Combining topic modeling and bibliometric analysis to understand the evolution of technological innovation adoption in the healthcare industry

Abstract

Purpose

This paper explores the Adoption of Technological Innovation (ATI) in the healthcare industry. It investigates how the literature has evolved, and what are the emerging innovation dimensions in the healthcare industry adoption studies.

Design/methodology/approach

We followed a mixed-method approach combining bibliometric methods and topic modeling, with 57 papers being deeply analyzed.

Findings

Our results identify three latent topics. The first one is related to the digitalization in healthcare with a specific focus on the COVID-19 pandemic. The second one groups up the word combinations dealing with the research models and their constructs. The third one refers to the healthcare systems/professionals and their resistance to ATI.

Research limitations/implications

The study’s sample selection focused on scientific journals included in the Academic Journal Guide and in the FT Research Rank. However, the paper identifies trends that offer managerial insights for stakeholders in the healthcare industry.

Practical implications

ATI has the potential to revolutionize the health service delivery system and to decentralize services traditionally provided in hospitals or medical centers. All this would contribute to a reduction in waiting lists and the provision of proximity services.

Originality/value

The originality of the paper lies in the combination of two methods: bibliometric analysis and topic modeling. This approach allowed us to understand the ATI evolutions in the healthcare industry.

Citation

Cobelli, N. and Blasi, S. (2024), "Combining topic modeling and bibliometric analysis to understand the evolution of technological innovation adoption in the healthcare industry", European Journal of Innovation Management, Vol. 27 No. 9, pp. 127-149. https://doi.org/10.1108/EJIM-06-2023-0497

1. Introduction

The rapid increase of technology has catalyzed a profound fusion between medicine and cutting-edge technologies, such as digitalization, artificial intelligence and the internet of Things (IoT), revolutionizing the healthcare sector (Tani et al., 2022). These innovations are driving improvements in diagnostics, personalized treatment options, remote patient monitoring and telemedicine, ultimately creating a more efficient, accessible and patient-centered healthcare future (Ciasullo et al., 2022). IoT, in particular, stands at the forefront of this transformative process, delivering efficiency improvements and cost reductions, all the while emphasizing the enhancement of patient care through features like continuous monitoring, tracking, secure storage of vital statistics and medical information (Cannavale et al., 2022). Furthermore, healthcare services are no longer confined to traditional clinical settings, being accessible in various environments. Wearable medical devices equipped with sensors play a new and fascinating role. These devices collect essential health data from the human body, measuring key health indicators (Vargheese and Viniotis, 2014). This technological integration is paving the way for a more interconnected and technologically driven healthcare landscape (Tani et al., 2022).

However, despite the increasing emphasis on implementing new knowledge in practical healthcare contexts, prior research has underscored the adoption of technology innovation (ATI) at the organizational and systemic levels (Godfrey et al., 2023). A deeper investigation into the necessity to explore the adoption of technological innovation in the healthcare industry is imperative to address these challenges comprehensively.

Venkatesh et al. (2003, 2012) investigated the variables affecting technology adoption in the business-to-consumer and business-to-business markets and developed the Unified Theory of Acceptance and Use of Technology (UTAUT) and the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2).

This paper, starting with a comprehensive literature review and employing topic modeling techniques to explore UTAUT and UTAUT2, aims to gain a deeper understanding of the prevailing research trends, key themes and emerging insights in the context of technology acceptance and adoption in the healthcare industry. Accordingly, our research questions are:

RQ1.

How has the literature evolved in relation to technological acceptance in the healthcare industry, with a particular emphasis on the research models UTAUT and UTAUT2?

RQ2.

What are the emerging innovation dimensions investigated in the healthcare industry?

To address this research questions, we first conduct a comprehensive bibliometric analysis, and second, we employ a topic modeling technique. A bibliometric analysis is a technique for identifying patterns and trends in a volume of literature. To assess publications, authors, journals or research institutes, as well as their relationships, it entails statistical analysis of bibliographic data, including citations. By giving insights into the significance and influence of research output, this approach aids in measuring and assessing the academic contributions made within a particular field and identify the steps forward research can and should do. Topic modeling was then chosen since it allows to identify and track emerging trends, also in healthcare industry (Jayaraman et al., 2020), providing insights into which areas require further exploration. Additionally, through an examination of the distribution of topics in the literature, topic modeling can pinpoint knowledge gaps and areas that have been underexplored. This information proves valuable for researchers aiming to direct their efforts toward uncharted territory and for policymakers looking to tackle overlooked healthcare challenges.

2. Theory

The landscape of technology adoption and use is underpinned by a range of influential models, each offering distinct advantages and encountering their own set of disadvantages. The Theory of Reasoned Action (TRA) provides a structured foundation for understanding individual attitudes and subjective norms (Fishbein, 1979), while the Technology Acceptance Model (TAM) and the Motivational Model delve into user motivations and perceptions (Davis et al., 1992), facilitating effective system design and evaluation. The Theory of Planned Behavior (TPB) extends TRA by incorporating Perceived Behavioral Control, enabling better predictions for behaviors that require personal control (Ajzen, 1991). Combining TAM and TPB offers a more comprehensive view, capturing both motivational and control aspects (Taylor and Todd, 1995). Innovation Diffusion Theory guides understanding of the adoption process and the influence of early adopters and opinion leaders (Rogers and Williams, 1983), while the Social Cognitive Theory emphasizes observational learning and self-regulation in technology adoption (Compeau et al., 1999). However, these models may oversimplify the complexities of real-world technology adoption, potentially lacking relevance in diverse cultural and contextual settings and facing challenges from rapid technological advancements.

The Unified Theory of Acceptance and Use of Technology (UTAUT) has emerged as an influential model that builds upon and intends to overcome limitations of the earlier technology adoption models (Gupta et al., 2008).

The UTAUT model incorporates multiple factors overcoming the oversimplification of the previous models. In details, Venkatesh et al. (2003) proposes four pivotal dimensions influencing information technology intention and usage. The first is Performance Expectancy, meaning the extent to which an individual believes that using the system will assist them in improving job performance. The second dimension is the Effort Expectancy, which refers to the perceived ease with which the system can be used. The third dimension, Facilitating Conditions represents the extent to which a person believes that an organizational and technical infrastructure exists to support system use. The fourth dimension, Social Influence, represents the extent to which a person believes that others believe they should use the new system. UTAUT2 combines the above-described dimensions with three new ones (Venkatesh et al., 2012). Hedonic Motivation is defined as the enjoyment or pleasure obtained from utilizing a system; Price Value is as the cognitive tradeoff between the technology apparent advantages and the expense of utilizing them (Dodds et al., 1991); Habit represents as the amount of activities people tend to do automatically as a result of learning, whereas Kim et al. (2007) associate habit with automaticity. Although conceived identically, Habit has been operationalized in two unique ways: first, habit is considered as past behavior, and second, habit is measured as an individual’s belief that the activity is automatic (Kim and Malhotra, 2005).

Moreover, UTAUT and UTAUT2 address the potential lack of applicability to diverse cultural and organizational contexts, a limitation that some of the previous models, such as the TAM, TRA and TPB, might encounter (Alshammari and Rosli, 2020). TAM, for example, focuses on perceived ease of use and perceived usefulness and it does not extensively consider cultural nuances or contextual variations. TRA is centered on individual attitudes and subjective norms, overlooking cultural and organizational influences. TPB extends TRA by including perceived behavioral control but still focuses on individual-level factors. UTAUT and UTAUT2 acknowledge that the impact of factors on technology adoption can vary depending on the specific context, that is the user, and the type of technology.

In addition, UTAUT and UTAUT2 recognize the dynamic nature of technology and its evolution aiming at overcoming the possibilities to become outdated. Other existing model, like TAM, since it was developed at a time when technological advancements were not as rapid as they are today, tends to be more static.

Finally, UTAUT and UTAUT2 go beyond individual factors and incorporate the collective influences that earlier models might not fully account for, emphasizing the influence of social factors on technology acceptance, and recognizing that user behavior is often shaped by social influences and expectations.

Although the UTAUT2 model was originally conceived to study the propensity for adoption in the business-to-consumer market, both UTAUT and UTAUT2 have been used in many studies concerning the healthcare industry from a business-to-business perspective (Schmitz et al., 2022). Having highlighted the advantages described by the authors of these models over their antecedents, we use the UTAUT and UTAUT2 models to deepen our understanding of ATI in the healthcare industry and to empirically verify whether they have been used as conceived or adapted or extended, due to the application industry, that is healthcare.

3. Methods

In this paper, we perform a bibliometric analysis and topic modeling to provide a comprehensive and multi-faceted perspective on the research field. The combination of bibliometric analysis and topic modeling provides a comprehensive perspective on the evolution of literature in the healthcare industry, particularly regarding the utilization of UTAUT and UTAUT2 research models (Donthu et al., 2021). While bibliometric analysis offers a quantitative and historical view of this evolution, topic modeling introduces a qualitative dimension to the analysis. Topic modeling, a statistical technique in natural language processing (NLP), extracts themes and topics from extensive textual datasets, with the widely used Latent Dirichlet Allocation (LDA) method assuming that each document in a corpus comprises a mixture of various topics, each associated with a unique set of words (Vayansky and Kumar, 2020). LDA employs a probabilistic approach to estimate the distribution of topics across documents and the distribution of words within topics, refining these estimates iteratively until convergence is achieved (Gurcan et al., 2021). This combined approach enables a holistic understanding of the trends and insights within the healthcare literature.

In detail, we performed a five-stage examination.

1. we created a data collection of documents using UTAUT and UTAUT2 in the healthcare industry and analyzed basic descriptive statistics,

2. we focused on the UTAUT and UTAUT2 models to map the history of the collection of documents to determine the evolutionary trajectory of publications in the area,

3. we investigated the fundamental works by doing a co-citation network analysis on the complete collection of 57 papers,

4. we explore the integration patterns of UTAUT and UTAUT2 models and

5. we investigated the knowledge structure of existing works on UTAUT and UTAUT2 by performing topic modeling.

For the construction of our sample, we searched in Title, Abstract and Keywords on the Scopus database, ISI Web of Science and the FT Research Rank. Specifically, we employed the following search string: (“health*”) OR (“health care”)) AND ((utaut*) OR (“utaut 2”) OR (“Unified Theory of Acceptance and Use of Technolog*”) OR (“Unified Theory of Acceptance and Use of Technolog* 2”)) AND ((adoption) OR (acceptance)) to comprehensively capture and identify relevant literature and studies within the scope of our research questions.

The search string was limited to papers published in English within a specific subset of subject categories on the Scopus database, which includes Business, Management, Accounting, Social Sciences, Economics, Econometrics and Finance. Furthermore, we focused on papers published in international academic journals that are featured in the 2021 ranking of the Academic Journal Guide (AJG) by the Chartered Association of Business Schools or listed in the FT Rank. We decided to focus on the AJG and FT Rank since they guide the range, subject matter and relative quality of journals in which business and management academics publish their research. AJG and FT Rank have high internal and external reliability; they are sensitive to small variations in the ratings of journals and are generally accepted as a fair means of ranking journals within their user community (Federkeil et al., 2012). The data were collected in February 2023.

Unlike a typical narrative literature review, our technique yields scientific and transparent conclusions, which help to reduce study bias of the researcher performing the review process. Bibliometric data were analyzed with Bibliometrix, an R program developed by Aria and Cuccurullo (2017) for thorough science mapping analysis. The Bibliometrix R package (http://www.bibliometrix.org) contains tools for doing quantitative research with bibliometrics and scientometrics.

We conducted a topic modeling analysis using the R program, employing a sequence of steps to enhance the literature review. Initially, we identified abstracts from articles in journals ranked on the AJG and FT Rank that were published in English. Subsequently, we preprocessed the collected data, which involved cleaning the text, eliminating stop words, punctuation and non-informative terms. Following the preprocessing stage, we applied a topic modeling algorithm to the refined data. The output from this analysis was then carefully reviewed to identify the most significant topics and themes. This assessment included pinpointing the key terms associated with each topic and identifying the documents with the strongest connections. Ultimately, the insights gained from the topic modeling analysis played a vital role in shaping the literature review, offering valuable information about key authors, relevant studies and pertinent issues related to each topic.

4. Results

4.1 Descriptive statistics

We conducted an analysis of 57 English-language papers spanning from 2012 to 2023. The results of our analysis indicate that the exploration of ATI in the healthcare industry has primarily centered on the application of the UTAUT and UTAUT2 models, particularly in English-speaking countries. Within Europe, France stands out as the sole country with a significant scientific production. Additionally, studies on this topic have gained substantial prominence in countries including India, China, Bangladesh and Malaysia. In our sample, Technological Forecasting & Social Change is the leading journal in terms of the number of publications (six articles) related to this topic. Following closely is Behaviour and Information Technology with three articles.

The interest within the scientific community regarding the ATI by healthcare industry has significantly increased, particularly since 2018. This interest reached its peak in 2020, coinciding with the COVID-19 pandemic, when authors (e.g. Baudier et al., 2021, 2023) dedicated their research to the realm of e-Health. During this period, in fact, healthcare workers swiftly adopted solutions such as telemedicine to maintain connections with their patients (Grinin et al., 2022), while simultaneously many IT and biotechnology companies accelerated digitization efforts to improve the well-being of both healthcare providers and patients (Brem et al., 2021). The exploration of ATI extends into various subdomains. Alam et al. (2022) and Singh (2022) have delved into the realm of mobile health (mHealth) and the adoption of innovative smart wearable technologies. Talukder et al. (2019) has expanded this exploration into the sphere of ATI within fitness and well-being, a dimension also explored by Gupta et al. (2008). In a separate avenue of study, Sabbir et al. (2021a, b) has channeled their efforts into investigating telemedicine and the online pharmacy sector.

Table 1 elucidates the evolution and significance of research models related to the ATI, with a focal point on the UTAUT and UTAUT2 models, shedding light on their origins and influential contributors. It is revealed that prior to the formulation of UTAUT, other authors, including Agarwal and Prasad (1998), delved into research models on ATI. Notably, the text emphasizes the pivotal role of authors like Venkatesh and Davis (2000) in the inception of the UTAUT model in 2003. In this vein, Venkatesh et al. (2012) conducted a systematic literature review to assess the applications of the UTAUT model, specifically in different sectors and product contexts, which laid the groundwork for the subsequent development of the UTAUT2 model. The analysis underscores a prevalent practice in the field, which is the combination of UTAUT and UTAUT2 models with statistical techniques like Structural Equation Modeling. This combined approach facilitates a deeper understanding of the complex relationships within these models, enhancing their applicability and explanatory power. Lastly, it is interesting the notable impact of the extension of the UTAUT model proposed by Hoque and Sorwar (2017), as evidenced by its high citation count in the sample.

4.2 Historiography

The historiography is built on direct citations. It draws the intellectual linkages in historical order.

The analysis presented in Figure 1 offers valuable insights into the chronological development of research on ATI within the healthcare domain. It is evident that the UTAUT model was initially applied to investigate ATI in healthcare by Khan et al. (2018), focusing on the topic of e-prescription. However, this early exploration remained relatively limited and received citations primarily from the same first author’s publication in 2022 (Khan et al., 2022). In 2019, a significant shift occurred, with three subsequent studies embracing the use of these models. Duarte and Pinho (2019), Nisha et al. (2019) and Hossain et al. (2019) studied the acceptance of remote care and mobile care. As a result, the need to understand the determinants of acceptance by healthcare providers and patients emerged. The years 2020 and 2022 emerged as pivotal periods for contributions in this field. In 2020, the research focus shifted to the acceptance of wearable health devices, as exemplified by Wang et al. (2020). Subsequently, in 2021, research on ATI, in the context of the pandemic and acceptance of wearable health and wellness devices, gained importance. In 2022, alongside these issues, scholarly attention expanded to privacy concerns associated with the disclosure of sensitive patient data on digital platforms (Pietronudo et al., 2022). In summary, the early contributions were closely linked to the potential of providing healthcare services at a distance. Over time, researchers extended their inquiries to encompass issues beyond health, placing an increased emphasis on the security of patient information, except for studies directly associated with the unique circumstances of the pandemic period.

4.3 Co-citation network

The co-citation network analysis is one of the main classic techniques in bibliometrics. It shows the structure of a specific field through the linkages between nodes (e.g. authors, papers, journals) and uses cited journals as a unit of analysis (Aria and Cuccurullo, 2017).

Figure 2 presents the co-citation network, revealing the emergence of three distinct clusters.

The green cluster mainly focuses on ATI, focusing research on how the issues of privacy and trust influence the choices related to ATI. In particular, the studies by Gefen et al. (2003), Pavlov (2003) and Hsu et al. (2013) examine the key factors addressing privacy concerns and aiming at building trust in the absence of human interactions. Moreover, Duarte and Pinho (2019), Zhao et al. (2018) and Kohnke et al. (2014) explore how safeguarding personal data plays a critical role in building trust in technology and subsequently influences the adoption of mobile health solutions. In essence, this cluster concentrates on the intricate relationship between privacy, trust and the ATI.

The blue cluster brings together studies on ATI by end-users (Ghasemzadeh et al., 2022). This cluster exhibits two distinct eras: one occurring before 2003, and the other emerging after 2003, with a notable time jump to 2010. In the period preceding 2003, the focus of research within this cluster revolves around the general concept of behavioral intention towards ATI by end-users (Compeau et al., 1999; Fishbein, 1979). In the subsequent era, which spans from 2003 to 2010, the research within this cluster took on a more methodological approach. Scholars like Hair et al. (2011, 2013), Zhou et al. (2010) engaged in methodological studies to empirically test end-users’ intentions towards ATI. These studies contributed valuable insights into the practical aspects of ATI adoption. Therefore, the blue cluster represents the evolution of research related to end-users’ attitudes and intentions towards ATI, progressing from foundational theoretical work to methodological investigations aimed at gaining practical insights into ATI by end-users (Ghasemzadeh et al., 2022).

The red cluster is characterized by the significant influence of two seminal contributions by Venkatesh et al. (2003, 2012), which gave rise to the UTAUT and its extension UTAUT2. It is noted that in the cluster, there is the co-presence of contributions that study ATI both from a business-to-business perspective (Aggelidis and Chatzoglou, 2009; Moores, 2012; Phichitchaisopa and Naenna, 2013; Williams et al., 2009) and business-to-consumer perspective (Alalwan et al., 2017, 2019; Featherman and Pavlou, 2003). This cluster exhibits a rich diversity of research models employed to study ATI adoption. Notably, it includes extensions of the UTAUT and UTAUT2 models (Alalwan et al., 2017, 2019; Alam et al., 2020; Kijsanayotin et al., 2009), as well as comparison of various research models aimed at assessing the validity and effectiveness of UTAUT and UTAUT2 (Phichitchaisopa and Naenna, 2013; Sun et al., 2013).Furthermore, the cluster encompasses the adoption of different and preceding models like the Technology Acceptance Model (TAM), Theory of Planned Behavior (TPB) and Theory of Reasoned Action (TRA) (Featherman and Pavlou, 2003; Moores, 2012; Williams et al., 2009). Lastly, some studies combine the use of UTAUT or UTAUT2 with other models to comprehensively study ATI (Aggelidis and Chatzoglou, 2009). In summary, the red cluster represents a rich and multifaceted landscape of research in the field of ATI, characterized by its in-depth exploration of ATI adoption from various angles, including different perspectives, research models and their combinations.

4.4 Exploring integration patterns of UTAUT and UTAUT2 models

Through a comprehensive analysis of published articles, which was made possible through the accurate use of bibliometric analysis and its results, we explore how UTAUT and UTAUT2 models have been adapted, extended or integrated with other constructs. The goal is to understand the evolving landscape of ATI research in healthcare industry and the different applications of UTAUT models in understanding user behavior and perceptions. Figure 3 provides a graphical representation of the results obtained.

In this study, we systematically organize relevant constructs into distinct clusters, each playing a crucial role in shaping the integration of UTAUT and UTAUT2 models. From the analysis, four distinct clusters arise among the independent variables. First, the “User Experience and Perception” cluster intricately captures the subjective dimensions of user engagement, ranging from emotional states like anxiety, assurance, enjoyment (AlQudah et al., 2022; Baudier et al., 2020b; Chong et al., 2022; Khan et al., 2022; Nisha et al., 2019) to cognitive elements like perceived credibility and perceived privacy (Nisha et al., 2019; Yousaf et al., 2021), and individual characteristics like innovativeness and self-efficacy (AlQudah et al., 2022; Baudier et al., 2020b; Mukhopadhyay et al., 2019; Sabbir et al., 2021a, b). Second, the “Quality and Service Design” cluster, emphasizes the pivotal role of system and service quality, drawing on established models like the DeLone & McLean IS Model (Nisha et al., 2019; Okumus et al., 2018; Rahi et al., 2021; Schmitz et al., 2022; Talukder et al., 2020). Third, the “Adoption and Usage Dynamics” cluster, encapsulates a broad spectrum of factors, providing a comprehensive understanding of the dynamic process of technology adoption. It includes factors related to the user’s perception of the technology’s benefits (Lo et al., 2019), social influences (Lo et al., 2019), and system-related factors (Baudier et al., 2023; Baudier et al., 2020b; Mukhopadhyay et al., 2019). Lastly, the “Concerns and Perceptions” cluster delves into various concerns and perceptions that might affect technology adoption. It includes factors related to external events (Lu and Kosim, 2022), individual responses (Baudier et al., 2021) and perceptions of the technology’s severity and responsiveness (Nisha et al., 2019; Rahi et al., 2021; Sun et al., 2013), shedding light on the multifaceted nature of user considerations.

Our results reveal two distinct groups of moderating variables in the integration of the UTAUT and UTAUT2 models. The first, labeled “Contextual Dynamics,” explores the nuanced influence of contextual factors, distinguishing between hospital and non-hospital settings and incorporating broader country-level considerations such as life expectancy and health quality (Chong et al., 2022). This cluster illuminates the pivotal role of the environment in shaping technology adoption. The second cluster, denoted as “User-System Interactions,” delves into the intricate interplay of individual and organizational dynamics. From user expertise and experience to health-related user considerations, the inclusion of constructs like perceived organizational support, perceived severity/vulnerability, privacy concerns and various facets of self-efficacy, the cluster provides a holistic understanding of user-system interactions (Beh et al., 2021; Calisto et al., 2022; Chong et al., 2022; Engin and Gürses, 2019; Mukhopadhyay et al., 2019; Nisha et al., 2019; Okumus et al., 2018; Rahi et al., 2021).

The constructs grouped under “ATI and User Perception,” serve as mediating variables in the integration of the UTAUT and UTAUT2 models. Within this group, the construct “Artificial Intelligence Guidance” recognizes the central role of guidance in shaping users’ adoption decisions in the intricate field of artificial intelligence. The integration of the “Attitude” construct provides a solid theoretical foundation, emphasizing the importance of user attitudes within the framework of the TPB (AlQudah et al., 2022). While the consideration of “Cost of Switching” sheds light on the economic implications that users face when switching to ATI (Kim et al., 2022). In addition, the treatment of “Privacy Concern,” “Risk,” “Security,” and “Trust” acknowledges the multifaceted nature of users’ perceptions, which mediate their adoption decisions (Arfi et al., 2021; Calisto et al., 2022; Choudhury et al., 2022; Trkman et al., 2023; Wei et al., 2022). Finally, the inclusion of “User Experience” emphasizes the mediating role of overall satisfaction and usability in shaping attitudes toward AI adoption (Chong et al., 2022).

Finally, the constructs classified within “Healthcare Evaluation and Engagement Metrics” serve as dependent variables in the integration/modification of the UTAUT and UTAUT2 models. In this cluster, we encounter constructs such as “Clinician Resistance,” underscoring the significant influence health professionals wield in shaping the path of health technology adoption (Kim and Malhotra, 2005); “Health Satisfaction,” emphasizing the pivotal role of user contentment in evaluating the success and overall impact of health technologies (Yousaf et al., 2021) and “Intention to Recommend,” highlighting the importance of social dynamics and network effects in adoption decision-making (Talukder et al., 2019). Additionally, incorporating “Self-reported use” and “Usage behavior” as dependent variables provides a dual perspective, merging users’ subjective insights with objective metrics, thereby offering a comprehensive assessment of technology use patterns (AlQudah et al., 2022; Trkman et al., 2023).

4.5 Topic modeling

Before delving into the analysis of the topic model, Figure 4 presents the coherence and prevalence measures. These are two important measures used in topic modeling to assess the quality of the generated topics (Kherwa and Bansal, 2021).

Coherence measures the degree of semantic similarity between the words within a topic. It is an important metric for evaluating the interpretability and usefulness of a topic. Higher coherence values indicate that the words within a topic are semantically similar and reflect a coherent theme. Prevalence, on the other hand, measures the extent to which a topic is distributed across the corpus. It is an important metric for evaluating the importance and relevance of a topic (Spooren and Degand, 2010). Higher prevalence values indicate that a topic is more widespread and relevant to the corpus (Jelodar et al., 2019; Misra et al., 2011; O’Callaghan et al., 2015).

Figure 5 provides a graphical representation of the most recurrent word combinations and allows identification of three different topics. The first topic is related to the digitalization in healthcare with a specific focus on COVID-19 pandemic. The second topic encompasses word combinations related to research models such as UTAUT, UTAUT2 and their extensions, along with their associated constructs. The third topic refers to the healthcare systems and their resistance to ATI. Accordingly, Table 2 presents the assignment probabilities for each contribution in our sample, indicating their most probable topic category.

Topic 1 (red – Digitalization in healthcare and COVID-19 pandemic) can be divided in two subgroups. The first subgroup predominantly focuses on studies related to ATI from the perspective of healthcare professionals (Weeger and Gewald, 2015). In this context, the most recurring keywords pertain to technologies that support healthcare operators, such as “assistive system” (Calisto et al., 2022; Tsao et al., 2022), “assistive technology” (Xavier Macedo de Azevedo et al., 2022), “intelligent agents” (Choudhury et al., 2022), “medical apps” (Agyei and Adzobu, 2020) and “medical records” (Badran, 2019; Mukhopadhyay et al., 2019). Conversely, the second subgroup within Topic 1 is oriented towards investigating ATI from the perspective of end-users. Here, the focus is on issues related to access to health service (Bhandari and Snowdon, 2012) with a particular focus on the COVID-19 pandemic (Baudier et al., 2021, 2023; Maleka and Matli, 2022). The studies within this subgroup examine the adoption and acceptance of telemedicine (Dwivedi et al., 2016; Rahi et al., 2021). Additionally, this second subgroup explores the perceived usefulness of the technology itself (AlQudah et al., 2022; Aria and Archer, 2018), generational perspectives (Baudier et al., 2020a, b; Jang et al., 2016; Sabbir et al., 2021a, b) and loyalty to healthcare services (Alismaili et al., 2021). In summary, Topic 1 encompasses diverse research strands within the overarching theme of healthcare digitalization and its significance during the COVID-19 pandemic, highlighting the distinct but interconnected interests of healthcare professionals and end-users in the realm of ATI.

Topic 2 (green – UTAUT Innovations: Fitness Apps, Wearables and M-Health Adoption) groups studies putting UTAUT and UTAU2 models to the test, employing advanced analytical techniques such as neural network analysis, as in the case of Kunnapapdeelert and Pitchayadejanant (2020), Sabbir et al. (2021a, b) and Yee-Loong Chong et al. (2015). Researchers extended the models by introducing additional dimensions. These extensions encompass concepts like personal innovativeness (Dhiman et al., 2019; Lo et al., 2019; Okumus et al., 2018; Sabbir et al., 2021a, b; Yee-Loong Chong et al., 2015) and perceived risk (Lu and Kosim, 2022; Schmitz et al., 2022). Finally, the UTAUT and UTAUT2 models were used in conjunction with other existing constructs, showcasing the versatility of these frameworks (Beh et al., 2021; Damberg, 2022; Dhiman et al., 2019).

Within this extensive field of research, two distinct sub-areas emerge, each focusing on specific dimensions of technology adoption. The first sub-area delves into the adoption of fitness apps and wearable technologies (Beh et al., 2021; Bianchi et al., 2022; Damberg, 2022; Dhiman et al., 2019; Sergueeva et al., 2020; Talukder et al., 2019; Yousaf et al., 2021). A second sub-area studies the adoption of mobile health services (or m-health services) (Duarte and Pinho, 2019; Kaur et al., 2023; Nisha et al., 2019; Okumus et al., 2018; Sun et al., 2013). Furthermore, within this second sub-area, there is a notable concentration on m-health services in developing countries (Alam et al., 2020, 2022; Khan et al., 2018; Mukred et al., 2017). In summary, Topic 2 provides a comprehensive view of research that revolves around the application and extension of the UTAUT and UTAUT2 models, fostering a nuanced exploration of technology adoption in the realms of fitness apps, wearable technologies and mobile health services.

Topic 3 (blue – Healthcare system and system resistance to adopt) offers a distinct socio-political perspective on ATI, with a particular focus on the responsibilities of health information systems (Pandey et al., 2021). The contributions within this topic collectively underscore the critical role of these systems in addressing privacy concerns and the perception of risk. These concerns are examined from multiple angles, encompassing the viewpoints of private citizens (Arfi et al., 2021; Ben Arfi et al., 2021; Baudier et al., 2020a; Khan et al., 2022; Singh, 2022; Talukder et al., 2020) and public organizations, such as hospitals (Engin and Gürses, 2019; Trkman et al., 2023). In the context of public organizations, with a specific focus on hospitals, the research investigates challenges related to the adoption of ATI, particularly considering the resistance to change among clinicians. Works by Hossain et al. (2019) and Kim et al. (2022) delve into the complexities of this aspect, where the protection of personal data managed by public operators takes center stage (Mukerjee et al., 2020). In this topic the perceived sense of trust and transparency emerges as central determinants in understanding the dynamics of technology adoption among both citizens and health professionals (Wei et al., 2022; Wu et al., 2021). In summary, Topic 3 sheds light on the intricate interplay of privacy concerns, risk perceptions, trust and transparency in shaping the decisions of private citizens and healthcare professionals, emphasizing the multifaceted nature of ATI within the healthcare industry.

The three topics, examined in this analysis, converge to offer a holistic and nuanced perspective on ATI in the healthcare industry shedding on the intricate dynamics involving various stakeholders, including healthcare professionals, end-users and healthcare organizations. Through these lenses, we gain a deeper understanding of the diverse interests and perspectives that underpin the adoption of healthcare technologies. Healthcare professionals, driven by the pursuit of enhanced efficiency and patient care, seek innovative solutions that support their work, while end-users, particularly in the context of the COVID-19 pandemic, navigate the adoption of technologies to access and benefit from healthcare services. Moreover, the topics underscore the complex interplay of factors influencing technology adoption, spanning the technological, sociopolitical and individual realms. This comprehensive view recognizes that the adoption of healthcare technologies is not solely a technological endeavor, but a multifaceted process deeply entwined with sociopolitical considerations and the unique characteristics of individuals and organizations.

5. Discussion and conclusions

In this paper, we use bibliometric methods and topic modeling to explore how the literature evolved concerning technological acceptance in the healthcare industry and to identify the emerging innovation trends and opportunities in the healthcare industry. The exploration of ATI in healthcare industry, with a focus on the UTAUT and UTAUT2 models, has developed especially in recent years driven in part by the COVID-19 pandemic, and shows a solid and persistent interest in understanding user behavior and perceptions.

The analysis of the co-citation network, particularly within the red cluster, provides valuable insights into the prevailing usage patterns of the UTAUT and UTAUT2 models within the literature on technological acceptance in the healthcare industry. The undeniable influence of seminal contributions by Venkatesh et al. (2003, 2012) in this cluster marks the inception of the UTAUT model and its subsequent extension, UTAUT2. Surprisingly, only one contribution, namely Kunnapapdeelert and Pitchayadejanant (2020), directly adopts the UTAUT model without extensions. This rarity raises intriguing questions about the perceived limitations or evolving needs that might drive researchers to extend these models in various ways. The prevalent trend across the sampled articles is the extension of UTAUT and UTAUT2 with additional dimensions, moderators and mediators, as visually depicted in the provided Figure 3. This widespread practice underscores the versatility and adaptability of these frameworks to the intricate contexts of healthcare technology adoption. While offering a more nuanced exploration of factors, the extensive use of extensions prompts reflection on potential concerns, including the risk of model proliferation and challenges in synthesizing findings. This observation calls for a critical discussion within the academic community about the balance between customization and standardization, aiming for a more robust and comparable body of knowledge in the evolving landscape of healthcare technology acceptance.

The analysis through topic modelling revealed several emerging trends in the adoption of healthcare technologies. The first topic emphasizes the diversity of perspectives in the digitization of healthcare. The dual focus on healthcare professionals and end-users highlights the need for a comprehensive approach that takes into account the diverse needs and concerns of these stakeholders. This trend suggests an emerging emphasis on inclusivity, recognizing that successful digitization efforts must consider the perspectives of those providing and receiving healthcare services. The second topic demonstrates the extensive use of UTAUT and UTAUT2 models in studying the adoption of fitness apps, wearables and mobile health services. The application of advanced analytical techniques and the introduction of additional dimensions reflect a trend towards a more nuanced exploration of technology adoption. The third topic highlights the socio-political dimension of healthcare technology adoption, emphasizing the responsibilities of health information systems. The focus is on privacy concerns, perceived risks and resistance to change within the healthcare system. Across all topics, the recurring emphasis on trust and transparency emerges as a central trend. Whether examining the dynamics between healthcare professionals and technology or the concerns of private citizens, the importance of building trust in technology adoption processes is evident. This trend signals a recognition that successful healthcare technology adoption is not solely a technological challenge but also a socio-political and ethical imperative, requiring transparent communication and ethical considerations.

5.1 Managerial implications

The identified trends in this paper offer managerial insights for stakeholders in the healthcare industry. Managers overseeing the implementation of healthcare technologies should recognize the existence of diverse perspectives, advocating for customized strategies that address the distinct needs of both healthcare professionals and end-users. The adaptability of UTAUT models underscores the necessity for technological solutions that can traverse varied contexts, prompting managers to encourage the development of technologies adaptable to diverse socio-economic and cultural settings. Trust and transparency emerge as central themes, highlighting the need for informed decision-making by prioritizing user trust through transparent communication about technology features and risks. For managers dealing with healthcare information systems, addressing privacy concerns and understanding resistance to change becomes crucial. Strategies must align technology implementation with privacy regulations and guidelines while mitigating resistance among healthcare professionals. Lastly, managers, especially in developing countries, are urged to adopt a global perspective, considering the unique challenges and opportunities in resource-constrained settings. This approach facilitates the creation of adaptable and scalable solutions on a global scale, ensuring successful healthcare technology adoption in an ever-evolving landscape.

5.2 Limitations and further research

This study has some limitations that should be acknowledged. Firstly, it is pertinent to note that the study would benefit from longitudinal research encompassing multiple countries. This would enable a comprehensive understanding of trends in the acceleration or deceleration of the ATI process within the healthcare industry. While the COVID-19 pandemic has been identified as a catalyst for ATI, this study does not provide conclusive evidence that it alone was sufficient for systematic ATI. Thus, future research could delve into the sustained impact of events like the pandemic on ATI processes over time. Second, Moreover, the study’s sample selection focused exclusively on scientific journals included in the Academic Journal Guide (AJG) 2021 ranking by the Chartered Association of Business Schools (CABS) or in the FT Rank. Although this approach offers a well-defined and reputable subset, it introduces potential issues. Future research should consider broadening the sample selection to encompass a wider range of sources, including grey literature, conference proceedings and reports. Additionally, the study’s temporal scope, despite its flexibility, reveals a concentration of scientific contributions emerging post-2015 and gaining significant traction around 2020. Future research could explore earlier or future periods to trace the evolution of ATI-related studies. Finally, addressing the objection raised about the sample’s narrow focus on AJG and FT50 journals, it’s crucial to reiterate that innovation in the healthcare industry, especially concerning ATI, demands careful planning, interdisciplinary collaboration and diverse skill sets. However, future research may benefit from expanding the scope to include a broader spectrum of journals while ensuring a rigorous selection process to maintain the quality and relevance of the data.

Figures

Figure 1

Historiograph citation network

Figure 2

Co-citation network plot

Figure 3

Integration patterns of UTAUT and UTAUT2 models

Figure 4

Coherence and prevalence measures

Figure 5

The topics identified

Table 1

Top 10 most cited references

Cited references	Local citations
Davis, F.D., 1989. Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Q. 13, 319	12
Hoque, R., Sorwar, G., 2017. Understanding factors influencing the adoption of mHealth by the elderly: An extension of the UTAUT model. Int. J. Med. Inform. 101, 75–84	12
Venkatesh, Morris, Davis, Davis, 2003. User Acceptance of Information Technology: Toward a Unified View. MIS Q. 27, 425	12
Venkatesh, V., Davis, F.D., 2000. A Theoretical Extension of the Technology Acceptance Model: Four Longitudinal Field Studies. Manage. Sci. 46, 186–204	11
Ajzen, I., 1991. The theory of planned behavior. Organ. Behav. Hum. Decis. Process. 50, 179–211	10
Fornell, C., Larcker, D.F., 1981. Structural Equation Models with Unobservable Variables and Measurement Error: Algebra and Statistics. J. Mark. Res. 18, 382–388	9
Davis, F.D., Bagozzi, R.P., Warshaw, P.R., 1992. Extrinsic and Intrinsic Motivation to Use Computers in the Workplace1. J. Appl. Soc. Psychol. 22, 1111–1132	7
Henseler, J., Ringle, C.M., Sarstedt, M., 2015. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J. Acad. Mark. Sci. 43, 115–135	7
Venkatesh, Thong, Xu, 2012. Consumer Acceptance and Use of Information Technology: Extending the Unified Theory of Acceptance and Use of Technology. MIS Q. 36, 157	7
Agarwal, R., Prasad, J., 1998. A Conceptual and Operational Definition of Personal Innovativeness in the Domain of Information Technology. Inf. Syst. Res. 9, 204–215	6

Table 2

Probabilities of assignment to the most likely topic for all documents

ID	Authors and year	Topic 1	Topic 2	Topic 3
1	Kaur et al. (2023)	0.058	0.492	0.45
2	Calisto et al. (2022)	0.804	0.016	0.18
3	Khan et al. (2022)	0.001	0.37	0.629
4	Choudhury et al. (2022)	0.519	0.333	0.147
5	Damberg (2022)	0.056	0.943	0.001
6	Chong et al. (2022)	0.178	0.603	0.219
7	Trkman et al. (2023)	0.001	0.026	0.973
8	AlQudah et al. (2022)	0.683	0.001	0.316
9	Singh (2022)	0.001	0.001	0.998
10	Xavier Macedo de Azevedo et al. (2022)	0.977	0.023	0.000
11	Lu and Kosim (2022)	0.012	0.986	0.001
12	Kim et al. (2022)	0.083	0.011	0.905
13	Maleka and Matli (2022)	0.945	0.034	0.021
14	Baudier et al. (2023)	0.771	0.228	0.002
15	Bianchi et al. (2022)	0.01	0.989	0.001
16	Alismaili et al. (2021)	0.988	0.001	0.011
17	Wei et al. (2022)	0.001	0.057	0.942
18	Tsao et al. (2022)	0.997	0.002	0.002
19	Alam et al. (2022)	0.012	0.976	0.012
20	Yousaf et al. (2021)	0.015	0.983	0.001
21	Arfi et al. (2021)	0.002	0.002	0.997
22	Pandey et al. (2021)	0	0	0.999
23	Baudier et al. (2021)	0.682	0.216	0.102
24	Arfi et al. (2021)	0.013	0.037	0.95
25	Sabbir et al. (2021a, b)	0.871	0.027	0.102
26	Beh et al. (2021)	0.001	0.998	0.001
27	Baudier et al. (2020a, b)	0.001	0.042	0.957
28	Dhiman et al. (2019)	0.001	0.906	0.093
29	Mukhopadhyay et al. (2019)	0.96	0.001	0.039
30	Baudier et al. (2020a)	0.83	0.012	0.158
31	Sergueeva et al. (2020)	0.002	0.869	0.129
32	Alam et al. (2020)	0.014	0.699	0.288
33	Sabbir et al. (2021a)	0.001	0.998	0.001
34	Kunnapapdeelert and Pitchayadejanant (2020)	0.032	0.946	0.022
35	Agyei and Adzobu (2020)	0.832	0.155	0.013
36	Mukerjee et al. (2020)	0.012	0.218	0.77
37	Talukder et al. (2020)	0.001	0.433	0.566
38	Engin and Gürses (2019)	0.012	0.179	0.809
39	Duarte and Pinho (2019)	0.016	0.968	0.016
40	Lo et al. (2019)	0.002	0.628	0.371
41	Badran (2019)	0.79	0.131	0.079
42	Talukder et al. (2019)	0.001	0.624	0.375
43	Hossain et al. (2019)	0.069	0.061	0.87
44	Nisha et al. (2019)	0.068	0.931	0.001
45	Khan et al. (2018)	0.022	0.707	0.271
46	Okumus et al. (2018)	0.017	0.966	0.017
47	Mukred et al. (2017)	0.001	0.998	0.001
48	Jang et al. (2016)	0.985	0.014	0.001
49	Dwivedi et al. (2016)	0.883	0.001	0.116
50	Yee-Loong Chong et al. (2015)	0.001	0.998	0.001
51	Sun et al. (2013)	0.027	0.933	0.04
52	Bhandari and Snowdon (2012)	0.985	0.001	0.014
53	Schmitz et al. (2022)	0.032	0.844	0.124
54	Rahi et al. (2021)	0.817	0.141	0.042
55	Wu et al. (2021)	0.012	0.045	0.942
56	Weeger and Gewald (2015)	0.843	0.001	0.156
57	Aria and Archer (2018)	0.989	0.001	0.01

Note(s): Topic 1: Digitalization in healthcare and the COVID-19 pandemic

Topic 2: Research models

Topic 3: Healthcare system and system resistance to adopt