Trust in AI applications and intention to use them in cardiac care among cardiologists in the UK: A Structural Equation Modeling Approach

Trust in AI applications and intention to use them in cardiac care among cardiologists in the UK: A Structural Equation Modeling Approach | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Trust in AI applications and intention to use them in cardiac care among cardiologists in the UK: A Structural Equation Modeling Approach Mahdi Mahdavi, Daniel Frings This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4114716/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background . The widespread use of Artificial Intelligence (AI)-driven applications among consultant cardiologists remains relatively low due to trust issues and perceived threat to professional autonomy, patient safety, and legal liability of misdiagnoses. There is a paucity of empirical research investigating the relationships between trust in AI applications and an intention to use (AI-Use) them among cardiologists. To address this gap, we surveyed a sample of cardiologists to examine the determinants of trust in AI and trust’s effects on AI-Use based on the organisational trust model. Methods . We conducted a cross-sectional survey of consultant cardiologists (n = 61) in the UK. Given the small sample size, we used a partial least square structural equation model (SEM) analysis approach to assess the measurement and structural models. We utilized factor loadings and weights for the measurement model assessment and coefficients, the redundancy indices, and goodness of fit (GoF) for the structural model assessment. We also undertook a content analysis of open-text responses around perceived risks, enablers, and barriers to AI use in cardiac care. We performed analyses in the R programme. Results . The GoF of the final SEM model was 63%, showcasing a substantial improvement over the original model (GoF=51%). The final model encompassed all latent constructs from the original model and explained 70% of the variance in trust and 37% in AI use. The AI application ability (accuracy and reliability) significantly influenced trust (β=0.55, p<.001), while lower benevolence correlated with decreased trust (β=0.19, p<.05). Trust in AI emerged as the sole significant contributor to AI-Use (β=0.48, p<.001), indicating higher trust associated with increased future use. Participants perceived diagnosis accuracy as a prominent theme, mentioned 20 times about AI risk and frequently cited as both an enabler (n=39 times) and a barrier (n=29 times). Conclusions . The enhanced GoF in the final model indicates an improved final SEM model compared with the original SEM model. Addressing diagnosis accuracy concerns and building trust in AI systems is crucial to facilitate increased AI adoption among cardiologists and seamless integration into cardiac care. Trust Artificial Intelligence Cardiologist Cardiac Care Intention to Use Technology Adoption Figures Figure 1 Figure 2 Figure 3 Background Artificial intelligence (AI)-driven applications have the potential to significantly improve the accuracy, efficiency, and reliability of diagnosis of coronary artery diseases (CADs). Using large datasets of previous heart images, AI algorithms have been trained to identify CADs by detecting abnormalities such as wall motion abnormalities, cardiac valves’ dysfunction, and ischemia [1]. However, the use of AI applications in healthcare remains relatively low due to several unresolved challenges facing AI in healthcare [2]. There is a paucity of research on determinants of the use of AI-driven applications from the perspective of medical practitioners across specialities [3]. The extant research, however, suggests that a perception that AI applications may produce potential false negative diagnoses (thus posing a considerable threat to the patient’s health) hinders the widespread adoption of AI applications [2]. Other features of AI applications such as explainability and transparency are believed to affect the adoption of AI-enabled applications in clinical care [4–7]. Focusing on cardiac care specifically, contextual factors such as perceived threat to professional autonomy, fear of replacement, concerns about patient safety, and legal liability of misdiagnosis hinder the widespread use of AI applications by cardiologists [8,9]. The challenge of a limited evidence base around AI in cardiac care is exacerbated as evidence base in this area dates rapidly due to changes in technology and attitudes. These issues raise several questions regarding the use of AI tools in AI health care. In this research, we address three questions below: Do cardiologists have a willingness to trust AI for the diagnosis of heart diseases and factors that predict this? How do clinicians perceive the potential consequences of using AI for their clinical autonomy and responsibility? How do contextual factors such as risk associated with the use of AI applications influence the intention to use AI? Review of conceptual models Few frameworks are used to study the use of technology in health care. Technology adoption frameworks and trust theoretical frameworks present two major perspectives to study the use of AI tools in healthcare. In this research, we used the terms ‘AI tool’ and ‘AI application’ interchangeably, however ‘AI tool’ is a technical term typically referring to the technologies or software libraries used to develop and implement AI solutions. In contrast, AI application refers to the systems and solutions that deploy AI technologies to perform analysis or identify patterns [10]. Within technology adoption frameworks, a dominant framework is the Unified Theory of User Acceptance of Technology (UTAUT), which integrates eight prominent technology acceptance theories in the healthcare service adoption literature [11]. Building on this framework, Praksash et al. report that the future use of AI is determined by performance expectancy, effort expectancy, social influence, initial trust, and resistance to change [4]. The seminal framework of the Organisational Model of Trust developed by Mayer et al. [12] presents the second major perspective to study the use of AI tools. This framework provides a basis for most studies around trust in AI and similar technologies under the umbrella term 'automation'. This framework was originally developed for studying organizational trust, but it has been well-received by researchers studying trust in AI and automation [13–15]. It has six primary components, of which only one is trust itself and those others are antecedents, context, and product of trust. Factors of perceived trustworthiness, trust, propensity of trustors, risk-taking behaviour, perceived risks, and outcome are key components of the trust model [12]. A review study reports that most studies about trust in automation study variables which fit into one or more components of Mayer’s organisational model of trust [16]. Starke et al. assess three features of AI tools consisting of reliability, competence, and intention to assess physician trust in AI systems [17]. They emphasize the relevance of contextual factors and concerns in the use of AI systems; for instance dependence on AI, loss of professional autonomy, and the fear of losing control by adopting the AI systems that in the future may disrupt patient-physician relationships [17]. While helpful in identifying these factors, their framework provides no explicit top-level constructs, nor specifies the relationships between them. In this research, we adapted the organizational model of trust to accommodate the research question of this study. Since most healthcare professionals have no or limited access to AI tools in their day-to-day practice, measurement of actual usage would lead to biased conclusions about the adoption of AI tools in healthcare. Intention rather than actual use would be an accurate outcome variable in the adoption model [18]. Our conceptual model and hypotheses The current paper builds a conceptual model based on Mayer’s Organisational Model of Trust [19]. We slightly adapted the Mayer model of trust for our study and extended it by adding two constructs innovativeness and peer influence (Figure 1) based on previous studies [20]. The components of the trust model consist of the factors of perceived trustworthiness, trust, propensity of trustors, risk-taking behaviour, perceived risks, and outcome. This model starts with the factors of perceived trustworthiness which determine the level of trust. The model differentiates between trust (as an attitude) and outcome of trust (intention to use AI in the future, an intention). This distinction in our study implies that there might be trust in AI but it may not lead to the use of AI in cardiac care settings. Trust is associated with risk-taking behaviour (in the current case, future use of AI systems, described in the original model of trust as risk-taking in relationships). The outcome (in the original model) refers to the consequences associated with the use of technology by cardiologists (i.e. clinical outcomes) which was not considered in this study. Instead of actual use of AI applications, we considered an intention to use AI in future. Below we describe our model’s components. Ability, benevolence, and integrity determine perceived trustworthiness [19]. Ability refers to skills, quality, and characteristics of AI tools, which are context- and tool-specific and are believed to be significant contributors to the formation of trust [13,16,17]. In the context of AI in cardiac care settings, ability measures the accuracy of diagnosis made by AI applications, reliability, and whether AI applications produce results without breaching security. The second factor, benevolence, refers to a situation where an AI tool intends to do good for users. Integrity refers to situations where the AI tool adheres to a set of principles that the user finds acceptable. We postulate that ability (hypothesis (H)1), benevolence (H2), and integrity (H3) are positively associated with trust in AI. As a propensity of cardiologists, knowledge of AI applications (both theoretical and practical) affects trust in AI. We postulate that knowledge of AI applications next to ability, benevolence, and integrity are associated with trust (H4). Trust in AI is the key intermediate outcome linked to usage. Based on the Mayer model, trust itself determines risk-taking to use technology and thus intention to use [12]. We postulate that trust in AI has a positive impact on an intention to use AI (H5). Li et al. reported a direct association between social influence and trusting beliefs in information system settings [20]. They argue that when individuals have inadequate knowledge and lack hands-on experience with technology, it is highly likely that they rely on trusted others’ opinions and incorporate their views into trust formation [21]. We postulate that peer influence (in terms of others being greater adopters) has a positive influence on the future use of AI applications (H6). Personal innovativeness refers to the willingness of an individual to try out an innovation. Studies support that individuals with a high degree of innovativeness are more likely to use computer-based applications and have positive perceptions about the ease of use of technology [22,23]. We postulate that personal innovativeness has a positive impact on the future use of AI applications (H7). The ‘perceived risk’ in the Mayer model of trust, refers to situational factors that, next to trust in AI, determine the use of AI. These may include how AI is perceived in working environments. The rise of AI tools and their potential have triggered fear of a substitution crisis (where valued roles and livelihoods are taken over by AI tools). In clinical settings, the capacity of AI tools to imitate human decision-making processes poses threats that skills and competencies possessed by physicians may be replaced by AI tools and thus can replace physicians [24]. We therefore postulate that perceived substitution threats have a negative influence on intention to use AI applications in future (H8). Alongside these hypothesized relationships, we also aimed to explore what cardiologists felt were the key risks and benefits of the use of AI, and how they would likely respond in instances when their clinical judgement and that of an AI system diverged. Methods We conducted a cross-sectional survey of consultant cardiologists in the UK. We used the SEM approach to validate a measurement model and to test our path model. We collected data from cardiologists (as a convenience sample) and included 61 participants in the final data analysis. Since our sample size was small, we used partial least square regression models to assess the measurement model and path analysis. We also used qualitative content analysis to elicit the main themes on perceived risks of using AI and the enablers and barriers of AI use in cardiac care. The study population consisted of consultant cardiologists in the UK (see below). Variables and measurement The demographic and socio-economic variables measured comprised age groups, gender, ethnicity, and years of experience as a cardiologist. We classified participants based on their ages into three age groups: 30–39, 40–49, and 50–70 years. We categorized participants into White, Asian/Asian British, Black/African/Caribbean/Black British, Mixed/Multiple Ethnic Groups, or Other Ethnic Group. Concerning the years of experience working as a cardiologist, study participants were categorized into four groups: 1–5, 6–10, 11–15, or 16 + years of experience. We also categorized participants based on the number of monthly SE images reviewed by them into four groups: 0–20, 21–50, 51–100, and 101 + images. Attributes of cardiologists were explored through two questions that measured familiarity with AI applications and feeling knowledgeable about AI systems (Supplementary file 1. The ability of AI systems was assessed via outcome, security, and reliability items. This refers to the AI system's capacity to produce accurate results without breaching patient privacy and without breaking down during the process [ 25 ]. Benevolence was measured using three items referring to whether AI systems are deceptive, participants are suspicious of AI, and they are wary of AI systems. Integrity was measured using two items. Trust was formed by two items measuring whether participants have trust and confidence in AI systems. Attributes of cardiologist, ability, benevolence, and integrity were measured using a Likert scale of 1–5. We provided questionnaire items to measure the perceived risk of using AI, innovativeness, and peer influence in Supplementary File 1. We measured the perceived risk of using AI [ 17 ] by measuring the adverse impact of using AI applications on confidence and trust between patients and doctors, adverse impacts on patient-doctor relationships, deskilling doctors’ roles, and needs for cardiologists in a Likert scale of 1–5. The innovativeness of cardiologists was measured by four items. Peer influence was measured through three items. The use of AI in the next 12 months was measured by a single item. Peer influence, innovativeness and future use of AI were measured using a Likert scale of 1–7. Participants described the potential risks of following the AI’s recommendation and further actions in response to two hypothetical scenarios: A) Where AI contradicts the clinician’s diagnosis of CAD being present; and B) AI contradicts the clinician's diagnosis of CAD not being present. We also asked participants to determine the extent of confidence in their initial diagnosis to retain it without taking further actions in response to the above scenarios. We also asked participants to describe barriers and enablers of using AI decision-support software in cardiac care. Questionnaire and data collection We used questionnaires for data collection. Demographic questions were developed based on a framework proposed by UK’s office for National Statistics [ 26 ]. We used 12-item instrument developed by Jian et al. to measure trust in AI and the factors of perceived trustworthiness of AI applications [ 27 ]. We also embark on the framework proposed by Starke et al. to incorporate questions on the substitutions threats of using AI, such as threats to patient-doctor relationship [ 28 ]. For items measuring the dimensions of innovativeness and peer influence, we used instrument developed Fan et al. [ 18 ]. We collected data through an online survey tool (Qualtrics). We applied screening questions to identify eligible participants i.e., consultant cardiologists. These included minimum age expected to start a consultant cardiologist career, and working in a cardiac care setting. 88 completed responses were collected, of which 27 were excluded due to ineligibility. We considered 61 participants in the final analysis. Analysis methods We conducted the descriptive analysis of sociodemographic and outcome parameters using summary statistics (mean and SD). We used the structural equation modelling (SEM) approach for the path analysis. An SEM model encompassed two components: a measurement model to operationalize the conceptual framework of trust and the future use of AI and a structural model to describe coefficients of relationships between constructs. We evaluated multiple measurement models and compared their performance in terms of goodness of fit (GoF) and other indices to reach the final measurement model for SEM. The SEM analysis relied on the partial least square (PLS) regression model, which is a suitable method for studies with small sample sizes [ 29 ]. We iterated model selection based on the loading of variables on a construct, cross-loading, and communalities. A factor loading presents correlations between a variable (i.e. a scale item) and a latent construct (e.g., trust in AI). Cross loading refers to loadings of variables on the rest of latent constructs i.e. constructs other than the hypothesized construct. Communalities represent the amount of variability in a latent construct explained by a variable. Variables with loadings lower than 0.7 were removed from the model. The quality of the structural model was evaluated using regression equations and by \({R}^{2}\) coefficient, the redundancy index, and GoF. \({R}^{2}\) indicates the amount of variance in latent constructs (trust and future use of AI) explained by their independent items. The values for \({R}^{2}\) were classified into low (R < 0.30), moderate (0.30 < R 0.60) [ 29 ]. Redundancy refers to the ability of the independent variables to explain variation in trust and future use of AI. The average communality indicates how much of each construct’s variability is reproducible by its latent constructs or items. GoF accounts for both the measurement and the structural model quality. GoF is calculated as the mean of the communality and the average \({\varvec{R}}^{2}\) value. GoF higher than 70% is perceived as acceptable [ 29 ]. We used the thematic analysis method for qualitative analysis of participants' responses to scenarios A and B as well as the enablers and barriers of using AI applications [ 30 ]. We followed six steps for this analysis: 1) familiarization with data, 2) generating initial codes, 3) searching for themes, 4) reviewing themes, 5) defining and naming themes, and 6) producing a report. We took an inductive analysis approach, with the themes identified strongly linked to the data themselves. Results Most participants were male (85%). The age of participants ranged from 30 to 70 years; 46% of participants were aged between 40–49 years. The largest ethnicity group was White comprising 64% of participants (see Table 1 for other ethnicities). Most participants have worked as a cardiologist for longer than five years (Table 1 ). More than half of participants (54%) reported current use of AI in a cardiac care pathway (Table 2 ). Table 1 Demographic characteristics of the survey participants Demographic information Frequency Percent (%) Sex Female 8 14 Male 50 85 Other 2 3 Age groups 30–39 16 28 40–49 26 46 50–70 15 26 Ethnicity White 39 64 Asian/Asian British 12 20 Black/African/Caribbean/Black British 3 5 Mixed/Multiple Ethnic Groups 2 3 Other Ethnic Group 5 8 Years working as a cardiologist 1–5 12 20 6–10 15 25 11–15 17 28 16+ 17 28 Table 2 Experience of cardiologists on the use of AI applications in cardiac care and reviewing stress echocardiography tests AI in the cardiac care pathway Frequency Percent (%) Currently using AI tools in cardiac care Yes 33 54 No 27 44 Unknown 1 2 Currently using auto-indexing software in cardiac care Yes 33 54 No 28 46 Currently using Decision Support Software in cardiac care Yes 23 38 No 38 62 Number of monthly stress echo tests reviewed 1–20 tests 12 20 21–50 tests 29 48 51–100 tests 7 11 101 + tests 13 21 The measurement model assessment The results of assessing the final measurement models are given in Table 3 . In the final model, all loadings were greater than 0.7 and therefore acceptable. A loading greater than 0.7 means that \({0.7}^{2}=50\%\) of the variability in a latent construct is captured by a variable. Lastly, since we only have a single-item scale for the integrity of AI, substitution threat, peer influence, innovativeness, and the use of AI in the next 12 months, the factor loadings and communality for them are 1. Compared with the original conceptual model, the number of items per construct in the final model was markedly reduced, yet the final model encompassed the main constructs suggested by the conceptual framework. Table 3 Assessment of final measurement model Weight Loading Communality Redundancy Attributes of cardiologist Familiarity with AI system 0.56 0.94 0.88 Feeling knowledgeable about AI 0.51 0.92 0.85 The ability of AI systems Security of AI systems 0.61 0.83 0.69 Reliability of AI systems 0.60 0.83 0.69 Benevolence Being wary of AI systems 0.73 0.89 0.80 Perceived harmfulness of AI systems 0.48 0.73 0.53 Perceived integrity of AI systems 1.00 1.00 1.00 Perceived trust in AI systems Perceived trust in AI systems 0.64 0.95 0.89 0.63 Perceived confidence in AI systems 0.45 0.88 0.78 0.55 Substitution threat 1.00 1.00 1.00 Peer influence 1.00 1.00 1.00 Innovativeness 1.00 1.00 1.00 Use of AI in the next 12 months 1.00 1.00 1.00 0.37 The cross-loading analysis (Supplementary File 1) showed that the loadings of all variables/items on their construct were higher than 0.7 and no variable loaded on constructs other than the corresponding construct. The structural model assessment The quality of the structural model was evaluated using PLS regression and regression coefficient, the redundancy index, and GoF. As shown in Fig. 2 , ability (β = 0.55, p < .001), benevolence (β = 0.19, p < .05), and attributes of cardiologists (β = 0.25, p < .05) were significant contributors to trust in AI tool. Therefore, hypotheses H1, H2, and H4 were supported. The perceived ability of the AI tool made the largest contribution to trust; one unit (on a Likert scale between 1 and 5) increase in ability was associated with a 0.55 unit (scale between 1 and 5) increase in trust. Lower benevolence (i.e., lower level of good intention of AI) was associated with lower trust. H3 postulating the relationship between integrity and trust was not supported. Among the constructs that were used to explain the future use of AI, only trust in AI was a significant contributor (β = 0.48, p < .001), with higher trust being linked to increased future use. Therefore, H5 was supported. Other hypotheses postulating relationships between peer influence, innovativeness and perceived risk of AI and the use of AI (H6, H7, and H8 respectively) were not supported. For the structural model, we evaluated the quality of the PLS model of trust and the future use of AI based on the \({R}^{2}\) coefficient (Table 4 ). \({R}^{2}\) indicated that 70% of the variance in trust can be explained by attributes of cardiologist, ability, benevolence, and integrity. Only 37% of the variance in the future use of AI could be explained by its independent latent constructs. In terms of classification used to assess the regression models, \({R}^{2}\) for trust was high and for the future use of AI was moderate [ 29 ]. Table 4 Quality indicators for structural model assessment Construct \({\varvec{R}}^{2}\) Communality Redundancy Attributes of cardiologist 0.86 Ability 0.69 Benevolence 0.66 Integrity 1 Trust 0.70 0.84 0.59 Perceived risk 1 Peer influence 1 Innovativeness 1 Future use of AI 0.37 1 0.37 The value of redundancy indicated that the dependent latent variables of trust (ability, benevolence, integrity, and attributes of cardiologists) explain around 59% of the variability in trust (Table 4 ). Latent variables including trust, perceived risk, peer influence, and innovativeness explained 37% of the variability in the future use of AI. The average communality indicated that two variables of attributes of cardiologists reproduced 87% of the variability in attributes of cardiologists. 69%, 66%, and 84% of the variability in ability, benevolence, and trust respectively were reproducible by their variables. Based on GoF, the predictive power of the model is 64%. Scenario analysis In a situation where the cardiologist made a diagnosis of CAD being present or not present, but an AI tool contradicted this, the mean values of confidence in their diagnosis were 69% and 65%, respectively. The results of regression analysis testing the relationships between mean values of confidence in own diagnosis and years of experience as a cardiologist, number of monthly images reviewed, and current use of AI in cardiac care pathway are given in Table 5 . Table 5 Regression models for scenario analysis Scenario A Scenario B Model β Std error β Std error Intercept 35.69* 13.12 58.38* 15.51 Years of experience as a cardiologist 1–5 (reference) 6–10 -1.51 7.39 0.45 8.74 11–15 10.21 7.48 -1.21 8.85 16+ 0.90 7.19 -0.52 8.50 Number of monthly stress test reviews 0–20 tests (reference) 21–50 tests 3.61 6.83 -11.57 8.07 51–100 tests 4.13 8.92 -8.72 10.54 101 + tests 4.90 7.83 -3.93 9.26 AI in the cardiac care pathway Yes (reference) No 3.68 5.06 -2.88 5.98 Trust 7.73* 3.11 4.58 3.67 Note: * = p < .05. As shown in Table 5 we found that confidence in own diagnosis was not associated with years of experience, the number of monthly images reviewed, or the current use of AI in the cardiac care pathway. However, when cardiologists had a higher trust in AI tools, they would need to have higher confidence in their own diagnosis to choose not to embark on further tests or seek a second opinion. Perceived risk of and further actions taken following AI recommendations Four main themes emerged during the content analysis of the perceived risk of following AI recommendations. The risk of inaccurate diagnosis or treatment was the most repeated theme (n = 20 instances) (Supplementary file 1). The second theme was the risk of adverse events (n = 15). The risk of litigation and legal responsibility was also mentioned (n = 6). Lastly, some participants indicated that they thought AI was only a decision-support tool and there was no risk (n = 2). These numbers refer to the sum of themes in scenarios A and B (Supplementary File 1). In terms of further actions taken when AI contradicts cardiologist judgement, the most repeated theme was the need to embark on further tests, imaging, scans or examinations (n = 61 times). The second most mentioned action was to verify or confirm evidence/information through wider assessment (n = 27). The third most repeated action was to ask for a second opinion or to consult with a peer (n = 26). In a small number of cases, participants reported taking no action (n = 10). Enablers and barriers to using AI Our analysis of enablers and barriers to using AI applications identified multiple themes that were iterated as both enablers and barriers (Supplementary file 1). The theme ‘improving accuracy’ was the most repeated enabler (n = 39 times). The second most repeated enabler was ‘saving time and more speedy diagnosis’ (n = 27). And the third most repeated enabler was ‘cost saving and efficiency’ (n = 18) and ‘ease of use and automation’ (n = 18). The most repeated barrier was the concern over the accuracy of diagnosis (n = 29). The second most mentioned barrier was concerns over costs and efficiency (n = 26). The third most repeated barrier was concern over time (n = 10) related to the production of AI reports may take time and/or AI reports may arrive late). Discussion This study aimed to explore trust in AI applications in cardiac care among cardiologists, based on an adapted version of Mayer’s model of organizational trust [ 12 ]. The findings present mixed support for the hypotheses generated by this model. Our findings indicate that two factors of trustworthiness (ability and benevolence) and propensity of trustors have significant associations with trust in AI. The positive influence of trust on the future use of AI aligns with existing evidence emphasizing the importance of trust in fostering acceptance and adoption of technological innovations [ 31 ]. The overall SEM structure confirms established evidence that trust is understood by its determinants and that it is positioned between factors of trustworthiness and AI use [ 32 ]. The positive impact of the AI system’s ability emphasizes the pivotal role of technical competence i.e., reliability and security, in the formation of cardiologists' trust in such systems. This finding aligns with the already recognized importance of security in building trust in AI [ 33 ]. Prior knowledge of and familiarity with AI were also identified as significant contributors to trust in AI (H4). Integrity was not associated with trust, which is in contrast with Mayer's model, where it is a key factor of trustworthiness [ 12 ]. AI applications offering a higher level of integrity tend to build higher subjective trust [ 34 ]. Contrary to our hypotheses and findings reported by other studies [ 18 , 31 ], peer influence, perceived risk of using AI, and personal innovativeness did not emerge as significant factors influencing the future use of AI. Cardiologists’ intention to use AI was not significantly influenced by peer influence (operationalized by the amount that colleagues used AI). This is in contrast with existing findings around the positive role of important people in shaping attitudes towards AI adoption within professional settings [ 21 ]. This may be because the question focused on increased AI use, and the sample was broadly positive about AI. Our findings around the lack of relationships between the intention to use AI and peer influence and innovativeness of cardiologists are in contrast with UTUAT which suggests these components as key determinants of the technology use [ 11 ]. However, this might be rooted in the fact that we incorporated these two components from UTUAT which is a distinctive conceptual framework from Mayer’s trust model, and used a different set of items to capture this dimension. In the context of a study designed based on UTUAT, relationships between peer influence and innovativeness with the use of technology may be significant. However, it does suggest that the context of measurement is important for UTUAT. Our findings therefore provide empirical grounds for studying trust and intention to use AI, responding to the call for an empirical-based conceptual framework of trust in AI in healthcare [ 31 ]. Our SEM analysis only explained a modest extent of variance in the intention to use AI, and only some predictors were statistically significant. These findings do not contradict the validity of the Mayer model of trust since we did not measure risk-taking behaviour and outcomes of trust according to the original trust model. Rather they suggest that while elements of Mayer's model are important (see Fig. 2 ), predictors of AI may be more nuanced or situation-specific. To address this in the context of cardiac AI, we conducted a qualitative assessment of the risks (and benefits) of following AI recommendations, especially in situations where the AI recommendation diverges from that of clinicians. Future research needs to embark on full-fledged operationalization of the trust model in this context or extend SEM models to explain a large extent of unexplained variance in the intention to use AI. The implications of this study are far-reaching for both researchers and the cardiac care community. Our study provided evidence around the multiple dimensions related to trust in AI and its adoption. These are likely crucial for designing trustworthy AI and facilitating the widespread adoption of AI in cardiac care (Fig. 3 ). This study has some limitations. Our sample size (n = 61) was relatively small which has implications for the generalizability of findings within the cardiologist community and beyond. Caution should be exercised when extrapolating findings to the broader medical community or to non-cardiologist healthcare professionals. Furthermore, females and non-white ethnic groups were under-represented in our sample, consequently, our findings might not adequately reflect the perceptions of female cardiologists and other ethnic communities. Due to the small sample size, achieving optimal model fit became challenging; yet the GoF of our final SEM was close to the optimal level (70%) and our analysis identified alternative models that improved overall model fit. The sample size needs to be cautiously viewed as the British consultant cardiologist community is small (~ 2,600 members), meaning we sampled only 2.3% of the population. We used the PLS method for SEM analysis which is developed to deal with shortcomings of small samples, but the small sample might have affected detecting true effects within the model. However, the observation that we did not face wide confidence intervals and model identification issues does provide some reassurance that the sample was sufficient. Conclusions The enhanced GoF in the final model indicates an improved final SEM model compared with the original SEM model despite the reduction in the number of items included in the final model. Addressing specific concerns related to accuracy and adverse events, while fostering trust in AI systems, will be essential for ensuring the seamless integration of AI applications into routine clinical practices. In contrast to previous studies findings, effects of peer influence, perceived substitution threats, and innovativeness on the use of AI applications were not found in this study (4 out of 8 hypotheses in this research were not supported). However, ability, benevolence, and previous familiarity and knowledge of AI were found to be key factors in trust in AI. The subsequent use of AI was associated with trust itself rather than contextual factors such as peer influence, personal innovativeness or perceived risk of AI. The Integration of quantitative and qualitative data provided a comprehensive understanding of the complex interplay of factors influencing cardiologists'’ trust in AI and the use of AI in the future. The study findings have practical implications for healthcare organizations aiming to implement AI technologies successfully, and for technology developers to ensure core concerns are reassured through robust design and testing. Addressing specific concerns related to accuracy, costs, and time, while fostering trust in AI systems, will be essential for ensuring the seamless integration of these technologies into routine clinical practices. Abbreviations AI: artificial intelligence. SE: Stress echocardiography. CAD: coronary artery disease. SEM: structural equation model. UTAUT: Unified Theory of User Acceptance of Technology. H: hypothesis. GoF: goodness of fit. PLS: partial least square. Declarations Ethics approval and consent to participate This research has received a favourable opinion from the NHS Health Research Authority (IRAS No: 315284) and approval from the London South Bank University Ethics Panel (ETH2223-0164). Informed Consent was obtained from the participants in the study before giving access to study questionnaires or collecting any data. We did not collect any personal data. Availability of data and materials The dataset generated and analysed during the current study is available in the Open Science Framework at osf.io/eknfz [35]. All other materials relevant to this study including method statements and analysis plan are available at the same place. Competing interests The authors declare that they have no competing interests. Funding This work was supported by NHSx and NHS Accelerated Access Collaborative grant number REI2122-0030 (AI Award in Health and Care, Technology Specific Evaluation Team). This report is independent research and the views expressed in this publication are those of the authors and not necessarily those of the NHS, NHSX or the Department of Health and Social Care. Authors' contributions DF acquired the funding for the study. MM and DJF conceptualised the study, developed the study design, and implemented the study. MM analysed the data and drafted the manuscript. DJF critically reviewed the manuscript. Both authors approved the final version of the manuscript. Acknowledgements We would like to express our sincere gratitude to all individuals and organizations who contributed to the completion of this paper. Special thanks to Virgil Griffith and Sharanya Jayadev for their support and assistance throughout the research process. Additionally, we extend our appreciation to the participants who generously shared their insights and experiences, without whom this study would not have been possible. References O’Driscoll JM, Hawkes W, Beqiri A, et al. Left ventricular assessment with artificial intelligence increases the diagnostic accuracy of stress echocardiography. Eur Heart J Open. 2022;2. 10.1093/ehjopen/oeac059 . Rasheed K, Qayyum A, Ghaly M, et al. Explainable, trustworthy, and ethical machine learning for healthcare: A survey. Comput Biol Med. 2022;149. https://doi.org/10.1016/j.compbiomed.2022.106043 . Shinners L, Aggar C, Grace S, et al. Exploring healthcare professionals’ understanding and experiences of artificial intelligence technology use in the delivery of healthcare: An integrative review. Health Inf J. 2020;26. 10.1177/1460458219874641 . Prakash AV, Das S. Medical practitioner’s adoption of intelligent clinical diagnostic decision support systems: A mixed-methods study. Inform Manage. 2021;58. 10.1016/j.im.2021.103524 . Slostad B, Karnik A, Appadurai V, et al. Applications of Artificial Intelligence in Echocardiography. Curr Cardiovasc Risk Rep. 2023;17. https://doi.org/10.1007/s12170-023-00721-6 . Laad M, Kotecha K, Patil K, et al. Cardiac Diagnosis with Machine Learning: A Paradigm Shift in Cardiac Care. Appl Artif Intell. 2022;36. https://doi.org/10.1080/08839514.2022.2031816 . Khalsa RK, Khashkhusha A, Zaidi S, et al. Artificial intelligence and cardiac surgery during COVID-19 era. J Card Surg. 2021;36. https://doi.org/10.1111/jocs.15417 . Jussupow E, Spohrer K, Dibbern J et al. AI changes who we are - Doesn’t it? Intelligent decision support and physicians’ professional identity. 26th European Conference on Information Systems: Beyond Digitization - Facets of Socio-Technical Change, ECIS. 2018. 2018. Hengstler M, Enkel E, Duelli S. Applied artificial intelligence and trust-The case of autonomous vehicles and medical assistance devices. Technol Forecast Soc Change. 2016;105. 10.1016/j.techfore.2015.12.014 . Russell S, Norivg P, Artificial Intelligence. A Modern Approach (Global Edition). Artificial Intelligence: A Modern Approach. 2021. Williams MD, Rana NP, Dwivedi YK. The unified theory of acceptance and use of technology (UTAUT): A literature review. J Enterp Inform Manage. 2015;28. https://doi.org/10.1108/JEIM-09-2014-0088 . Mayer RC, Davis JH, David Schoorman F. An Integrative Model of Organizational Trust. 1995. https://www.jstor.org/stable/258792?seq=1&cid=pdf- Jiun-Yin Jian, Ann M, Bisantz CGD. Foundations for an Empirically Determined Scale of Trust in Automated Systems. Int J Cogn Ergon. 2000;4:53–71. Sousa S, Martins P, Cravino J. Measuring Trust in Technology: A Survey Tool to Assess Users’ Trust Experiences. 29TH INTERNATIONAL CONFERENCE ON INFORMATION SYSTEMS DEVELOPMENT. 2021:2–7. Wojton HM, Porter D, Lane T. Initial validation of the trust of automated systems test (TOAST). J Soc Psychol. 2020;160. 10.1080/00224545.2020.1749020 . Kohn SC, de Visser EJ, Wiese E, et al. Measurement of Trust in Automation: A Narrative Review and Reference Guide. Front Psychol. 2021;12:4138. Starke G, van den Brule R, Elger BS, et al. Intentional machines: A defence of trust in medical artificial intelligence. Bioethics. 2022;36:154–61. Fan W, Liu J, Zhu S, et al. Investigating the impacting factors for the healthcare professionals to adopt artificial intelligence-based medical diagnosis support system (AIMDSS). Ann Oper Res. 2020;294. 10.1007/s10479-018-2818-y . Mayer RC, Davis JH, Schoorman FD. An Integrative Model Of Organizational Trust. Acad Manage Rev. 1995;20:709–34. Li X, Hess TJ, Valacich JS. Using Attitude and Social Influence to Develop an Extended Trust Model for Information Systems. Data Base Adv Inform Syst. 2006;37. 10.1145/1161345.1161359 . Li X, Hess TJ, Valacich JS. Why do we trust new technology? A study of initial trust formation with organizational information systems. J Strategic Inform Syst. 2008;17. 10.1016/j.jsis.2008.01.001 . O’Connor Y, O’Connor S, Heavin C, et al. Sociocultural and Technological Barriers Across all Phases of Implementation for Mobile Health in Developing Countries. Appl Comput Med Health. 2015. https://doi.org/10.1016/B978-0-12-803468-2.00010-2 . Jeong SC, Choi BJ. Moderating Effects of Consumers’ Personal Innovativeness on the Adoption and Purchase Intention of Wearable Devices. Sage Open. 2022;12:1–14. Tran AQ, Nguyen LH, Nguyen HSA, et al. Determinants of Intention to Use Artificial Intelligence-Based Diagnosis Support System Among Prospective Physicians. Front Public Health. 2021;9. 10.3389/fpubh.2021.755644 . Starke G, van den Brule R, Elger BS, et al. Intentional machines: A defence of trust in medical artificial intelligence. Bioethics. 2022;36:154–61. Office for National Statistics. Demography question development for Census 2021. 2024. Jian J-Y, Bisantz AM, Drury CG. Foundations for an Empirically Determined Scale of Trust in Automated Systems. http://dx.doi.org/101207/S15327566IJCE0401_04 . 2010;4:53–71. Starke G, Brule R, Elger BS, et al. Intentional machines: A defence of trust in medical artificial intelligence. Bioethics. 2022;36:154–61. Hair JF, Sarstedt M, Hopkins L, et al. Partial least squares structural equation modeling (PLS-SEM): An emerging tool in business research. Eur Bus Rev. 2014;26. https://doi.org/10.1108/EBR-10-2013-0128 . Clarke V, Braun V, Hayfield N. Thematic analysis. Qualitative psychology: A practical guide to research methods. 2015;222:248. Gille F, Jobin A, Ienca M. What we talk about when we talk about trust: Theory of trust for AI in healthcare. Intell Based Med. 2020;1–2:100001. Steerling E, Siira E, Nilsen P, et al. Implementing AI in healthcare—the relevance of trust: a scoping review. Front Health Serv. 2023;3. 10.3389/frhs.2023.1211150 . Anson KC, Li IARKK. Knowledge is not all you need to generate trust in AI use in healthcare. medRxiv. 2024;01. Mehrotra S, Centeio Jorge C, Jonker CM et al. Building Appropriate Trust in AI: The Significance of Integrity-Centered Explanations. Frontiers in Artificial Intelligence and Applications. 2023. https://doi.org/10.3233/FAIA230121 . Mahdavi Mahdi FDJS. Evaluating Artificial Intelligence Driven Stress Echocardiography Analysis System (EASE Study). WP4: Trust in AI and Future Use of AI. 2024. osf.io/eknfz . (accessed 15 March 2024). Additional Declarations No competing interests reported. Supplementary Files SupplementaryFile1v1.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4114716","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":283049791,"identity":"a9cab26f-8246-4672-b84c-b7a806d9ee6c","order_by":0,"name":"Mahdi Mahdavi","email":"","orcid":"","institution":"London South Bank University","correspondingAuthor":false,"prefix":"","firstName":"Mahdi","middleName":"","lastName":"Mahdavi","suffix":""},{"id":283049792,"identity":"5bec56c2-c9a4-4ae0-b43c-1bf737a02302","order_by":1,"name":"Daniel Frings","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIie3RoQrCQBjA8U+EpU3rXXGvcGuKD+ErbEXLtFhMIghaBOuarzCbwfAdFywn1tlmMRmsa945xOSpTfD+sDs29oPvOACb7RdD9YyAqLXKsTIpPz7210TeiRN+RXQu+4zUdnsEvm0O/I688mIjoD5HhyYGQuUgBDyTYZD1U+FJAUSGDk0NhGHMqlck0TrxUlGZCYAMHJqbyOHCADVZypwXivhvSRaXZAUxoKcI08Q0GM0uDDVJSZeps/TcQEbTlun4tUMc5IjjaLUUp1OxaTcaO8GPCwPR4X1CLF/cdxf5zP/wP5vNZvvDbujkW8/Lgrc5AAAAAElFTkSuQmCC","orcid":"","institution":"London South Bank University","correspondingAuthor":true,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Frings","suffix":""}],"badges":[],"createdAt":"2024-03-16 21:29:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4114716/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4114716/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":53491818,"identity":"ece90e05-9a96-423f-b2ce-1b008e78d8db","added_by":"auto","created_at":"2024-03-26 15:47:52","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":50049,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual framework for measurement and analysis of trust in AI applications in cardiac care settings.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4114716/v1/772381cd78f3539c7340fdb8.png"},{"id":53491820,"identity":"3cf758a1-45ec-4193-ab1f-246759bb7f1a","added_by":"auto","created_at":"2024-03-26 15:47:53","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":66233,"visible":true,"origin":"","legend":"\u003cp\u003eStructure model and path coefficients\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4114716/v1/601466ba740847ffdf97074c.png"},{"id":53491819,"identity":"650451ff-cc2a-43e4-97b6-3fd3c8249a22","added_by":"auto","created_at":"2024-03-26 15:47:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":26713,"visible":true,"origin":"","legend":"\u003cp\u003eStatistically significant contributors to trust and future use of AI\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4114716/v1/081cc7140da6419092e48a5b.png"},{"id":60139094,"identity":"def1d2ab-60b0-4d81-a9d8-471af3d3f512","added_by":"auto","created_at":"2024-07-12 08:35:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":924856,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4114716/v1/347aa642-105b-4aa1-b14e-6a7592399730.pdf"},{"id":53491817,"identity":"27bc7cbe-a669-4777-81f4-ae2c6d279342","added_by":"auto","created_at":"2024-03-26 15:47:52","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":154441,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFile1v1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4114716/v1/6c3e85ba4c7f4f825100e98c.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Trust in AI applications and intention to use them in cardiac care among cardiologists in the UK: A Structural Equation Modeling Approach","fulltext":[{"header":"Background","content":"\u003cp\u003eArtificial intelligence (AI)-driven applications have the potential to significantly improve the accuracy, efficiency, and reliability of diagnosis of coronary artery diseases (CADs). Using large datasets of previous heart images, AI algorithms have been trained to identify CADs by detecting abnormalities such as wall motion abnormalities, cardiac valves\u0026rsquo; dysfunction, and ischemia [1]. However, the use of AI applications in healthcare remains relatively low due to several unresolved challenges facing AI in healthcare [2]. There is a paucity of research on determinants of the use of AI-driven applications from the perspective of medical practitioners across specialities [3]. The extant research, however, suggests that a perception that AI applications may produce potential false negative diagnoses (thus posing a considerable threat to the patient\u0026rsquo;s health) hinders the widespread adoption of AI applications [2]. Other features of AI applications such as explainability and transparency are believed to affect the adoption of AI-enabled applications in clinical care [4\u0026ndash;7].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFocusing on cardiac care specifically, contextual factors such as perceived threat to professional autonomy, fear of replacement, concerns about patient safety, and legal liability of misdiagnosis hinder the widespread use of AI applications by cardiologists [8,9]. The challenge of a limited evidence base around AI in cardiac care is exacerbated as evidence base in this area dates rapidly due to changes in technology and attitudes.\u003c/p\u003e\n\u003cp\u003eThese issues raise several questions regarding the use of AI tools in AI health care. In this research, we address three questions below: Do cardiologists have a willingness to trust AI for the diagnosis of heart diseases and factors that predict this? How do clinicians perceive the potential consequences of using AI for their clinical autonomy and responsibility? How do contextual factors such as risk associated with the use of AI applications influence the intention to use AI?\u003c/p\u003e\n\u003ch2\u003eReview of conceptual models \u0026nbsp;\u003c/h2\u003e\n\u003cp\u003eFew frameworks are used to study the use of technology in health care. Technology adoption frameworks and trust theoretical frameworks present two major perspectives to study the use of AI tools in healthcare. In this research, we used the terms \u0026lsquo;AI tool\u0026rsquo; and \u0026lsquo;AI application\u0026rsquo; interchangeably, however \u0026lsquo;AI tool\u0026rsquo; is a technical term typically referring to the technologies or software libraries used to develop and implement AI solutions. In contrast, AI application refers to the systems and solutions that deploy AI technologies to perform analysis or identify patterns [10].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWithin technology adoption frameworks, a dominant framework is the Unified Theory of User Acceptance of Technology (UTAUT), which integrates eight prominent technology acceptance theories in the healthcare service adoption literature [11]. Building on this framework, Praksash et al. report that the future use of AI is determined by performance expectancy, effort expectancy, social influence, initial trust, and resistance to change [4].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe seminal framework of the Organisational Model of Trust developed by Mayer et al. [12] presents the second major perspective to study the use of AI tools. This framework provides a basis for most studies around trust in AI and similar technologies under the umbrella term \u0026apos;automation\u0026apos;. This framework was originally developed for studying organizational trust, but it has been well-received by researchers studying trust in AI and automation [13\u0026ndash;15]. It has six primary components, of which only one is trust itself and those others are antecedents, context, and product of trust. Factors of perceived trustworthiness, trust, propensity of trustors, risk-taking behaviour, perceived risks, and outcome are key components of the trust model [12]. A review study reports that most studies about trust in automation study variables which fit into one or more components of Mayer\u0026rsquo;s organisational model of trust [16]. Starke et al. assess three features of AI tools consisting of reliability, competence, and intention to assess physician trust in AI systems [17]. They emphasize the relevance of contextual factors and concerns in the use of AI systems; for instance dependence on AI, loss of professional autonomy, and the fear of losing control by adopting the AI systems that in the future may disrupt patient-physician relationships [17]. While helpful in identifying these factors, their framework provides no explicit top-level constructs, nor specifies the relationships between them.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn this research, we adapted the organizational model of trust to accommodate the research question of this study. Since most healthcare professionals have no or limited access to AI tools in their day-to-day practice, measurement of actual usage would lead to biased conclusions about the adoption of AI tools in healthcare. Intention rather than actual use would be an accurate outcome variable in the adoption model [18]. \u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eOur conceptual model and hypotheses\u003c/h2\u003e\n\u003cp\u003eThe current paper builds a conceptual model based on Mayer\u0026rsquo;s Organisational Model of Trust [19]. We slightly adapted the Mayer model of trust for our study and extended it by adding two constructs innovativeness and peer influence (Figure 1) based on previous studies [20]. The components of the trust model consist of the factors of perceived trustworthiness, trust, propensity of trustors, risk-taking behaviour, perceived risks, and outcome. This model starts with the factors of perceived trustworthiness which determine the level of trust. The model differentiates between trust (as an attitude) and outcome of trust (intention to use AI in the future, an intention). This distinction in our study implies that there might be trust in AI but it may not lead to the use of AI in cardiac care settings.\u003c/p\u003e\n\u003cp\u003eTrust is associated with risk-taking behaviour (in the current case, future use of AI systems, described in the original model of trust as risk-taking in relationships). The outcome (in the original model) refers to the consequences associated with the use of technology by cardiologists (i.e. clinical outcomes) which was not considered in this study. Instead of actual use of AI applications, we considered an intention to use AI in future. Below we describe our model\u0026rsquo;s components.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAbility, benevolence, and integrity determine perceived trustworthiness [19]. Ability refers to skills, quality, and characteristics of AI tools, which are context- and tool-specific and are believed to be significant contributors to the formation of trust [13,16,17]. In the context of AI in cardiac care settings, ability measures the accuracy of diagnosis made by AI applications, reliability, and whether AI applications produce results without breaching security. The second factor, benevolence, refers to a situation where an AI tool intends to do good for users. Integrity refers to situations where the AI tool adheres to a set of principles that the user finds acceptable. We postulate that ability (hypothesis (H)1), benevolence (H2), and integrity (H3) are positively associated with trust in AI.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs a propensity of cardiologists, knowledge of AI applications (both theoretical and practical) affects trust in AI. We postulate that knowledge of AI applications next to ability, benevolence, and integrity are associated with trust (H4). Trust in AI is the key intermediate outcome linked to usage.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBased on the Mayer model, trust itself determines risk-taking to use technology and thus intention to use [12]. We postulate that trust in AI has a positive impact on an intention to use AI (H5). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eLi et al. reported a direct association between social influence and trusting beliefs in information system settings [20]. They argue that when individuals have inadequate knowledge and lack hands-on experience with technology, it is highly likely that they rely on trusted others\u0026rsquo; opinions and incorporate their views into trust formation [21]. We postulate that peer influence (in terms of others being greater adopters) has a positive influence on the future use of AI applications (H6).\u003c/p\u003e\n\u003cp\u003ePersonal innovativeness refers to the willingness of an individual to try out an innovation. Studies support that individuals with a high degree of innovativeness are more likely to use computer-based applications and have positive perceptions about the ease of use of technology [22,23]. We postulate that personal innovativeness has a positive impact on the future use of AI applications (H7).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe \u0026lsquo;perceived risk\u0026rsquo; in the Mayer model of trust, refers to situational factors that, next to trust in AI, determine the use of AI. These may include how AI is perceived in working environments. The rise of AI tools and their potential have triggered fear of a substitution crisis (where valued roles and livelihoods are taken over by AI tools). In clinical settings, the capacity of AI tools to imitate human decision-making processes poses threats that skills and competencies possessed by physicians may be replaced by AI tools and thus can replace physicians [24]. We therefore postulate that perceived substitution threats have a negative influence on intention to use AI applications in future (H8).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAlongside these hypothesized relationships, we also aimed to explore what cardiologists felt were the key risks and benefits of the use of AI, and how they would likely respond in instances when their clinical judgement and that of an AI system diverged.\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eWe conducted a cross-sectional survey of consultant cardiologists in the UK. We used the SEM approach to validate a measurement model and to test our path model. We collected data from cardiologists (as a convenience sample) and included 61 participants in the final data analysis. Since our sample size was small, we used partial least square regression models to assess the measurement model and path analysis. We also used qualitative content analysis to elicit the main themes on perceived risks of using AI and the enablers and barriers of AI use in cardiac care. The study population consisted of consultant cardiologists in the UK (see below).\u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eVariables and measurement\u003c/h2\u003e \u003cp\u003eThe demographic and socio-economic variables measured comprised age groups, gender, ethnicity, and years of experience as a cardiologist. We classified participants based on their ages into three age groups: 30\u0026ndash;39, 40\u0026ndash;49, and 50\u0026ndash;70 years. We categorized participants into White, Asian/Asian British, Black/African/Caribbean/Black British, Mixed/Multiple Ethnic Groups, or Other Ethnic Group. Concerning the years of experience working as a cardiologist, study participants were categorized into four groups: 1\u0026ndash;5, 6\u0026ndash;10, 11\u0026ndash;15, or 16\u0026thinsp;+\u0026thinsp;years of experience. We also categorized participants based on the number of monthly SE images reviewed by them into four groups: 0\u0026ndash;20, 21\u0026ndash;50, 51\u0026ndash;100, and 101\u0026thinsp;+\u0026thinsp;images.\u003c/p\u003e \u003cp\u003eAttributes of cardiologists were explored through two questions that measured familiarity with AI applications and feeling knowledgeable about AI systems (Supplementary file 1. The ability of AI systems was assessed via outcome, security, and reliability items. This refers to the AI system's capacity to produce accurate results without breaching patient privacy and without breaking down during the process [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Benevolence was measured using three items referring to whether AI systems are deceptive, participants are suspicious of AI, and they are wary of AI systems. Integrity was measured using two items. Trust was formed by two items measuring whether participants have trust and confidence in AI systems. Attributes of cardiologist, ability, benevolence, and integrity were measured using a Likert scale of 1\u0026ndash;5.\u003c/p\u003e \u003cp\u003eWe provided questionnaire items to measure the perceived risk of using AI, innovativeness, and peer influence in Supplementary File 1. We measured the perceived risk of using AI [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] by measuring the adverse impact of using AI applications on confidence and trust between patients and doctors, adverse impacts on patient-doctor relationships, deskilling doctors\u0026rsquo; roles, and needs for cardiologists in a Likert scale of 1\u0026ndash;5. The innovativeness of cardiologists was measured by four items. Peer influence was measured through three items. The use of AI in the next 12 months was measured by a single item. Peer influence, innovativeness and future use of AI were measured using a Likert scale of 1\u0026ndash;7.\u003c/p\u003e \u003cp\u003eParticipants described the potential risks of following the AI\u0026rsquo;s recommendation and further actions in response to two hypothetical scenarios: A) Where AI contradicts the clinician\u0026rsquo;s diagnosis of CAD being present; and B) AI contradicts the clinician's diagnosis of CAD not being present. We also asked participants to determine the extent of confidence in their initial diagnosis to retain it without taking further actions in response to the above scenarios. We also asked participants to describe barriers and enablers of using AI decision-support software in cardiac care.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eQuestionnaire and data collection\u003c/h2\u003e \u003cp\u003eWe used questionnaires for data collection. Demographic questions were developed based on a framework proposed by UK\u0026rsquo;s office for National Statistics [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. We used 12-item instrument developed by Jian et al. to measure trust in AI and the factors of perceived trustworthiness of AI applications [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. We also embark on the framework proposed by Starke et al. to incorporate questions on the substitutions threats of using AI, such as threats to patient-doctor relationship [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. For items measuring the dimensions of innovativeness and peer influence, we used instrument developed Fan et al. [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe collected data through an online survey tool (Qualtrics). We applied screening questions to identify eligible participants i.e., consultant cardiologists. These included minimum age expected to start a consultant cardiologist career, and working in a cardiac care setting. 88 completed responses were collected, of which 27 were excluded due to ineligibility. We considered 61 participants in the final analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis methods\u003c/h2\u003e \u003cp\u003eWe conducted the descriptive analysis of sociodemographic and outcome parameters using summary statistics (mean and SD). We used the structural equation modelling (SEM) approach for the path analysis. An SEM model encompassed two components: a measurement model to operationalize the conceptual framework of trust and the future use of AI and a structural model to describe coefficients of relationships between constructs. We evaluated multiple measurement models and compared their performance in terms of goodness of fit (GoF) and other indices to reach the final measurement model for SEM. The SEM analysis relied on the partial least square (PLS) regression model, which is a suitable method for studies with small sample sizes [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. We iterated model selection based on the loading of variables on a construct, cross-loading, and communalities. A factor loading presents correlations between a variable (i.e. a scale item) and a latent construct (e.g., trust in AI). Cross loading refers to loadings of variables on the rest of latent constructs i.e. constructs other than the hypothesized construct. Communalities represent the amount of variability in a latent construct explained by a variable. Variables with loadings lower than 0.7 were removed from the model.\u003c/p\u003e \u003cp\u003eThe quality of the structural model was evaluated using regression equations and by \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e coefficient, the redundancy index, and GoF. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e indicates the amount of variance in latent constructs (trust and future use of AI) explained by their independent items. The values for \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e were classified into low (R\u0026thinsp;\u0026lt;\u0026thinsp;0.30), moderate (0.30\u0026thinsp;\u0026lt;\u0026thinsp;R\u0026thinsp;\u0026lt;\u0026thinsp;0.60), or high (R\u0026thinsp;\u0026gt;\u0026thinsp;0.60) [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Redundancy refers to the ability of the independent variables to explain variation in trust and future use of AI. The average communality indicates how much of each construct\u0026rsquo;s variability is reproducible by its latent constructs or items. GoF accounts for both the measurement and the structural model quality. GoF is calculated as the mean of the communality and the average \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\varvec{R}}^{2}\\)\u003c/span\u003e\u003c/span\u003evalue. GoF higher than 70% is perceived as acceptable [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe used the thematic analysis method for qualitative analysis of participants' responses to scenarios A and B as well as the enablers and barriers of using AI applications [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. We followed six steps for this analysis: 1) familiarization with data, 2) generating initial codes, 3) searching for themes, 4) reviewing themes, 5) defining and naming themes, and 6) producing a report. We took an inductive analysis approach, with the themes identified strongly linked to the data themselves.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eMost participants were male (85%). The age of participants ranged from 30 to 70 years; 46% of participants were aged between 40\u0026ndash;49 years. The largest ethnicity group was White comprising 64% of participants (see Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e for other ethnicities). Most participants have worked as a cardiologist for longer than five years (Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). More than half of participants (54%) reported current use of AI in a cardiac care pathway (Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eDemographic characteristics of the survey participants\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDemographic information\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFrequency\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePercent (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e85\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge groups\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30\u0026ndash;39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40\u0026ndash;49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e46\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50\u0026ndash;70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthnicity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWhite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAsian/Asian British\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack/African/Caribbean/Black British\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMixed/Multiple Ethnic Groups\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther Ethnic Group\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eYears working as a cardiologist\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u0026ndash;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u0026ndash;10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11\u0026ndash;15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e16+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eExperience of cardiologists on the use of AI applications in cardiac care and reviewing stress echocardiography tests\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAI in the cardiac care pathway\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFrequency\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePercent (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCurrently using AI tools in cardiac care\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e44\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUnknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCurrently using auto-indexing software in cardiac care\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e46\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCurrently using Decision Support Software in cardiac care\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e38\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e62\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eNumber of monthly stress echo tests reviewed\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u0026ndash;20 tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u0026ndash;50 tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e48\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e51\u0026ndash;100 tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e101\u0026thinsp;+\u0026thinsp;tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eThe measurement model assessment\u003c/h2\u003e\n \u003cp\u003eThe results of assessing the final measurement models are given in Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e. In the final model, all loadings were greater than 0.7 and therefore acceptable. A loading greater than 0.7 means that \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({0.7}^{2}=50\\%\\)\u003c/span\u003e\u003c/span\u003e of the variability in a latent construct is captured by a variable. Lastly, since we only have a single-item scale for the integrity of AI, substitution threat, peer influence, innovativeness, and the use of AI in the next 12 months, the factor loadings and communality for them are 1. Compared with the original conceptual model, the number of items per construct in the final model was markedly reduced, yet the final model encompassed the main constructs suggested by the conceptual framework.\u003c/p\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eAssessment of final measurement model\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eWeight\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLoading\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCommunality\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRedundancy\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAttributes of cardiologist\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eFamiliarity with AI system\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eFeeling knowledgeable about AI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eThe ability of AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSecurity of AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eReliability of AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBenevolence\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBeing wary of AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerceived harmfulness of AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerceived integrity of AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerceived trust in AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerceived trust in AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerceived confidence in AI systems\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubstitution threat\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePeer influence\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eInnovativeness\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eUse of AI in the next 12 months\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003eThe cross-loading analysis (Supplementary File 1) showed that the loadings of all variables/items on their construct were higher than 0.7 and no variable loaded on constructs other than the corresponding construct.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003eThe structural model assessment\u003c/h2\u003e\n \u003cp\u003eThe quality of the structural model was evaluated using PLS regression and regression coefficient, the redundancy index, and GoF. As shown in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e, ability (\u0026beta;\u0026thinsp;=\u0026thinsp;0.55, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), benevolence (\u0026beta;\u0026thinsp;=\u0026thinsp;0.19, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05), and attributes of cardiologists (\u0026beta;\u0026thinsp;=\u0026thinsp;0.25, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05) were significant contributors to trust in AI tool. Therefore, hypotheses H1, H2, and H4 were supported. The perceived ability of the AI tool made the largest contribution to trust; one unit (on a Likert scale between 1 and 5) increase in ability was associated with a 0.55 unit (scale between 1 and 5) increase in trust. Lower benevolence (i.e., lower level of good intention of AI) was associated with lower trust. H3 postulating the relationship between integrity and trust was not supported.\u003c/p\u003e\n \u003cp\u003eAmong the constructs that were used to explain the future use of AI, only trust in AI was a significant contributor (\u0026beta;\u0026thinsp;=\u0026thinsp;0.48, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), with higher trust being linked to increased future use. Therefore, H5 was supported. Other hypotheses postulating relationships between peer influence, innovativeness and perceived risk of AI and the use of AI (H6, H7, and H8 respectively) were not supported.\u003c/p\u003e\n \u003cp\u003eFor the structural model, we evaluated the quality of the PLS model of trust and the future use of AI based on the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e coefficient (Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e indicated that 70% of the variance in trust can be explained by attributes of cardiologist, ability, benevolence, and integrity. Only 37% of the variance in the future use of AI could be explained by its independent latent constructs. In terms of classification used to assess the regression models, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({R}^{2}\\)\u003c/span\u003e\u003c/span\u003e for trust was high and for the future use of AI was moderate [\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e\n \u003ctable id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eQuality indicators for structural model assessment\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eConstruct\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\varvec{R}}^{2}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCommunality\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRedundancy\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAttributes of cardiologist\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAbility\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBenevolence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntegrity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTrust\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.59\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePerceived risk\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePeer influence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eInnovativeness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFuture use of AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003eThe value of redundancy indicated that the dependent latent variables of trust (ability, benevolence, integrity, and attributes of cardiologists) explain around 59% of the variability in trust (Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). Latent variables including trust, perceived risk, peer influence, and innovativeness explained 37% of the variability in the future use of AI.\u003c/p\u003e\n \u003cp\u003eThe average communality indicated that two variables of attributes of cardiologists reproduced 87% of the variability in attributes of cardiologists. 69%, 66%, and 84% of the variability in ability, benevolence, and trust respectively were reproducible by their variables. Based on GoF, the predictive power of the model is 64%.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003eScenario analysis\u003c/h2\u003e\n \u003cp\u003eIn a situation where the cardiologist made a diagnosis of CAD being present or not present, but an AI tool contradicted this, the mean values of confidence in their diagnosis were 69% and 65%, respectively. The results of regression analysis testing the relationships between mean values of confidence in own diagnosis and years of experience as a cardiologist, number of monthly images reviewed, and current use of AI in cardiac care pathway are given in Table \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e\n \u003ctable id=\"Tab5\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eRegression models for scenario analysis\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eScenario A\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eScenario B\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModel\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026beta;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStd error\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026beta;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStd error\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntercept\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e35.69*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e58.38*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15.51\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYears of experience as a cardiologist\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u0026ndash;5 (reference)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u0026ndash;10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-1.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11\u0026ndash;15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e10.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7.48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-1.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.85\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e16+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.50\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNumber of monthly stress test reviews\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u0026ndash;20 tests (reference)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u0026ndash;50 tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-11.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.07\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e51\u0026ndash;100 tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-8.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10.54\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e101\u0026thinsp;+\u0026thinsp;tests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-3.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAI in the cardiac care pathway\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes (reference)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-2.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTrust\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7.73*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.67\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003eNote: * = \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003cp\u003eAs shown in Table \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e we found that confidence in own diagnosis was not associated with years of experience, the number of monthly images reviewed, or the current use of AI in the cardiac care pathway. However, when cardiologists had a higher trust in AI tools, they would need to have higher confidence in their own diagnosis to choose not to embark on further tests or seek a second opinion.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003ePerceived risk of and further actions taken following AI recommendations\u003c/h2\u003e\n \u003cp\u003eFour main themes emerged during the content analysis of the perceived risk of following AI recommendations. The risk of inaccurate diagnosis or treatment was the most repeated theme (n\u0026thinsp;=\u0026thinsp;20 instances) (Supplementary file 1). The second theme was the risk of adverse events (n\u0026thinsp;=\u0026thinsp;15). The risk of litigation and legal responsibility was also mentioned (n\u0026thinsp;=\u0026thinsp;6). Lastly, some participants indicated that they thought AI was only a decision-support tool and there was no risk (n\u0026thinsp;=\u0026thinsp;2). These numbers refer to the sum of themes in scenarios A and B (Supplementary File 1).\u003c/p\u003e\n \u003cp\u003eIn terms of further actions taken when AI contradicts cardiologist judgement, the most repeated theme was the need to embark on further tests, imaging, scans or examinations (n\u0026thinsp;=\u0026thinsp;61 times). The second most mentioned action was to verify or confirm evidence/information through wider assessment (n\u0026thinsp;=\u0026thinsp;27). The third most repeated action was to ask for a second opinion or to consult with a peer (n\u0026thinsp;=\u0026thinsp;26). In a small number of cases, participants reported taking no action (n\u0026thinsp;=\u0026thinsp;10).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eEnablers and barriers to using AI\u003c/h2\u003e\n \u003cp\u003eOur analysis of enablers and barriers to using AI applications identified multiple themes that were iterated as both enablers and barriers (Supplementary file 1). The theme \u0026lsquo;improving accuracy\u0026rsquo; was the most repeated enabler (n\u0026thinsp;=\u0026thinsp;39 times). The second most repeated enabler was \u0026lsquo;saving time and more speedy diagnosis\u0026rsquo; (n\u0026thinsp;=\u0026thinsp;27). And the third most repeated enabler was \u0026lsquo;cost saving and efficiency\u0026rsquo; (n\u0026thinsp;=\u0026thinsp;18) and \u0026lsquo;ease of use and automation\u0026rsquo; (n\u0026thinsp;=\u0026thinsp;18).\u003c/p\u003e\n \u003cp\u003eThe most repeated barrier was the concern over the accuracy of diagnosis (n\u0026thinsp;=\u0026thinsp;29). The second most mentioned barrier was concerns over costs and efficiency (n\u0026thinsp;=\u0026thinsp;26). The third most repeated barrier was concern over time (n\u0026thinsp;=\u0026thinsp;10) related to the production of AI reports may take time and/or AI reports may arrive late).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study aimed to explore trust in AI applications in cardiac care among cardiologists, based on an adapted version of Mayer\u0026rsquo;s model of organizational trust [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. The findings present mixed support for the hypotheses generated by this model. Our findings indicate that two factors of trustworthiness (ability and benevolence) and propensity of trustors have significant associations with trust in AI. The positive influence of trust on the future use of AI aligns with existing evidence emphasizing the importance of trust in fostering acceptance and adoption of technological innovations [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. The overall SEM structure confirms established evidence that trust is understood by its determinants and that it is positioned between factors of trustworthiness and AI use [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe positive impact of the AI system\u0026rsquo;s ability emphasizes the pivotal role of technical competence i.e., reliability and security, in the formation of cardiologists' trust in such systems. This finding aligns with the already recognized importance of security in building trust in AI [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Prior knowledge of and familiarity with AI were also identified as significant contributors to trust in AI (H4).\u003c/p\u003e \u003cp\u003eIntegrity was not associated with trust, which is in contrast with Mayer's model, where it is a key factor of trustworthiness [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. AI applications offering a higher level of integrity tend to build higher subjective trust [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eContrary to our hypotheses and findings reported by other studies [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], peer influence, perceived risk of using AI, and personal innovativeness did not emerge as significant factors influencing the future use of AI. Cardiologists\u0026rsquo; intention to use AI was not significantly influenced by peer influence (operationalized by the amount that colleagues used AI). This is in contrast with existing findings around the positive role of important people in shaping attitudes towards AI adoption within professional settings [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. This may be because the question focused on increased AI use, and the sample was broadly positive about AI.\u003c/p\u003e \u003cp\u003eOur findings around the lack of relationships between the intention to use AI and peer influence and innovativeness of cardiologists are in contrast with UTUAT which suggests these components as key determinants of the technology use [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. However, this might be rooted in the fact that we incorporated these two components from UTUAT which is a distinctive conceptual framework from Mayer\u0026rsquo;s trust model, and used a different set of items to capture this dimension. In the context of a study designed based on UTUAT, relationships between peer influence and innovativeness with the use of technology may be significant. However, it does suggest that the context of measurement is important for UTUAT. Our findings therefore provide empirical grounds for studying trust and intention to use AI, responding to the call for an empirical-based conceptual framework of trust in AI in healthcare [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOur SEM analysis only explained a modest extent of variance in the intention to use AI, and only some predictors were statistically significant. These findings do not contradict the validity of the Mayer model of trust since we did not measure risk-taking behaviour and outcomes of trust according to the original trust model. Rather they suggest that while elements of Mayer's model are important (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), predictors of AI may be more nuanced or situation-specific. To address this in the context of cardiac AI, we conducted a qualitative assessment of the risks (and benefits) of following AI recommendations, especially in situations where the AI recommendation diverges from that of clinicians. Future research needs to embark on full-fledged operationalization of the trust model in this context or extend SEM models to explain a large extent of unexplained variance in the intention to use AI.\u003c/p\u003e \u003cp\u003eThe implications of this study are far-reaching for both researchers and the cardiac care community. Our study provided evidence around the multiple dimensions related to trust in AI and its adoption. These are likely crucial for designing trustworthy AI and facilitating the widespread adoption of AI in cardiac care (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThis study has some limitations. Our sample size (n\u0026thinsp;=\u0026thinsp;61) was relatively small which has implications for the generalizability of findings within the cardiologist community and beyond. Caution should be exercised when extrapolating findings to the broader medical community or to non-cardiologist healthcare professionals. Furthermore, females and non-white ethnic groups were under-represented in our sample, consequently, our findings might not adequately reflect the perceptions of female cardiologists and other ethnic communities. Due to the small sample size, achieving optimal model fit became challenging; yet the GoF of our final SEM was close to the optimal level (70%) and our analysis identified alternative models that improved overall model fit. The sample size needs to be cautiously viewed as the British consultant cardiologist community is small (~\u0026thinsp;2,600 members), meaning we sampled only 2.3% of the population.\u003c/p\u003e \u003cp\u003eWe used the PLS method for SEM analysis which is developed to deal with shortcomings of small samples, but the small sample might have affected detecting true effects within the model. However, the observation that we did not face wide confidence intervals and model identification issues does provide some reassurance that the sample was sufficient.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThe enhanced GoF in the final model indicates an improved final SEM model compared with the original SEM model despite the reduction in the number of items included in the final model. Addressing specific concerns related to accuracy and adverse events, while fostering trust in AI systems, will be essential for ensuring the seamless integration of AI applications into routine clinical practices. In contrast to previous studies findings, effects of peer influence, perceived substitution threats, and innovativeness on the use of AI applications were not found in this study (4 out of 8 hypotheses in this research were not supported). However, ability, benevolence, and previous familiarity and knowledge of AI were found to be key factors in trust in AI. The subsequent use of AI was associated with trust itself rather than contextual factors such as peer influence, personal innovativeness or perceived risk of AI.\u003c/p\u003e \u003cp\u003eThe Integration of quantitative and qualitative data provided a comprehensive understanding of the complex interplay of factors influencing cardiologists'\u0026rsquo; trust in AI and the use of AI in the future. The study findings have practical implications for healthcare organizations aiming to implement AI technologies successfully, and for technology developers to ensure core concerns are reassured through robust design and testing. Addressing specific concerns related to accuracy, costs, and time, while fostering trust in AI systems, will be essential for ensuring the seamless integration of these technologies into routine clinical practices.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAI: artificial intelligence. SE: Stress echocardiography. CAD: coronary artery disease. SEM: structural equation model. UTAUT: Unified Theory of User Acceptance of Technology. H: hypothesis. GoF: goodness of fit. PLS: partial least square.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eThis research has received a favourable opinion from the NHS Health Research Authority (IRAS No: 315284) and approval from the London South Bank University Ethics Panel (ETH2223-0164). Informed Consent was obtained from the participants in the study before giving access to study questionnaires or collecting any data. We did not collect any personal data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe dataset generated and analysed during the current study is available in the Open Science Framework at osf.io/eknfz [35]. All other materials relevant to this study including method statements and analysis plan are available at the same place.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis work was supported by NHSx and NHS Accelerated Access Collaborative grant number REI2122-0030 (AI Award in Health and Care, Technology Specific Evaluation Team). This report is independent research and the views expressed in this publication are those of the authors and not necessarily those of the NHS, NHSX or the Department of Health and Social Care.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAuthors\u0026apos; contributions\u003c/p\u003e\n\u003cp\u003eDF acquired the funding for the study. MM and DJF conceptualised the study, developed the study design, and implemented the study. MM analysed the data and drafted the manuscript. DJF critically reviewed the manuscript. Both authors approved the final version of the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eWe would like to express our sincere gratitude to all individuals and organizations who contributed to the completion of this paper. Special thanks to Virgil Griffith and Sharanya Jayadev for their support and assistance throughout the research process. Additionally, we extend our appreciation to the participants who generously shared their insights and experiences, without whom this study would not have been possible.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eO\u0026rsquo;Driscoll JM, Hawkes W, Beqiri A, et al. Left ventricular assessment with artificial intelligence increases the diagnostic accuracy of stress echocardiography. Eur Heart J Open. 2022;2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ehjopen/oeac059\u003c/span\u003e\u003cspan address=\"10.1093/ehjopen/oeac059\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRasheed K, Qayyum A, Ghaly M, et al. Explainable, trustworthy, and ethical machine learning for healthcare: A survey. Comput Biol Med. 2022;149. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compbiomed.2022.106043\u003c/span\u003e\u003cspan address=\"10.1016/j.compbiomed.2022.106043\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShinners L, Aggar C, Grace S, et al. Exploring healthcare professionals\u0026rsquo; understanding and experiences of artificial intelligence technology use in the delivery of healthcare: An integrative review. Health Inf J. 2020;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/1460458219874641\u003c/span\u003e\u003cspan address=\"10.1177/1460458219874641\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrakash AV, Das S. Medical practitioner\u0026rsquo;s adoption of intelligent clinical diagnostic decision support systems: A mixed-methods study. Inform Manage. 2021;58. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.im.2021.103524\u003c/span\u003e\u003cspan address=\"10.1016/j.im.2021.103524\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlostad B, Karnik A, Appadurai V, et al. Applications of Artificial Intelligence in Echocardiography. Curr Cardiovasc Risk Rep. 2023;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12170-023-00721-6\u003c/span\u003e\u003cspan address=\"10.1007/s12170-023-00721-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaad M, Kotecha K, Patil K, et al. Cardiac Diagnosis with Machine Learning: A Paradigm Shift in Cardiac Care. Appl Artif Intell. 2022;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/08839514.2022.2031816\u003c/span\u003e\u003cspan address=\"10.1080/08839514.2022.2031816\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhalsa RK, Khashkhusha A, Zaidi S, et al. Artificial intelligence and cardiac surgery during COVID-19 era. J Card Surg. 2021;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/jocs.15417\u003c/span\u003e\u003cspan address=\"10.1111/jocs.15417\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJussupow E, Spohrer K, Dibbern J et al. AI changes who we are - Doesn\u0026rsquo;t it? Intelligent decision support and physicians\u0026rsquo; professional identity. 26th European Conference on Information Systems: Beyond Digitization - Facets of Socio-Technical Change, ECIS. 2018. 2018.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHengstler M, Enkel E, Duelli S. Applied artificial intelligence and trust-The case of autonomous vehicles and medical assistance devices. Technol Forecast Soc Change. 2016;105. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.techfore.2015.12.014\u003c/span\u003e\u003cspan address=\"10.1016/j.techfore.2015.12.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRussell S, Norivg P, Artificial Intelligence. A Modern Approach (Global Edition). Artificial Intelligence: A Modern Approach. 2021.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliams MD, Rana NP, Dwivedi YK. The unified theory of acceptance and use of technology (UTAUT): A literature review. J Enterp Inform Manage. 2015;28. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/JEIM-09-2014-0088\u003c/span\u003e\u003cspan address=\"10.1108/JEIM-09-2014-0088\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer RC, Davis JH, David Schoorman F. An Integrative Model of Organizational Trust. 1995. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.jstor.org/stable/258792?seq=1\u0026amp;cid=pdf-\u003c/span\u003e\u003cspan address=\"https://www.jstor.org/stable/258792?seq=1\u0026amp;cid=pdf-\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiun-Yin Jian, Ann M, Bisantz CGD. Foundations for an Empirically Determined Scale of Trust in Automated Systems. Int J Cogn Ergon. 2000;4:53\u0026ndash;71.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSousa S, Martins P, Cravino J. Measuring Trust in Technology: A Survey Tool to Assess Users\u0026rsquo; Trust Experiences. 29TH INTERNATIONAL CONFERENCE ON INFORMATION SYSTEMS DEVELOPMENT. 2021:2\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWojton HM, Porter D, Lane T. Initial validation of the trust of automated systems test (TOAST). J Soc Psychol. 2020;160. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/00224545.2020.1749020\u003c/span\u003e\u003cspan address=\"10.1080/00224545.2020.1749020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKohn SC, de Visser EJ, Wiese E, et al. Measurement of Trust in Automation: A Narrative Review and Reference Guide. Front Psychol. 2021;12:4138.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStarke G, van den Brule R, Elger BS, et al. Intentional machines: A defence of trust in medical artificial intelligence. Bioethics. 2022;36:154\u0026ndash;61.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFan W, Liu J, Zhu S, et al. Investigating the impacting factors for the healthcare professionals to adopt artificial intelligence-based medical diagnosis support system (AIMDSS). Ann Oper Res. 2020;294. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10479-018-2818-y\u003c/span\u003e\u003cspan address=\"10.1007/s10479-018-2818-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer RC, Davis JH, Schoorman FD. An Integrative Model Of Organizational Trust. Acad Manage Rev. 1995;20:709\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Hess TJ, Valacich JS. Using Attitude and Social Influence to Develop an Extended Trust Model for Information Systems. Data Base Adv Inform Syst. 2006;37. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1145/1161345.1161359\u003c/span\u003e\u003cspan address=\"10.1145/1161345.1161359\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Hess TJ, Valacich JS. Why do we trust new technology? A study of initial trust formation with organizational information systems. J Strategic Inform Syst. 2008;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jsis.2008.01.001\u003c/span\u003e\u003cspan address=\"10.1016/j.jsis.2008.01.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO\u0026rsquo;Connor Y, O\u0026rsquo;Connor S, Heavin C, et al. Sociocultural and Technological Barriers Across all Phases of Implementation for Mobile Health in Developing Countries. Appl Comput Med Health. 2015. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/B978-0-12-803468-2.00010-2\u003c/span\u003e\u003cspan address=\"10.1016/B978-0-12-803468-2.00010-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeong SC, Choi BJ. Moderating Effects of Consumers\u0026rsquo; Personal Innovativeness on the Adoption and Purchase Intention of Wearable Devices. Sage Open. 2022;12:1\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTran AQ, Nguyen LH, Nguyen HSA, et al. Determinants of Intention to Use Artificial Intelligence-Based Diagnosis Support System Among Prospective Physicians. Front Public Health. 2021;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fpubh.2021.755644\u003c/span\u003e\u003cspan address=\"10.3389/fpubh.2021.755644\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStarke G, van den Brule R, Elger BS, et al. Intentional machines: A defence of trust in medical artificial intelligence. Bioethics. 2022;36:154\u0026ndash;61.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOffice for National Statistics. Demography question development for Census 2021. 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJian J-Y, Bisantz AM, Drury CG. Foundations for an Empirically Determined Scale of Trust in Automated Systems. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/101207/S15327566IJCE0401_04\u003c/span\u003e\u003cspan address=\"http://dx.doi.org/101207/S15327566IJCE0401_04\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. 2010;4:53\u0026ndash;71.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStarke G, Brule R, Elger BS, et al. Intentional machines: A defence of trust in medical artificial intelligence. Bioethics. 2022;36:154\u0026ndash;61.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHair JF, Sarstedt M, Hopkins L, et al. Partial least squares structural equation modeling (PLS-SEM): An emerging tool in business research. Eur Bus Rev. 2014;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/EBR-10-2013-0128\u003c/span\u003e\u003cspan address=\"10.1108/EBR-10-2013-0128\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarke V, Braun V, Hayfield N. Thematic analysis. Qualitative psychology: A practical guide to research methods. 2015;222:248.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGille F, Jobin A, Ienca M. What we talk about when we talk about trust: Theory of trust for AI in healthcare. Intell Based Med. 2020;1\u0026ndash;2:100001.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteerling E, Siira E, Nilsen P, et al. Implementing AI in healthcare\u0026mdash;the relevance of trust: a scoping review. Front Health Serv. 2023;3. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/frhs.2023.1211150\u003c/span\u003e\u003cspan address=\"10.3389/frhs.2023.1211150\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnson KC, Li IARKK. Knowledge is not all you need to generate trust in AI use in healthcare. medRxiv. 2024;01.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMehrotra S, Centeio Jorge C, Jonker CM et al. Building Appropriate Trust in AI: The Significance of Integrity-Centered Explanations. Frontiers in Artificial Intelligence and Applications. 2023. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3233/FAIA230121\u003c/span\u003e\u003cspan address=\"10.3233/FAIA230121\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahdavi Mahdi FDJS. Evaluating Artificial Intelligence Driven Stress Echocardiography Analysis System (EASE Study). WP4: Trust in AI and Future Use of AI. 2024. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eosf.io/eknfz\u003c/span\u003e\u003cspan address=\"http://osf.io/eknfz\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (accessed 15 March 2024).\u003c/span\u003e\u003c/li\u003e \u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Trust, Artificial Intelligence, Cardiologist, Cardiac Care, Intention to Use, Technology Adoption","lastPublishedDoi":"10.21203/rs.3.rs-4114716/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4114716/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e. The widespread use of Artificial Intelligence (AI)-driven applications among consultant cardiologists remains relatively low due to trust issues and perceived threat to professional autonomy, patient safety, and legal liability of misdiagnoses. There is a paucity of empirical research investigating the relationships between trust in AI applications and an intention to use (AI-Use) them among cardiologists. To address this gap, we surveyed a sample of cardiologists to examine the determinants of trust in AI and trust’s effects on AI-Use based on the organisational trust model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e. We conducted a cross-sectional survey of consultant cardiologists (n = 61) in the UK. Given the small sample size, we used a partial least square structural equation model (SEM) analysis approach to assess the measurement and structural models. We utilized factor loadings and weights for the measurement model assessment and coefficients, the redundancy indices, and goodness of fit (GoF) for the structural model assessment. We also undertook a content analysis of open-text responses around perceived risks, enablers, and barriers to AI use in cardiac care. We performed analyses in the R programme.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e. The GoF of the final SEM model was 63%, showcasing a substantial improvement over the original model (GoF=51%). The final model encompassed all latent constructs from the original model and explained 70% of the variance in trust and 37% in AI use. The AI application ability (accuracy and reliability) significantly influenced trust (β=0.55, p\u0026lt;.001), while lower benevolence correlated with decreased trust (β=0.19, p\u0026lt;.05). Trust in AI emerged as the sole significant contributor to AI-Use (β=0.48, p\u0026lt;.001), indicating higher trust associated with increased future use. Participants perceived diagnosis accuracy as a prominent theme, mentioned 20 times about AI risk and frequently cited as both an enabler (n=39 times) and a barrier (n=29 times).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e. The enhanced GoF in the final model indicates an improved final SEM model compared with the original SEM model. Addressing diagnosis accuracy concerns and building trust in AI systems is crucial to facilitate increased AI adoption among cardiologists and seamless integration into cardiac care.\u003c/p\u003e","manuscriptTitle":"Trust in AI applications and intention to use them in cardiac care among cardiologists in the UK: A Structural Equation Modeling Approach","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-26 15:47:45","doi":"10.21203/rs.3.rs-4114716/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f4035d0b-5dba-4fef-8a1c-4a4604a58edd","owner":[],"postedDate":"March 26th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-07-12T08:27:23+00:00","versionOfRecord":[],"versionCreatedAt":"2024-03-26 15:47:45","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4114716","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4114716","identity":"rs-4114716","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}