and reliability (e.g., Cronbach’s α or interrater agreement). This raises concerns regarding the methodological soundness of the instrument.
11. Although it is stated that participants were attendees of the 21st UAA Congress, the sampling method remains unclear (random, purposive, or voluntary sampling?). This is important when evaluating the representativeness of the results.
12. There is no information on how many invitations were sent and what the response rate was—an important methodological indicator in survey-based research.
13. Statistical processing steps are only briefly mentioned. This section should be expanded to provide a clear statistical analysis plan: which variables were analyzed, what type of data was processed, which indicators were used.
14. Although the Mann-Whitney U test is mentioned in the results section, there is a lack of information on statistical test outcomes: p-values, effect sizes, or confidence intervals.
15. Currently, only descriptive data are presented without deeper analysis. It is recommended to additionally perform regression or correlation analysis, especially to understand what determines a positive attitude toward AI or the willingness to adopt it in practice.
16. The results are presented almost exclusively as raw numbers. It would be appropriate to begin interpreting trends—for example, why younger professionals are more supportive of AI adoption, or why administrative benefits are considered most significant.
17. It is suggested to organize the discussion section into subsections based on key themes. This would enhance readability and logical structure.
18. Some ideas—particularly regarding the efficiency of administrative tasks or the gap between theoretical perception and practical use—are repeated several times. These could be shortened or combined into one paragraph.
19. Some statements remain overly general. For instance, “AI technologies can vary significantly between different healthcare settings” would benefit from a more specific explanation of what factors hinder implementation (e.g., technical infrastructure, cost, staff shortages, etc.).
20. Instead of “many health professionals lack the resources,” it would be better to specify what kind of resources are lacking (technical, human, financial?).
21. In addition to regional distribution, it would be useful to briefly discuss the possible self-selection bias—those more interested in AI were likely more inclined to respond to the survey.
22. It would also be beneficial to briefly assess how the questionnaire design may have influenced the results (e.g., clarity of question phrasing, limitations in response options).
23. Instead of general recommendations, more specific actions should be proposed—what kind of training is needed, which AI tools should be prioritized for development.
24. It is recommended that the conclusions more clearly specify which concrete research findings they are summarizing. For example, figures or trends could be provided.
25. Instead of general phrases like “clear demand for AI” or “universal barriers,” it would be more appropriate to specify concrete needs (e.g., lack of training, gaps in regulation, shortage of technological resources) and specific barriers (e.g., data security, ethical concerns, disparities in AI tool accessibility).

Author Response

Thank you for taking the time and effort in providing the below response and feedback. We have taken all these feedbacks on board and endeavour to address as much as we can. The changes in the manuscript are highlighted (yellow) as requested by the editor. Additionally, we have made short succinct point by point responses to each feedback and also provided line numbers to assist with the review. We hope the updated manuscript provides more robust results and explanation. We look forward to your review and response.

Comment 1: Although the study focuses on the application of artificial intelligence in urology, the abstract lacks even a minimal specification of AI methods. It is not indicated what type of AI solutions are being evaluated.

• Response: Our manuscript aims to evaluate the general attitudes and beliefs of AI application in urology. The focus on evaluating examples of AI technologies applied in urology is unfortunately beyond the scope of our paper. Hence, specification in AI methods are not delved into.

Comment 2: The phrase “AI has many uses in medicine including but not limited to diagnostics…” is overly generic. At least one specific example should be provided to avoid a vague impression.

• Response: We have made changes to the abstract to make it more robust and precise, please see highlighted changes.

Comment 3: The formulation “participants see value in improving urological care” is too general. It would be useful to briefly indicate in which areas this assessment is particularly evident or which challenges are considered most relevant.

• response: Similar to bullet point 2, we have made changes to the abstract, please see highlighted changes

Comment 4: The introduction mentions terms such as “machine learning” and “natural language processing,” but they are not explained further or linked to specific examples in urology. It is also not indicated which machine learning algorithms are most commonly used and why they are appropriate for specific situations.

• Response: Our response would be similar to bullet point 1, we were aiming for an introduction on the AI topic with broad strokes.

Comment 5: Instead of presenting scattered examples, the overview should be structured according to the areas of AI application.

• Response: Similar to bullet point 4, we were aiming for broad strokes to give wide examples of how AI is used in urology.

Comment 6: Although many references are provided, they mostly illustrate the use of AI rather than analyze methodological advantages, challenges, or limitations. A more critical evaluation should be included—for example, which algorithms have proven to be the most effective, what problems arise with data bias or model explainability.

• Response: Unfortunately, we cannot address this question as this goes beyond the scope of this paper, as this paper focuses on the general attitudes and beliefs of AI usage in Urology practise.

Comment 7: Phrases about market value growth or a “paradigm shift” resemble business or popular media style rather than scientific language. Such statements should either be scientifically substantiated or phrased more cautiously.

• Response: Agree, this statement is now removed from the introduction.

Comment 8: It is stated that attitudes toward AI in urology have not been studied so far, but there is no explanation why this aspect is important from a scientific or clinical perspective. The manuscript lacks reasoning on how these attitudes may affect AI adoption, clinical practice, or health policy.

• Response: Agree, we have added details in lines 79-81 and lines 240-242.

Comment 9: The use of a “modified Delphi method” is mentioned, but it is not explained in detail how exactly this method was applied: what were the criteria for question selection, how was feedback implemented, was expert consensus sought, and how was the reliability of the questions assessed.

• Response: Agree, to address these points we have made clarification in lines 87-97, lines 134-135, lines 140-142. The creation and validity are through expert consensus and we have performed unidimensional analysis for reliability assessment (using both Cronbach and Guttmann).

Comment 10: There is no qualitative or quantitative assessment of the questionnaire’s validity and reliability (e.g., Cronbach’s α or interrater agreement). This raises concerns regarding the methodological soundness of the instrument.

• Response: Agree, this point is related to bullet point 9, we have provided further clarification in lines 95-97, lines 140-142.

Comment 11: Although it is stated that participants were attendees of the 21st UAA Congress, the sampling method remains unclear (random, purposive, or voluntary sampling?). This is important when evaluating the representativeness of the results.

• Response: Agree, this was purposive sampling, we have provided clarification in lines 112-113.

Comment 12: There is no information on how many invitations were sent and what the response rate was—an important methodological indicator in survey-based research.

• Response: Agree, to provide clarity to the method section we have added details in lines 112-119, 138, and lines 240. The response rate is estimated to be 15.6%. the invitations were made to those delegates of the 21^st UAA congress and UAA members.

Comment 13: Statistical processing steps are only briefly mentioned. This section should be expanded to provide a clear statistical analysis plan: which variables were analyzed, what type of data was processed, which indicators were used.

• response: Agree, to provide more robust statistical analysis we have used the ordinal regression analysis, please see lines 126-133.

Comment 14: Although the Mann-Whitney U test is mentioned in the results section, there is a lack of information on statistical test outcomes: p-values, effect sizes, or confidence intervals.

• response: We have removed this test entirely, as ordinal regression analysis would be more robust.

Comment 15: Currently, only descriptive data are presented without deeper analysis. It is recommended to additionally perform regression or correlation analysis, especially to understand what determines a positive attitude toward AI or the willingness to adopt it in practice.

• Response: Agree, we have performed the ordinal regression analysis to focus on whether age, gender, occupation, locations of practise, institution, and previous history of AI training can influence questionnaire response related to attitudes/beliefs; see lines 212-225.

Comment 16: The results are presented almost exclusively as raw numbers. It would be appropriate to begin interpreting trends—for example, why younger professionals are more supportive of AI adoption, or why administrative benefits are considered most significant.

• response: Although we cannot provide analysis on the trend, we did endeavour to provide more robust analysis through logistic regression. See bullet point 15.

Comment 17: It is suggested to organize the discussion section into subsections based on key themes. This would enhance readability and logical structure.

• response: We have tried to organised the discussion in the following order: general results, perceived/proposed AI usage, inspection on the enablers/barriers to AI usage, limitation, and finally conclusion.

Comment 18: Some ideas—particularly regarding the efficiency of administrative tasks or the gap between theoretical perception and practical use—are repeated several times. These could be shortened or combined into one paragraph.

• response: We have removed the redundant statements that was previously in lines 249-250.

Comment 19: Some statements remain overly general. For instance, “AI technologies can vary significantly between different healthcare settings” would benefit from a more specific explanation of what factors hinder implementation (e.g., technical infrastructure, cost, staff shortages, etc.).

• Response: Agree, we have made changes, see lines 278-281.

Comment 20: Instead of “many health professionals lack the resources,” it would be better to specify what kind of resources are lacking (technical, human, financial?).

• response: This is similar to bullet point 19, we will address the same .

Comment 21: In addition to regional distribution, it would be useful to briefly discuss the possible self-selection bias—those more interested in AI were likely more inclined to respond to the survey.

• response: Agree, we have a made statement, added in lines 323-325.

Comment 22: It would also be beneficial to briefly assess how the questionnaire design may have influenced the results (e.g., clarity of question phrasing, limitations in response options).

• Response: This was briefly discussed as a limitation in the discussion, seen in lines 325-330.

Comment 23: Instead of general recommendations, more specific actions should be proposed—what kind of training is needed, which AI tools should be prioritized for development.

• Response: Agree, we have previously mentioned actions such as those in lines 302-309. Although this is a good feedback regarding AI tool, it goes beyond the scope of our paper to identify what sort of AI tools to be prioritise. Aforementioned, our paper is meant to illustrate a general pattern in attitudes and belief of AI usage in urology care.

Comment 24: It is recommended that the conclusions more clearly specify which concrete research findings they are summarizing. For example, figures or trends could be provided.

• Response: Agree, we have made the changes seen in lines 342-353.

Comment 25: Instead of general phrases like “clear demand for AI” or “universal barriers,” it would be more appropriate to specify concrete needs (e.g., lack of training, gaps in regulation, shortage of technological resources) and specific barriers (e.g., data security, ethical concerns, disparities in AI tool accessibility).

• Response: We presume this comment is for the conclusion section, we have made changes seen in lines 345-353.