Exploring clinical specialists’ perspectives on the future role of AI: evaluating replacement perceptions, benefits, and drawbacks

Artificial intelligence (AI) generally applies to computational technologies that emulate mechanisms assisted by human intelligence, such as thought, deep learning, adaptation, engagement, and sensory understanding [1, 2]. Some devices can execute a role that typically involves human interpretation and decision-making [3, 4]. These techniques have an interdisciplinary approach and can be applied to different fields, such as medicine and health. AI has been involved in medicine since as early as the 1950s when physicians made the first attempts to improve their diagnoses using computer-aided programs [5, 6]. Interest and advances in medical AI applications have surged in recent years due to the substantially enhanced computing power of modern computers and the vast amount of digital data available for collection and utilization [7]. AI is gradually changing medical practice. Several AI applications in medicine can be used in a variety of medical fields, such as clinical, diagnostic, rehabilitative, surgical, and predictive practices. Another critical area of medicine where AI is making an impact is clinical decision-making and disease diagnosis. AI technologies can ingest, analyze, and report large volumes of data across different modalities to detect disease and guide clinical decisions [8‐10].

It is possible that the widespread use of advanced technologies and the imminent challenges of AI could pose a threat to physicians’ jobs and experts feel that the role of physicians will customarily become a joint team effort between physicians and machines [4]. In a survey involving 791 psychiatrists from advanced countries, 83% of participants expressed their opinions about the future capability of technology to provide care, while they also showed concerns about the potential replacement of their roles [11]. Conversely, a study with radiologists showed that 77% displayed favorable attitudes towards AI implementation and 89% exhibited a lack of concern regarding job displacement [12]. Neurosurgeons, too, embraced AI, with 60% utilizing it for outcome prediction, and a staggering 90% of physicians across diverse specialties in Germany anticipated AI integration in the future of medicine [11, 12]. A global survey of pathologists revealed widespread acceptance of AI, as only 17.6% showed their apprehensions about job security [13]. This was also observed in a Saudi Arabian study, where physicians, nurses, and technologists expressed notable concerns about their future employability [14]. The optimism persisted in a study among neurosurgeons, reinforcing the anticipation of AI as a crucial component in the future of medicine [15]. Concerning patient perspectives on AI in clinical practice, a prevailing preference for human physicians over AI was observed [16‐19]. This inclination may be attributed to the proven value of AI algorithms in aiding radiologists to detect abnormalities in medical images, categorize conditions, hypothesize about underlying health issues, suggest suitable procedures, and interpret results [20, 21]. The capabilities of AI extended to dermatology, where it demonstrated robust diagnostic imaging and assessments for various dermatological pathologies. In ophthalmology, AI emerged as a valuable tool for identifying and understanding eye problems, aiding doctors in recognizing issues within the retina and other ocular structures [22‐25]. The exploration of physicians’ perspectives on AI in the Arab world revealed optimism among doctors in Oman, where a study with 300 doctors and 750 medical students unveiled a positive outlook on AI’s future role, coupled with minimal concerns about job displacement [26]. A similar study in the United Arab Emirates involving 119 individuals skilled in conducting X-rays and 34 specialized X-ray doctors demonstrated that 86% did not envision AI playing a role in X-ray procedures. Furthermore, 64% expressed a lack of concern about AI negatively impacting their careers in this specific domain [27].

Pakistan, with a population exceeding 220 million people, is facing distinctive healthcare challenges arising from its diverse demographics, varying economic conditions, and geographical disparities. Ranked as the fifth-most populous country globally, Pakistan exhibits a populace dispersed across urban and rural landscapes, with healthcare infrastructure discrepancies evident between urban centers and remote rural areas. Besides these challenges, the country has initiated an AI program under a presidential initiative, signifying a forward-looking approach [28]. However, the implementation of AI in the health sector encounters several hurdles, especially in medical healthcare. Recent advancements in AI application in Pakistan prompt attention to the need for rigorous evaluation, particularly through randomized controlled trials (RCTs) to establish effectiveness across diverse healthcare settings. A systematic review uncovered a deficiency of RCTs involving AI within the Pakistani healthcare system [29]. Furthermore, the absence of a published dataset for retinal images poses a notable challenge for doctors seeking to train AI models, limiting the integration of AI into routine medical practices [30].

Against this backdrop, to contribute to the existing literature on the impact of AI in Pakistan, this study aims to delve into the perspectives of clinical experts situated in the southern region of Punjab. This region, like numerous others in Pakistan, grapples with challenges related to healthcare infrastructure [31].

The study seeks to explore how AI might influence the daily routines of healthcare professionals in this region, examining the anticipated benefits and potential implications for patient confidentiality and job stability. Moreover, the findings of this study could have broader implications for healthcare systems in developing countries facing similar challenges. By examining the perceived benefits and drawbacks of AI adoption from the perspective of healthcare professionals in Pakistan, this study contributes to the global discourse on the role of AI in healthcare and provides valuable insights into AI strategies to meet the specific needs of diverse healthcare contexts. The impact of AI on healthcare in Pakistan is greater than in other countries due to its unique healthcare challenges, recent initiatives in AI implementation, and the potential for transformative change in addressing longstanding healthcare disparities. Understanding the perspectives of clinical experts in regions like southern Punjab not only informs local strategies but also contributes to broader discussions on AI adoption in healthcare systems worldwide.

By focusing on the southern region of Punjab, the research aims to find the answers to the following questions;

This study is structured as follows: The second section presents the materials and methods, encompassing the study design, participant demographics, and ethical considerations. The third section details the data analysis, followed by the presentation of results regarding AI knowledge as a replacement, its benefits, drawbacks, and acceptance. Finally, the study ends with a discussion and conclusion section, where it reflects on its findings and acknowledges the possible limitations of the study.

The study involved an extensive data collection process from a diverse group comprising 140 participants, consisting of clinical specialists in their fields with varying ages and experience levels. Among the respondents, there were more male participants (80 individuals, representing 57.1%) compared to female participants 60(42.9%). The age distribution of the participants was also analyzed, with 68 (48.6%) being above the age of 40, 42 (30%) of study participants falling below the age of 35, and 30 (21.4%) belonging to the age range of 36 to 40. Furthermore, the participants’ experience levels demonstrated considerable heterogeneity, as 59(42.1%) reported having less than six years of experience, 55(39.3%) had more than 11 years of experience, and 26(18.6%) fell within the 7 to 11 years experience range. The study encompassed multiple departments, with Surgery representing 24(17.1%), Pathology accounting for 31(22.1%), Radiology including 35(25%), Gynecology comprising 33(23.6%), and Pediatrics comprising 17(12.1%) of the participants. Surprisingly, only 34(24.3%) expressed a belief that hospitals were ready to embrace AI-driven diagnostic tools as a new tool in the field of diagnostics, while a substantial majority of 81(57.9%) displayed doubt about it, stating that hospitals were not adequately prepared for AI implementation as diagnostics. A smaller but still noteworthy proportion of 25(17.9%) were cautiously optimistic, asserting that hospitals were partially ready for AI integration. Table 1 provides a comprehensive breakdown of the socio-demographic characteristics of the participants, as well as their knowledge and acceptance levels concerning AI (Table 1).

In terms of participants’ perception of AI’s role in the future, a substantial 71(50.7%) of respondents believed that AI held a vital position in shaping the future of healthcare. Conversely, 41(29.3%) of participants expressed doubts about AI being the future, while 28(20.0%) thought AI would play a role to some extent. The survey also assessed the participants’ knowledge of AI, with the results showcasing diverse levels of understanding. A considerable proportion, 54(37.1%), displayed above-average knowledge, followed by 41(29.3%) with average knowledge, and 24(16.4%) with excellent knowledge about AI. On the other hand, 21(15.0%) of participants had a poor understanding of AI.

This sub-scale consisted of nine questions consisting of AI’s different roles as replacements of documentation, clinical care to patients, suggesting medication, conducting a physical examination, diagnosis, and patient history (for data analysis, Possible = 1, Not Possible = 2, Maybe Possible = 3) AI’s most prominent role in documenting patient information, including health records and patient history, is likely to be in Surgery 12(8.6%) and Pathology 9(6.4%), while Radiology and Gynecology show lower percentages at 4(2.9%) and Pediatrics at 3(2.1%) (p = 0.018**). However, in terms of providing good patient care in the future, the utilization of AI appears relatively low across all specialties, ranging from 2(1.4%) to 7(5.0%). Notably, Radiology stands out as the specialty where AI’s possible involvement in suggesting personalized medication based on patient history is most prevalent, with a significant percentage of 13.6% (p = 0.011).

When evaluating patients for treatments using AI, the task seems to be evenly distributed across all specialties, with no significant differences observed (p-value = 0.949). Similarly, AI’s possible role in examining patient information to establish prognoses shows consistent percentages across specialties, with no significant differences (p-value = 0.256). A noteworthy finding is that it is possible that in the future AI can be frequently employed to detect instances of self-harmful behavior in patient information, particularly in Surgery with the highest percentage at 10.7%. The p-value of 0.001*** highlights a highly significant association between AI use and the detection of self-harmful behavior. Likewise, AI’s “possible” use to discern potential indications of criminal behavior or assault in patient information varies significantly among specialties (p = 0.001***). Conducting a “possible” comprehensive physical examination with a detailed mental status assessment using AI is most prevalent according to the Surgery and Gynecology specialists, with percentages of 10.7% and 3.6%, respectively (p = 0.000**). Finally, AI’s “possible” involvement in conducting patient interviews across diverse settings to gather comprehensive medical histories showed significantly varying percentages across specialties, with the highest in Surgery at 10.7% (p = 0.042) (Table 2; Fig. 1).

This sub-scale consisted of eight questions about AI benefits and drawbacks including the concerns about violation of patients’ privacy, unemployment issues, reducing the work and paper burden, help for specialists, and part of the curriculum (For the data analysis, Yes = 1, No = 2, Maybe = 3). The study revealed that 42(30.0%) of them expressed the view that AI can be a violation of patients’ privacy. Looking at the responses from different departments, gynecology had the highest number of specialists who agreed with this statement, with 13(9.3%) respondents. On the other hand, 62(44.3%) specialists disagreed with the statement that AI can violate patients’ privacy. The majority of respondents in each department fell under this category, with Pathology having the highest number of specialists, 16(11.4%), who disagreed, followed by Radiology with 15 (10.7%) and Gynecology with 14 (10.0%) respondents. 36(25.7%), were unsure and responded with “maybe.” (Table 3; Fig. 2).

Additionally, 70(50%) of the clinical specialists expressed the opinion that AI can cause unemployment in the healthcare sector. Among the departments, Radiology had the highest number of specialists who agreed with this statement, with 19(13.6%) respondents. This was followed by Pathology with 17(12.1%) specialists agreeing, and Surgery with 14 (10%) specialists in agreement. Gynecology and Pediatrics departments had 15 (10.7%) and 5 (3.6%) specialists, respectively, who believed that AI could lead to unemployment in the field.

51(36.4%) specialists disagreed with the statement that AI can cause unemployment. Among the study participants, 119(85%) expressed the view that the computerization of AI of healthcare data by AI of patients can indeed offer an opportunity to improve patient care. Among the departments, Gynecology had the highest number of specialists who agreed with this statement, with 29 (20.7%) respondents. In contrast, 21(15%) specialists disagreed with the statement that computerization can improve patient care. Moreover, 106(75.7%) expressed the view that the utilization of AI has indeed substantially minimized the paperwork burden of keeping patient records. While 101(72.1%) expressed the view that AI can indeed be helpful for disease diagnosis accurately. Among the clinical specialists, radiologists specialists showed the highest numbers who agreed with this statement, with 25(17.9%) respondents. This was followed by pathology and Gynecology, with 23(16.4%) and 24(17.1%) respectively. 39(27.8%) specialists disagreed with the statement that AI can be helpful for disease diagnosis accurately. Out of a total of 140 participants, 108(77.1%) expressed the belief that AI can indeed be more accurate than doctors. Among the departments, 27(19.3%) of Gynecologists showed the highest percentage of responses among other specialties who agreed with this statement. 108(77.1%) expressed the view that AI should indeed be part of the curriculum of medical students. Among the departments, radiologists showed the highest percentage of agreement with 30(21.4%) respondents.

The study provides valuable insights into the perspectives of different clinical specialties regarding the potential roles of artificial intelligence (AI) in healthcare. Our study highlighted that 81(57.9%) of the participants hold the belief that hospitals are currently unprepared to incorporate AI into the future of healthcare which aligns with prior studies [9] that reported 83% of participants expressing doubts about AI’s ability to deliver empathetic patient care in the future. Moreover, our study examined participants’ grasp of AI, revealing varying levels of understanding among clinical specialists. Specifically, 54(37.1%) of the participants demonstrated above-average AI knowledge, a trend in line with recent studies [32‐52]. In contrast, 41 (29.3%) of the participants exhibited an average level of AI knowledge, indicating a moderate familiarity with AI concepts. This suggests the potential value of further education and exploration of AI applications.

Concerning AI’s roles within specific specialties, the study emphasized that AI is anticipated to have a notable presence in areas such as Surgery and Pathology, particularly in the realm of documenting patient information such as health records and medical history. An exception to this pattern was observed in radiology 19(13.6%), where AI’s potential for suggesting personalized medication based on patient history stood out. Our findings are supported by prior studies [27] which indicated that 14% of radiologists agreed with the idea of AI’s role in radiology.

The integration of AI in healthcare raised significant privacy concerns among specialists, emerging as a key focus of the study. In our study, approximately 42(30%) of clinical specialists expressed apprehensions about AI’s impact on patient privacy. The outcomes are consistent with previous global studies [11, 13], although some specific studies [25, 26] diverged from this trend.

Another noteworthy concern highlighted by specialists is the fear of AI contributing to unemployment. 70 (50%) of the participants indicated worries about job loss due to AI. A similar concern was noted in a separate study [32] that shared a comparable objective but utilized a different methodology. While some studies reported differing findings, such as a study in the UAE where 64% of participants were unconcerned about AI’s negative impact on their careers [27], this fear may be influenced by the preexisting high unemployment rates in Pakistan. Addressing this fear necessitates proper planning to ensure AI complements and enhances human skills rather than fully replacing them. Considering the benefits of AI, a substantial majority, comprising 85% of participants, expressed their support for the computerization of healthcare data. This opinion was accompanied by an acknowledgment of AI’s potential in achieving precise disease diagnosis, with a significant 72.1% of respondents acknowledging this potential accuracy. A further 77.3% of participants exhibited a belief that AI’s accuracy could potentially surpass that of doctors. Additionally, a considerable 88.5% of the respondents shared an agreement on AI’s potential to alleviate the demanding workload within the healthcare sector. However, it is worth noting that our study’s findings were not uniform across all dimensions. Certain aspects, such as “Provide documentation,” “personalized medication” for patients,” and “Perform a physical examination” tasks, garnered comparatively lower percentages (17%) of agreement, as indicated by previous research [11]. A recent study conducted in Malaysia [53], which featured a participant group consisting of nearly half of physicians, demonstrated parallels with our findings. In this study, approximately 60% of the participants believed that AI would exhibit lower error rates than human physicians. In our investigation, we observed a slightly higher percentage, with 77.3% of participants holding a similar belief.

Furthermore, the study noted that 108(77.1%) of clinical specialists agreed that AI could be integrated into the medical student curriculum. Similar findings have been given in the previous studies conducted in the USA [16, 54], where a majority of doctors expressed that AI should be part of the medical curriculum.

In conclusion, our study provides crucial insights into the perspectives of different clinical specialties on the potential roles of AI in healthcare. Overall, a positive attitude towards AI emerges, with many participants acknowledging its transformative impact on the future of healthcare. While reservations exist, a substantial number of specialists recognize the potential benefits and revolutionary capacity of AI in healthcare practices. There is widespread acknowledgment of AI’s diagnostic capabilities and a shared vision of its potential to enhance patient care, reduce workload, and integrate AI concepts into medical education, particularly in Surgery, Pathology, and Radiology specialties.

The study can help in providing a theoretical understanding of how healthcare professionals in developing countries adopt new technologies. It provides valuable insights that can inform existing theories on how individuals perceive and embrace technological innovations in healthcare. By exploring clinical specialists’ views on AI’s potential to replace, benefit, or disadvantage them, the study deepens our understanding of how these professionals assess and navigate emerging technologies. Additionally, its cross-sectional design enables us to analyze the relationships between various factors and AI acceptance among clinical specialists,

As far as practical implications are concerned, the findings of the study can guide policymaking in Pakistan by providing insights into the readiness of hospitals for AI integration and highlighting concerns regarding privacy violations and potential unemployment. This information can guide the development of policies and regulations aimed at facilitating the responsible adoption of AI in healthcare systems. Secondly, understanding clinical specialists’ perspectives on AI can aid healthcare organizations in allocating resources more effectively. For instance, if a significant number of specialists express apprehensions about privacy violations, resources can be directed toward implementing robust data protection measures. Thirdly, the study underscores the need for training programs designed to equip clinical specialists with the necessary knowledge and skills to utilize AI technologies effectively. These programs can address concerns and misconceptions while emphasizing the benefits of AI adoption.

The findings not only shed light on the current perspectives of clinical specialists regarding AI in healthcare but also open avenues for future research opportunities. Especially in high-stakes fields like Radiology and Gynecology, can be a focal point for developing robust and ethical AI applications. Moreover, the study identifies a noteworthy opportunity in the apprehensions expressed by specialists regarding job stability.

While our study has identified some significant factors related to the acceptance of AI as the future for clinical specialists, there is a significant limitation in conducting a survey study on AI in the healthcare sector specifically in the Southern Punjab region. The region faces limited access to technological resources like fast internet which could result in fewer specialists being able to respond during their job timing. Convenience sampling, while practical, poses limitations that must be acknowledged for the sake of research validity. The method may not yield a representative sample, introducing the risk of sampling bias as participants are selected based on accessibility rather than randomness. Generalizing findings beyond the specific sample and context can become challenging, compromising external validity. It is important to acknowledge that outcomes derived from this region may not generalize to other areas. The main reason is that the healthcare infrastructure in this region significantly lags behind that of other parts of Pakistan, accompanied by other factors such as facility availability, technological resources, healthcare staffing, and accessibility to advanced diagnostics and treatments.