Artificial Intelligence in Urology—A Survey of Urology Healthcare Providers

Abstract

Background/Objectives: Artificial intelligence (AI) has been utilised in urological conditions such as urolithiasis, urogynaecology and uro-oncology. The aim of this study is to examine the attitudes and beliefs about AI technology amongst urology healthcare providers. Methods: A structured online questionnaire, created from a modified Delphi method with a panel of urologists and urology surgical trainees, was delivered through the Urological Asia Association’s annual congress. The questionnaire, with 25 items of mixed type responses (five-point Likert scale, nominal-polytomous and open-ended), acquired data regarding demographics, perception and attitudes towards general usage of AI in urological care. Results: A total of 464 respondents from 47 different countries were collected. The results showed that 83.4% of participants believed AI will improve efficiency and 18.8% believed they are knowledgeable in AI technology, with ordinal logistic regression showing both urology specialists and trainees are more likely to agree to these responses. Overall, 51.5% believed AI adoption will not replace clinical practice, and regression analysis found those with previous AI training are more likely to agree to this response. We found AI is commonly used in research, patient education and administrative tasks and identified key enablers as regulatory approval, AI clinical effectiveness and access to AI training. Conclusions: Overall attitudes and beliefs towards the use of AI in urology is positive and encouraging. AI training and education and regulatory reform needs to be addressed to allow integration of AI into clinical practice. A limitation of the study lies in its generalisability to global settings due to the demographics of the respondents.

1. Introduction

Artificial intelligence (AI) is the computational ability of a computer to simulate and carry out cognitive activities or behaviour that is comparable to humans; examples include machine learning and natural language processing [1,2,3]. Preventive medicine and AI outpatient clinics are projected to become more prevalent in future healthcare systems with the availability of patient-related big data [4].

The power of AI is in its capacity to handle and evaluate big data for predictive modelling, which can improve disease detection/diagnosis, patient outcomes and treatment planning [1,2,5]. For example, it is used in urological cancers with the likes of radiomics for the nuclear grading and categorisation of renal and bladder cancer, prediction of gleason score and magnetic resonance imaging diagnosis for prostate cancer [1,6]. Other urological applications are in the field of urolithiasis, paediatric urology, benign prostate hyperplasia, robotic surgery and surgical education [1,7,8,9]. The use of AI is also demonstrated in health information systems such as electronic medical records (EMRs) [2]. AI integration can allow healthcare providers to tailor interventions more effectively, ensure patients receive effective personalised care, achieve improved patient outcomes and a higher quality of life for individuals affected by urological diseases [5,8]. There is also AI usage in patient education. A previous study showed the use of Chat-GPT as an adjunct consultation tool for bladder cancer [10]. Machine learning models are used in research to perform complex data analysis on genetic and biomarker databases to predict/model oncological survival rates, disease recurrence and treatment response [11].

From the literature, we can see the rapid development and usage of AI in urological care. However, no previous studies have investigated attitudes and beliefs towards AI amongst urology healthcare providers. Hence, the aim of this study is to use an online questionnaire to examine the attitudes and beliefs on the application of AI in urology care and practice. By gaining insight from these stakeholders, we can identify and prioritise facilitators and challenges in AI implementation and provide guidance in prospective AI-related health policy formulation.

2. Materials and Methods

2.1. Survey Content/Design

A structured online questionnaire was created from a modified Delphi method through a panel of urologists and urology surgical trainees and fellows [12,13]. A total of three rounds of expert panel discussion and feedback were required to distil the survey from >40 questions to the final consensus of 25 questions, designed to assess understanding, attitudes, perceptions and potential barriers to AI use in urology. The first two discussion rounds were through online correspondence and the third round was through a face-to-face meeting. Expert consensus was not sought from AI experts; however, the panel consisted of leading experts in the field of urology, representing multiple national societies from Asia. This panel consisted of RRD, MZ, SD, MS, SE, NRZ, AHMIT, TF, ARAH, JYCT and IAT. These members are also part of the Urological Asia Association (UAA) Young Leadership forum research project. Expert consensus was used to create the questionnaire to achieve content and face validity, whereas unidimensional analysis was used to evaluate the reliability of the questionnaire.

The questionnaire consisted of 25 items with a mixture of responses: nominal-polytomous, a five-point Likert scale and open-ended. Participants could select more than one response for the nominal-polytomous questions. The questionnaire covered (i) demographics (five items), (ii) current practice with AI (four items), (iii) attitudes and beliefs towards AI in improving urology care (six items), (iv) barriers and enablers for AI implementation (two items), (v) attitudes and beliefs towards the implementation of AI in urology practice (seven items) and (vi) final comments to address any use, application, pitfalls or future of AI technology in urological healthcare (one item). It was delivered in written English and there was no time restriction placed on participants to complete the questionnaire. The full questionnaire is available in the Supplementary Material.

2.2. Data Collection/Analysis

The questionnaire was delivered through the Google form application and made available through the period of 26 August to 7 September 2024. This was a purposive sampling method as the questionnaire was distributed through blast emails and personal communication with the assistance of UAA. The inclusion criteria were being amongst those who received an online invitation to the survey, which were UAA members and participants/delegates who had registered for the 21st UAA Congress 2024 in Bali, Indonesia. There were no exclusion criteria to participate in the survey. The estimated number of delegates who participated in the congress was 2962, and this number was used to calculate the response rate to the survey.

Responses were collected through the Google form application and, to avoid any incomplete or missing data, the questionnaire required all questions to be answered prior to submission. To avoid duplicate results from the survey, we implemented measures on IP restriction, where a single internet protocol (IP) address can only complete the survey once. Survey responses were made available only to the investigators. All data were exported to Microsoft Excel for data analysis and visualisation. JASP 0.17.3 (Intel, University of Amsterdam, The Netherlands) statistical software was used for descriptive statistics, ordinal logistic regression analysis and unidimensional reliability analysis. The p-value 0.05 was selected as a significant level. The ordinal logistic regression analysis was used to identify whether demographic differences (i.e., age, occupation, gender, institution, location of practice and history of AI training—nominal/ordinal variables) can influence questionnaire responses (Likert scale responses—ordinal variables). The analysis was reported as an odds ratio (OR) converted from coefficient estimates (log-odds), with 95% confidence interval (CI). A unidimensional analysis was performed to assess the reliability of the Likert scale questions. The analysis was reported as Cronbach’s α and Guttman’s λ6 with 95% CI.

3. Results

A total of 464 participants, representing a 15.6% estimated response rate, completed the survey between 26 August and 7 September 2024. Demographic details are available in

Table 1. Demographic details of respondents.

Characteristics	n (N = 464)	%
Gender
Male	403	86.85
Female	61	13.15
Age group (years)
<30	90	19.4
30–39	215	46.34
40–49	104	22.41
50–59	36	7.76
≥60	19	4.09
Continent
Africa	1	0.22
Asia	389	83.84
Oceania	44	9.48
Europe	5	1.08
North America	20	4.31
South America	5	1.08
Institution
Teaching hospital/academic institution	312	67.24
Nonacademic public hospital	67	14.44
Private practice	37	7.97
Mixed (public/private)	48	10.34
Occupation
Urology Consultant/Specialist	272	58.62
Urology Nurse/Nursing Consultant	2	0.43
Urology Resident/Trainee/Registrar	151	32.54
General Practitioner	31	6.68
Hospital staff/JMO/Others	8	1.72

JMO: Junior Medical Officer.

. A reliability analysis showed the questionnaire to have a Cronbach’s α = 0.871 (95% CI 0.854–0.886), Guttman’s λ6 = 0.901 (95% CI 0.881–0.921) and the average inter-item correlation was 0.32 (95% CI 0.272–0.365). In the field analysis, the most common gender was male (n = 403, 86.9%) and the most represented age group was 30–39 years old (n = 215, 46.3%). The participants came from 47 countries across 6 continents. Indonesia had the highest number of respondents (n = 154, 33.2%), followed by Australia, Japan, Hong Kong, India, Taiwan and Malaysia. The majority of participants were urology consultants/specialists (n = 272, 58.6%), and a large portion were affiliated with a teaching hospital or academic institution (n = 312, 67.2%).

3.1. Actual and Perceived Use of AI in Urological Practice

Within the participants, 18.9% (n = 88) have never used AI, the majority being <50 years old (n = 72, 81.8%), and 71.6% were urology consultants/specialists (n = 63). Up to 77.2% (n = 358) of participants had no prior training in AI. However, 73.9% (n = 343) were willing to participate in further training and 24.8% (n = 115) were unsure.

Amongst those who have had AI experience, the most common application was research-related, including literature search (n = 214, 46.1%), literature summary (n = 187, 40.3%), presentation preparation (n = 141, 30.4%), manuscript writing (n = 113, 24.4%) and data analysis (n = 109, 23.5%). Other usage included patient education (n = 128, 27.6%), administrative tasks (n = 126, 27.2%), disease detection (n = 84, 18.1%), treatment planning (n = 65, 14%) and disease prognosis (n = 57, 12.3%) (Figure 1).

Interestingly, whilst only 27% of participants had used AI to perform administrative tasks, 58% of participants perceived/proposed that the primary role of AI in urology was improving administrative tasks (n = 270, 58%). Other proposed uses were in patient education (n = 247, 53.2%), research-related (n = 238, 51.3%) and patient satisfaction improvement (n = 156, 33.6%). Many participants also proposed the role of AI in the clinical context, in particular in disease diagnosis/detection (n = 236, 50.9%), predicting patient outcomes (n = 196, 42.2%) and treatment planning (n = 188, 40.5%) (Figure 2).

3.2. Attitudes and Beliefs Towards AI in Urological Care

More than 75% of participants were positive towards AI in the improvement of urology care; for example, 76.2% (n = 354) in patient outcome; 75% (n = 346) in patient satisfaction; 83.4% (n = 387) in efficiency of care and 87.5% (n = 406) in urology research. Regarding the confidence in AI accuracy, 53.9% (n = 250) of participants were confident and 37.1% (n = 172) were neutral about the idea. Lastly, 66.3% (n = 308) of participants believed patients will accept AI usage in urology care and 26.9% (n = 125) were neutral to the idea.

More than 80% of participants were positive in the adoption of AI technology, 85% (n = 394) supported adoption in research and 80% (n = 370) in clinical practice. Amongst participants, 86.2% (n = 400) were willing to use AI in the future with their clinical practice, amongst which 76.1% were <50 years old, 41.6% were urology specialists and 26.9% were urology trainees/residents (

Table 2. Tabulated breakdown of responses for willingness to use artificial intelligence (AI) in future practice, organised by age group and occupation, presented as n (%).

Response to Question “Willing to Use AI in Future Clinical Practice”	Age Groups, n (%)
	<50 years old	≥50 years old
Agree	353 (76.1)	47 (10.1)
Urology Consultant/Specialist	193 (41.6)	46 (9.9)
Urology Nurse/Nursing Consultant	1 (0.2)	1 (0.2)
Urology Resident/Trainee/Registrar	125 (26.9)	-
General Practitioner	27 (5.8)	-
Hospital staff/JMO/Others	7 (1.5)	-
Neutral	46 (9.9)	8 (1.7)
Urology Consultant/Specialist	21 (4.5)	8 (1.7)
Urology Resident/Trainee/Registrar	20 (4.3)	-
General Practitioner	4 (0.9)	-
Hospital staff/JMO/Others	1 (0.2)	-
Disagree	10 (2.2)	-
Urology Consultant/Specialist	4 (0.9)	-
Urology Resident/Trainee/Registrar	6 (1.3)	-
Total	409 (88.2)	55 (11.9)

JMO: Junior Medical Officer.

).

Interestingly, most participants did not believe the adoption of AI will replace clinical practice, with 51.5% (n = 239) in disagreement and 25.2% (n = 117) being neutral to the idea. Only 18.8% (n = 87) of participants felt they were knowledgeable about the current state of AI technology in urology, and this is consistent with the finding that only 20% had experience in the past with AI tools designed for urology.

3.3. Perceived Enablers and Barriers for AI Use

The most popular perceived enabler for future AI usage was regulatory approval (n = 261, 56.3%), followed by evidence of clinical effectiveness (n = 260, 56%), access to appropriate training (n = 253, 54.5%), cost-effectiveness (n = 243, 52.4%) and open-source for AI development tools (n = 195, 42%) (Figure 3).

There were various perceived barriers to AI implementation, with data privacy/security concerns (n = 284, 61.2%) being the most common concern. This was followed by concerns about AI accuracy (n = 256, 55.2%), ethical concerns (n = 200, 43.1%), healthcare integration (n = 198, 42.7%), medicolegal concerns (n = 180, 38.8%), a lack of understanding/confusion from colleagues (n = 176, 37.9%), high implementation costs (n = 157, 33.8%), a lack of understanding/confusion from patients (n = 155, 33.4%), resistance/distrust from colleagues (n = 116, 25%) and resistance/distrust from patients (n = 113, 24.4%) (Figure 4).

3.4. Ordinal Logistic Regression

Of the 14 items with ordinal scale responses, 3 questions demonstrated statistical significance with regression analysis. The question “I believe AI technology can improve the efficiency of urology practices” was more likely to be agreed with by urology specialists OR 2.94 (95% CI 1.12–7.72) and urology residents/registrars/trainees OR 2.97 (95% CI 1.24–7.12). Similarly, the question “I am very knowledgeable about the current state of AI technology in urology” was more likely to be agreed with by urology specialists OR 2.82 (95% CI 1.16–6.89) and urology residents/registrars/trainees OR 2.56 (95% CI 1.14–5.75), p < 0.001. Interestingly, participants who were less likely to agree were those with a history of AI training OR 0.32 (95% CI 0.21–0.49).

Finally, for the question “I believe AI technology will replace my urology practice in the future”, we found those with a history of AI training were less likely to agree OR 0.61 (95% CI 0.40–0.92) and only urology residents/registrars/trainees were more likely to agree OR 2.4 (95% CI 1.09–5.28).

4. Discussion

This is the first study to examine the attitudes and beliefs on the application of AI technology in urology healthcare providers, using an online questionnaire with an estimated response rate of 15.6%. This study is unique as it represented respondents from a large and diverse group of urological healthcare providers from over 40 different countries. The response rate is considered reasonable given the number of invitations sent and the specificity in our target audience [14]. The questionnaire demonstrated good reliability, as both Cronbach’s α and Guttman’s λ6 were above 0.8 for the Likert scale questions. We demonstrated that urology healthcare providers have a favourable view of integrating AI into their practice despite a lack of training and knowledge in the field. Notably, through regression analysis, we found that urology specialists and trainees were more likely to agree that AI can improve efficiency in clinical care and that they were knowledgeable about AI technology. The latter finding is interesting as 76.7% of participants declared having no history of prior AI training and regression analysis identified those with prior training were less likely to agree that they were knowledgeable in the AI field. This may reflect the Dunning–Kruger effect, where those with lower literacy in a subject matter report greater confidence in knowledge and vice versa for those with higher literacy [15]. Up to 51.5% did not believe AI will replace clinical practice and regression analysis showed that those who have received AI training were less likely to believe AI will replace clinical practice. These findings illustrated why it is important to receive education in AI technology. Reassuringly, our study showed that the majority of respondents are open to receiving AI education and incorporating AI into their future work, with a strong belief that AI will enhance urology care and be widely adopted in clinical settings.

The majority of participants proposed the primary role of AI in urology to be administrative-task-related, and this is consistent with the existing literature [16]. AI significantly enhances the efficiency of administrative tasks in healthcare settings, such as the automation of routine tasks (i.e., data entry, appointment scheduling and patient record management), improving operational efficiency and, thereby, reducing workload and improving productivity [16]. Whereas in a clinical context, the majority proposed that the role of AI is in disease diagnosis/detection. Consistent with the literature, a large portion of AI technology is focused on the accuracy and efficiency of diagnosis, especially for histological and imaging of urological tumours [7,16,17,18].

Interestingly, our study found that there is a discrepancy between the perceived use of AI compared to actual use. Despite the fact that 58% of participants proposed the role of AI to be in administration tasks, only 27% of participants have implemented it. Similarly, regarding clinical use, 50.9% proposed the use of AI to be in disease detection/diagnosis but only 18.1% of participants have practised this. This discrepancy can be attributed to several factors we have identified in this study. Firstly, many health professionals lack the resources or training to implement these technologies in their practice [19]. In our study, 76.7% of participants have no history of related training. Secondly, access to AI technologies can vary significantly between different healthcare settings and regions, with potential attributable factors being access to appropriate healthcare infrastructure, scarcity in resource allocation, institution resistance and economic constraints [19,20,21]. In this study, we have identified a major gap in AI exposure, with 18.9% of participants having never engaged in any AI usage and only 20% having any experience with urology-specific AI tools. Thirdly, regulatory hurdles and ethical concerns can also impact the actual use of AI. Healthcare professionals may be cautious about adopting AI due to concerns about data privacy, security and the potential for bias in AI algorithms [21,22,23,24,25]. A previous study from Eppler et al. evaluated the awareness and use of Chat-GPT of urologists worldwide, in which 47.7% of participants reported the use of Chat-GPT/large language models in their academic practice, with fewer using the technology in clinical practice (19.8%). More than half (62.2%) believed there are potential ethical concerns when using Chat-GPT for scientific or academic writing, and 53% reported that they have experienced limitations when using Chat-GPT in academic practice [26].

Consistent with the literature, we identified notable barriers to AI implementation being distrust, lack of knowledge and ethics, as well as data privacy and security being the most common perceived barriers [24,25]. The perceived barriers can often be attributed to the lack of opportunities for patients or clinicians to deepen their understanding of AI through engagement and communication with experts in the AI field [24,25]. Hence, collaborations between different stakeholders such as healthcare organisations, regulatory bodies, education institutions, technology companies, clinicians and patients are necessary processes to create acceptance and trust towards AI application in clinical practice [19,20,27,28,29]. For example, (i) AI technology companies can provide training and professional development programmes for healthcare professionals to build the necessary skills and confidence to use AI technologies effectively [24,25,27,28]; (ii) government/regulatory bodies can provide incentives for the adoption of AI technologies, such as tax breaks or funding for AI research and development [29]; (iii) multidisciplinary collaborated guidelines and standards for the use of AI in healthcare will therefore help address any perceived barriers or misconceptions and define ethical and safety parameters for AI use in clinical practice [24,25,30]. These solutions are consistent with enablers we identified in our study, with regulatory approval as the most commonly perceived enabler, along with other enablers such as AI-related training, clinical effectiveness and access to AI tools. There are reasonable grounds for a peak urology organisation such as UAA to be a trailblazer in promoting these enablers for AI implementation.

There are limitations to consider in this study. Firstly, the participant group is predominantly from Asia. This regional dominance may introduce bias and limit the generalisability of the findings to other global regions. The cultural, economic and healthcare differences across continents could influence the adoption and perception of AI in urology. Therefore, future studies should aim for a more balanced geographical representation to ensure a comprehensive understanding of AI’s role in urological practice worldwide. Potential distribution channels could be other urological conferences such as Société Internationale d’Urologie Journal (SIUJ), American Urological Association (AUA), Canadian Urological Association (CUA), European Association of Urology (EAU) and many more. Secondly, biases such as self-reflection could be considered. For example, those who are interested in AI technology may be more inclined to complete the survey. Lastly, the designs of the questionnaire could be optimised, with potential considerations in evaluating the length of the questionnaire, number of response options available for questions with nominal-polytomous responses and other non-English language options. Hence, a multilanguage option to widen the selection of participants and a review of the questionnaire’s validity across multiple continents may be considered for future study.

The future of AI in urology is promising and several implications are important for the functional integration of AI into clinical practice. Considerations such as the development of best practice guidelines/statements for AI implementation may be prioritised by urological societies. The introduction/development of AI training programmes for health providers, new urology specific AI tools and improved governance and interdisciplinary collaborations (between academic institutes, hospitals, AI technology firms, government and local community) are other important future considerations for AI integration.

5. Conclusions

In conclusion, the overall positive attitude towards the future application of AI in urology highlights the potential benefits it can bring to the field. The demand for AI in urology is evident, with participants expressing a strong belief in its ability to improve patient outcomes, satisfaction, efficiency of care and research. Regression analysis shows that those with a prior history of AI training were less likely to agree that AI will replace clinical practice, and up to half of the respondents have the same belief. In fact, the lack of AI education and training is one of the many barriers we have identified in this study. Additional barriers we have identified include ethical and security concerns, AI accuracy and economic constraints, whilst enablers include regulatory approval, clinical and cost effectiveness of AI technology. Lastly, we have also identified discrepancies in the proposed role and actual usage of AI in clinical practice. These differences are likely attributed to the presence of enablers and barriers to AI implementation.