Influential factors in the adoption of clinical decision support systems in hospital settings: a systematic review and meta-synthesis of qualitative studies Free

There has been a rapid uptake of disrupting technologies over the past 3 decades, in particular, digital technologies pushed by the fourth industrial revolution (Maroufkhani et al., 2022). Hence, healthcare organisations have embarked on digital transformation in order to stay competitive with limited monetary and non-monetary resources (Ahmadi et al., 2018; Sutton et al., 2020). One of the most sophisticated advancements in healthcare IT enabling clinicians to drive meaningful use of electronic health records (EHRs) is clinical decision support systems (CDSS) that aim to prevent medical errors and improve patient safety and quality of care (Bates et al., 1998; Kaushal et al., 2003; Moja et al., 2014; Sutton et al., 2020).

CDSSs are computerised information systems integrating patient and clinical information in order to improve evidenced-based decision-making. These real-time systems carry out algorithms using computational reasoning aggregating multifaceted patient information through a knowledge base in order to generate patient-specific recommendations (Berner, 2016). Such systems assist clinicians in conducting clinical practice through simple (passive, non-specific information presented at the point of decision-making) and complex interactions (interactive communication by presenting patient-specific advice along with the clinical guidelines) (Fraccaro et al., 2015; Miller et al., 2005).

Berner (2016) (p.2) claims that “If used properly, CDSS have the potential to change the way medicine has been taught and practiced”. As such, it improves processes and patient outcomes (Bright et al., 2012) and hence promotes its adoption in the healthcare industry. Throughout the last decade, there has been a boom in adopting CDSS technologies across the globe, especially in developed countries forming a multibillion-dollar market (Chang and Gupta, 2015; Sutton et al., 2020). Further, governments in developed countries have been actively advocating and mandating CDSS initiatives, acknowledging CDSS as the focal innovation to provide evidence-based services and hence accounting for the growing CDSS market (Sutton et al., 2020).

Such an extensive investment calls for continues evaluation of the impact of CDSS systems. However, according to a plethora of studies in the field of CDSS, the outcomes do not justify the high cost of establishment. While a mild to moderate impact on hospital outcomes such as length of stay and mortality rate has been revealed by a number of studies, the majority of studies are showing no positive trend regarding hospital outcomes (Bright et al., 2012; Jaspers et al., 2011; Jia et al., 2016; Klarenbeek et al., 2020; Moja et al., 2014; Prgomet et al., 2017; Roshanov et al., 2013; Shahmoradi et al., 2021). To shed light on these controversial outcomes, researchers advocated for qualitative evaluation of CDSS interventions to dissect the discrepancies in healthcare settings, CDSS design and external factors that may be leading to different outcomes across healthcare settings (Ammenwerth et al., 2008; Georgiou et al., 2013; Greenes et al., 2018; Kawamoto et al., 2005; McKibbon et al., 2012; Moxey et al., 2010).

The results of the said studies have demonstrated a dramatic scepticism and rejection of CDSS by clinicians which can be attributed to organisational and environmental factors such as organisational infrastructure or government regulations that can be traced back to procurement phases of CDSS adoption (Ahmadi et al., 2018; Black et al., 2011; Kaplan, 2001; Moxey et al., 2010; Pu et al., 2018). According to the literature, these gaps and shortfalls then lead to low acceptance and usage of CDSS technology by clinicians in the go-live phase (Stilwell et al., 2022). This is in alignment with the results of the technology diffusion science claiming that adoption of a technology is a complex socio-technical phenomenon that goes beyond the relative advantage and compatibility studies (Wang et al., 2022).

The establishment and onboarding of a CDSS technology is a massive undertaking that encompasses a sequence of phases including adoption, pre-implementation, implementation, maintenance and promotion. Adoption is the first phase of CDSS onboarding that includes awareness, selection and development/contracting that takes place before the implementation of CDSS technology (Ahmadi et al., 2018). Adoption means that the system or innovation is new to the adopting organisation and is anticipated to cause changes in the organisation that will lead to better outcomes. During this process, managers recognise the need and initiate a search for possible solutions that are evaluated for their suitability and prospective benefits to the organisation (Damanpour and Schneider, 2006).

Damanpour and Schneider (2006) recognised that within the “adoption” phase, different antecedents effect the managerial decision in adopting or rejecting the innovative technology and adoption procedures. Therefore, identification of the antecedents and incorporation of rectifying strategies can empower leadership for a successful adoption that will have a positive knock-on effect on the following phases (pre-implementation, implementation, etc.). Similarly, literature has traced many of the setbacks occurring in the implementation and maintenance phase to initial decisions and initiatives made by the leadership of the organisation as to adopting CDSS (Ahmadi et al., 2015, 2018; White et al., 2023). This lends support to the calls for studies aggregating influential factors in the selection and contracting phases of CDSS adoption (Baysari et al., 2023; Liberati et al., 2017; Sutton et al., 2020; Venkatesh et al., 2020).

In order to discover the antecedents of the CDSS adoption, the Technology–Organization–Environment (TOE) theoretical framework was applied in this study (Tornatzky et al., 1990). The aim here was to go beyond the sole exploration of existing literature and pinpoint the gaps for future research (Seuring et al., 2020). In doing so, we aimed to evaluate the theoretical position of “adoption” literature that empowers the direction for future research (Weick, 1995). The use of the TOE theoretical framework enables a better understanding of the contextual differences related to the field of healthcare.

TOE is an organisation-level theory that uses a flexible approach and utilises a holistic lens to harness the influential factors in the adoption of an innovation. It has been strongly supported in the literature for its ability to explain the formation of “adoption” at an organisational level (Nguyen et al., 2022). TOE aspires to explain adoption behaviour in relation to three groups of antecedents, namely, technical context, organisational context and environmental context (Tornatzky et al., 1990). Technical context comprises all the technologies related to the organisation either currently owned by an organisation or prospective technologies relevant to the scope of services provided by the organisation. Organisational context includes characteristics and resources of the organisation such as culture, financial resources, physical infrastructure and so on. Finally, environmental context refers to the technology providers, dynamics of the industry and competitors as well as the regulatory bodies (Baker, 2012).

Although there are other theories that investigate the adoption of a new technology by an organisation (i.e. Diffusion of Innovation Theory), the capacity of TOE theory in investigating and the distinct foci that it has on environmental factors made it most applicable to the current study (Callen et al., 2008). Previous studies investigating the adoption of new technologies in healthcare settings provide further support for the suitability of TOE theory to this context (Abekah-Nkrumah et al., 2022; Ahmadi et al., 2015).

In order to gain an overview and accumulation of knowledge in innovation adoption literature (Duygan et al., 2023), we conducted a systematic review and meta-synthesis. In doing so, we accumulated the existing literature by synthesising relevant knowledge on the “adoption” phase to elucidate determinants specific to this phase. Accordingly, we investigated the literature incorporating both suppliers’ and buyers’ point-of-view regarding influential factors in the procurement and adoption process of CDSS. In addition to employing a multiphase lens for the synthesis, we applied a multidimensional approach, using the TOE theoretical framework, that empowered us to go “beyond the provision of a static list of obstacles and facilitators” (Liberati et al., 2017) and comprehend the dynamics of adoption antecedents. The aim of this SLR is to synthesise and examine the existing high- and medium-quality literature at the organisational level on CDSS adoption. Thus, not only the current knowledge of CDSS adoption is illustrated and the gaps are revealed but also it yields a granular understanding for developing strategies that will enable a successful adoption of CDSS technologies.

This study conducts a systematic review and meta-synthesis of the literature on CDSS procurement. According to Wang and Chung (2022, p. 444), a “Systematic literature review is a well-established and rigorous method of evaluating and reviewing research literature based on reproducible, scientific and transparent process, which is a key tool for building an evidence base and reducing bias”. In addition to serving the complementary purposes of secondary study analysis, meta-synthesis complements quantitative SRs by investigating research questions from a different angle, bringing diversity to the development of evidence-based practice and policy (Major and Savin-Baden, 2010). The other advantage of qualitative study synthesis is identifying similarities and differences that exist across various contexts and enabling managers to convey sustainability in implementation science (Glenton et al., 2013).

Our study protocol is registered on PROSPERO: the international prospective register of systematic reviews (CRD42020196418). This study was conducted in accordance with the PRISMA guidelines for reporting systematic reviews of studies that evaluate healthcare interventions. The guideline refers to a set of recommendations for conducting and reporting systematic reviews. PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses. The aim of PRISMA is to ensure transparency, accuracy and completeness of systematic review reports (Page et al., 2021). Following the 27-item checklist of PRISMA, authors are required to include important information ranging from selection of studies, data extraction, quality appraisal, etc. that makes it possible to ascertain the validity and applicability of the findings by readers.

Healthcare services are classified into three main categories including primary, secondary and tertiary healthcare settings. While primary care services provide preventive and curative services, the more complex cases that may need further investigation or possible interventions that require admission to the hospital are referred to secondary and tertiary healthcare settings (Almalki et al., 2011). In addition, the nature of CDSS being used in outpatient and primary settings are significantly different leading to different results and challenges (Ronan et al., 2022). In this study, papers were included if they were conducted in secondary or tertiary inpatient healthcare settings. The reason for excluding other healthcare settings is the unique complexity of systems and workflows in hospitals that make the adoption challenges exclusive to that particular context (Ahmadi et al., 2017).

The phenomena of interest to this meta-synthesis were the antecedents and determinant factors in the CDSS adoption phase from key stakeholders’ (vendors, healthcare managers and frontline users) point-of-view. With respect to Berner’s definition of CDSS, we reviewed CDSS interventions in the event of delivering information or recommendations at the point of decision-making with the aim of conveying specific guidelines into clinicians’ decision-making process (Berner, 2016). The communication must have happened through computerised modules through digital interfaces (workstation on wheels, tablets, mobile, etc.). All the qualitative studies reporting influential factors in the adoption phase of CDSS were reviewed.

Studies conducted in primary healthcare settings were excluded from this review. The exclusion of primary healthcare settings was made due to substantial differences in scope and impact on clinical workflows, magnitude of organisational challenges and external advocacies for adoption between hospital and primary care settings.

Further, decision automation interventions (machine learning, AI, expert systems, etc.), sole passive platforms (i.e. web-based tools), dashboard view or other personal digital assistant systems, patient decision support/aid and surveillance/early warning systems were excluded from the review. The AI interventions were excluded as to their radical difference in generating decision supports using AI algorithms. As per other systems, either the purpose is not providing decision support to clinicians or targeted populations are not clinicians.

Only interventions that were implemented in real, none-simulated, clinical settings were included in this meta-synthesis. Studies without an available full text and usability studies were also excluded. High-quality studies were selected for inclusion only, this is described further in the quality assessment section below.

The search strategy was conducted by one of the authors. Initially, a preliminary search in Cochrane, EMBASE, Medline, PubMed, Web of Science and PROSPERO was done to track published and in-progress systematic reviews. An extensive search was undertaken covering the period January 2010 and July 2020 in Medline, EMBASE, PsycINFO, Web of science and Scopus. The search strategy was conducted and tailored sensitively to each database. All the key terms and combinations are provided in  Appendix 1. Only studies published in the English language journals were included. The reference lists of all relevant studies and related systematic reviews were hand-searched. Grey literature search was also carried out using the Google search engine in order to identify additional relevant articles. The search results were subsequently updated until June 2022 in all the above-mentioned databases.

The study selection process was conducted via Covidence web software (Covidence, 2021) and all the steps of screening were conducted independently by two of the authors of this paper. The selection of relevant studies was done via screening the titles and abstracts against the inclusion and exclusion criteria after a hand-search and automatic removal (via Covidence) of duplicates. Selected documents were then full text-reviewed against the criteria. The disagreements were resolved by either a discussion between the two reviewers or a consensus involving a third member of the project team for a final decision. Figure 1 describes the screening and selection process in this study following the PRISMA guidelines. The extraction involved a line-by-line reading of the selected papers to extract the characteristics and to recognise the patterns which later framed our codes.

We used an expanded version of JBI’s appraisal tool ( Appendix 2) to investigate the qualitative methodology utilised in the chosen papers (Miller et al., 2015). This qualitative assessment framework provides a comprehensive and transparent set of questions (n = 20) that comprises all the critical quality dimensions in qualitative studies including credibility, transferability, dependability and confirmability (Leung, 2015; Saldana, 2014). Using this tool, the selected studies were scored between 0 and 40 (each question representing one score). Categorising the scores into four groups, namely: high, medium, low and very low. With respect to the aim of the study, including above-average quality studies for meta-synthesis, we excluded the first and second quartiles (0<score ≤ 20) as well as studies that failed to meet above 30% of the score in the abovementioned quality dimensions. This cut-off resulted in 10 final studies for further analysis.

Qualitative studies were synthesised in accordance with PRISMA guidelines for systematic reviews and the meta-synthesis of qualitative studies (Page et al., 2021). We used the framework synthesis approach to organise and analyse the challenges surrounding the adoption of CDSS. This method of synthesis has been widely recognised for its ability to rigorously analyse qualitative results in a highly structured manner using a priori framework (i.e. theory) as a foundation. During this largely deductive process, although mainly being classified into constructs of the priori framework, new constructs may be developed or constructs be expanded (Barnett-Page and Thomas, 2009). The data synthesis was conducted independently by two of the authors. The main findings, field observations and interviewees’ quotations were independently extracted through line-by-line reading of the papers in the form of codes. These codes were later reread and discussed by the authors through several meetings and were associated with the main three themes of the TOE theory, namely, Technology Context, Organisation Context and Environment Context. Disagreements were resolved by rereading the relevant sections to curb misinterpretation. The qualitative data that was allocated to more than one theme was consensually either broken down into meaningful constructs or resolved to a single theme. In the final step, the categories were presented to the rest of the research team for assessment of comprehensibility and congruence.

The screening and eligibility check, using the PRISMA diagram, is presented in Figure 1. Our initial electronic search identified 9,706 publications, with 6,035 excluded at the initial title and abstract screening stage due to not meeting the inclusion criteria and 263 excluded due to duplication. The remaining 117 (0.7%) papers proceeded for full-text screening and quality appraisal, resulting in seven (7.1%) studies included in the review. Additionally, grey literature and citation searching resulted in an additional six papers that were assessed against the eligibility and quality criteria of which three (50%) were chosen to be added to the final pool, resulting in 10 papers in total. Cohen’s k for agreement between the independent reviewers was 0.4. Throughout the screening process, articles that did not meet one or more inclusion criteria were eliminated. A summary of the included articles is presented in Table 1.

Four dimensions of qualitative research quality (transferability, dependability, credibility and confirmability) were considered for quality appraisal (Leung, 2015). Using the JBI tool, papers were divided into four quartiles and studies located within the third and fourth quartiles (score $≥$ 20) were included in the meta-synthesis. In the total sample of 27 articles passing through the full screening stage, 14 articles were excluded (51.8%) and 13 articles were included during the quality appraisal process. The quality score statistics were a mean of 24, a standard deviation of 9.9, median of 25 and range of 2–36 points. Six articles of medium quality (third quartile) with a mean score of 23 were excluded due to failing to meet less than 20% in any of the four quality dimensions. The authors believed that an 80% shortage in any dimension could be a significant sign of bias in the said research, thus it was considered appropriate to eliminate these studies. In total, the percentages of high, medium, low and very low quartiles were 41%, 30%, 11% and 19%, respectively.

Table 1 summarizes the studies included in the synthetic analysis. All the studies were conducted in the context of developing countries with the majority conducted in the UK (Bailey et al., 2020; Cresswell et al., 2015, 2017; Lee et al., 2016; Mozaffar et al., 2016a, b) and the USA (Ash et al., 2012, 2015; Joshi et al., 2022; Simon et al., 2013) as the frontrunners in healthcare digitisation leaving the developing countries context under-investigated. All but three studies investigated the adoption of commercial CDSS (Ash et al., 2012; Bailey et al., 2020; Joshi et al., 2022) in hospitals, leaving the home-grown context under-investigated. With respect to study participants, only three studies (Ash et al., 2015; Lee et al., 2016; Mozaffar et al., 2016a) took internal and external (users, managers, IT staff and vendor experts) stakeholders’ perceptions into perspective and only one study investigated vendor experts’ perception with regard to CDSS adoption (Cresswell et al., 2015). Methodologically, only one study used ethnography (Bailey et al., 2020) and three used combinations of an inductive and deductive approach using a theory-based deductive analytical framework (Cresswell et al., 2017; Mozaffar et al., 2016b; O’Connor et al., 2022). The remaining studies used inductive thematic or grounded theory analysis. Three studies collected data using three or more different sources of data including observations, interviews, filed data and documents (Ash et al., 2012, 2015; Cresswell et al., 2017; Mozaffar et al., 2016b). Three studies conducted interviews and observations (Bailey et al., 2020; Lee et al., 2016; Simon et al., 2013). One study used interviews and complementary focus groups to collect the data (Mozaffar et al., 2016a). There was a lack of consistency with regard to studies’ foci so only six studies solely focused on procurement and adoption as the purpose of the study (Baysari et al., 2023; Cresswell et al., 2015, 2017; Joshi et al., 2022; Lee et al., 2016; Liberati et al., 2017) while the rest mainly focused on development, implementation and utilisation of CDSS.

Table 2 illustrates the synthesis analysis with respect to each study. The synthesis incorporated the main dimensions of TOE theory as the main them. The information of each included paper in the study was inductively analysed and categorised under the main three dimensions including technology context, organisation context and environment context. Bailey et al. (2020) illustrated the concept of relative advantage in a comparative study of two British hospitals using CDSS. The authors demonstrated that no system can be perfect and one size does not fit all. While one hospital developed its own algorithm, which eradicated underdiagnosis, it faced overdiagnosis and a higher number of alerts were triggered, meaning that the system was not only more expensive to develop but also costlier to run and maintain. Despite the increased costs, the system was more effective in identifying prospective cases (e.g. potential acute kidney injuries). On the other hand, the second hospital did not aim for eradication but rather mitigation of underdiagnosis. As such, managers dealt with a trade-off between cost and effectiveness, trying to minimise the workload and alert-fatigue (Bailey et al., 2020).

Therefore, developing a framework guiding the procurement process and contracting through organisational readiness assessment was recommended. Using this framework, internal capacities are assessed and communicated to vendors before any contracting takes place so that they are required to declare the compatibility of their solution to organisation demands (Ash et al., 2012; Liberati et al., 2017). This is also emphasised in the following quote from the study conducted by Cresswell et al. (2015):

Additionally, the literature highlights that drawing a line between CDSS functionality myths on the managers’ “wish list” (digital hospitals), clinicians’ demands (real-time and remote access to patient information) and the real capacity of CDSS is necessary. When it comes to CDSS technology, which in part could be driven by a lack of believe in an evidence-based approach, adopters just presume that CDSS is safe and effective and the entire design and development of CDSS are informed by existing evidence. One of the participants in the study conducted by the Baysari et al. (2023) stated:

The cost of establishing CDSS was noted as an on-going challenge (Joshi et al., 2022). It includes the procurement of the hardware, and the software, as well as the necessary infrastructure and initial and ongoing training for users. Although this expenditure can be substantial and deterrent for some hospitals, the long-term expenditures that incorporate customisation, configuration, maintenance, training, monitoring and evaluation can be catastrophic as well. The long-term expenditures more often are overlooked by managers. In order to decrease the costs of the system at the procurement stage, some hospitals exclude the longitudinal support and maintenance services provided by vendors (Ash et al., 2012; Cresswell et al., 2015).

At the adoption phase, managers and clinicians are concerned with the integrity of CDSS and the impact that it will have on the workflow of end-users (O’Connor et al., 2022). Selection of a suitable CDSS is a challenging task. There is a wide range of CDSS available that managers can choose from, based on the type of logic applicable (Rule-based or Machine learning), the level of integration possible with legacy systems and whether the system is supplied by a vendor or is internally developed. Joshi et al. (2022) showed that the majority of hospitals eventually turned to a vendor-supplied CDSS due to ease of integration and customisability. A CDSS that generates generic alerts, is not integrated with the day-to-day workflow of clinicians or is not based on the most current knowledge will exacerbate the scepticism and resistance, particularly during the early stages that involve a radical change for the hospital (Ash et al., 2015). The following quote particularly illustrates the frustration of the physicians:

Managers who were vigilant of not causing alert fatigue leaned towards involvement of the human in the reliability detection of the alerts by retrospectively investigating the indication of the alerts before prompting clinicians. Although this approach succeeded in decreasing the alert fatigue, it is not real-time and requires significant ongoing investment. On the other hand, automated alerts are more integrated, and real-time as they eliminate processing the information by human, but the chances are that they increase overdiagnosis and increase of the alert fatigue. This shows the extent of variance in types of CDSS that can be chosen, each carrying its own significant burden, and involving trade-offs between efficacy and sustainability (Bailey et al., 2020).

Technical characteristics of CDSS should be compatible with the organisational finance and IT aspirations as well as with legacy systems and be interoperable with other non-integrated solutions. An important part of the adoption process is measuring the technical compatibility of the prospective CDSS solutions (Cresswell et al., 2015). The extent of integration and dependency of prospective CDSS is also critical to be measured. Non-integrated CDSS has a higher tendency to relay on other intermediary systems to convert and transmit the data to CDSS. This often led to long delays at the implementation or go-live phases due to logistics and the associated costs (Mozaffar et al., 2016a). This was demonstrated by Mozaffar et al. (2016a):

Considerable differences in workflows and standards of care between different countries were a part of the reason for the low adoptability of “foreign” CDSS. The complexity of the nuance of care in the variety of destination countries makes it convoluted and laborious for vendors to customise CDSS. This difference can lie in many aspects of the healthcare settings such as standards and accreditation, insurance and reimbursement systems and the relationship between primary and secondary care. For example, the discharge of patients who are prescribed take-home medications is significantly different between the USA and the UK While in the UK medication at discharge needs to be fulfilled at the hospital pharmacy, in the USA, they can be picked up at any pharmacy (Mozaffar et al., 2016b). This brings about a substantial change in the discharge workflow that should be translated to digital workflows within the CDSS which is not an easy feat. This significance was illustrated in the following:

The benefit of CDSS adoption is deemed to highly depend on the IT knowledge of managers and their expectations of the systems that were mostly overestimated and biased. Thus, decision-makers failed to consider their organisations’ capacity (i.e. skill shortages, IT infrastructure, etc.) for adopting CDSS which led suppliers to poorly plan unrealistic projects (Lee et al., 2016; Mozaffar et al., 2016a). This shortage is expressed by Cresswell et al. (2015) as follows:

The wrong impression of CDSS was noted as common among hospital top-managers. CEOs’ knowledge regarding CDSS’s types, discrepancies in the purpose they serve and critical factors to consider for selecting the CDSS in terms of compatibility with future IT aspirations were declared as inadequate (Cresswell et al., 2015; Liberati et al., 2017; Mozaffar et al., 2016a, b). The lack of evidence-based approach in the selection of CDSS by managers and leaders regarding the actual impact of CDSS on clinical performance and relying on vendors’ assurance was reported in the literature. It came as a surprise to clinicians that controlling mechanisms for medical and drug interventions are generally available but for CDSS technologies these rarely exist (Baysari et al., 2023). According to Ash et al. (2015), CEOs are heavily motivated to increase their revenue by adopting CDSS and their expectation of what a CDSS can do is surreal. Also, a lack of experience with activities involved in the adoption (such as business change management, benchmarking, product testing and contracting) hindered the efficient adoption of CDSS (Cresswell et al., 2015).

Literature suggests a frequent mutual communication path between the adopter and the supplier as a remedy to bridge the knowledge gap. Striking such a relationship will set a mutual communication path between hospitals and vendors before the tendering phase and assist hospitals in the selection process. The existing knowledge gap in terms of technical aspects and system requirements for the successful implementation of CDSS solutions and the unrealistic demands of adopters can be identified by the suppliers and prevented at an early stage (Baysari et al., 2023; Cresswell et al., 2015; Lee et al., 2016; Mozaffar et al., 2016b), however, often neglected by adopters (Mozaffar et al., 2016b). This was highlighted in the following statement:

Another approach that was endorsed as a reliable source of information regarding the alignment and quality of services provided by suppliers was benchmarking. Contacting peer organisations which are ahead in the digitisation journey can provide valuable information for future adopters (Cresswell et al., 2015; Lee et al., 2016). This is emphasised in the following quote:

Understanding the current organisations’ situation in terms of IT maturity and IT goals assists managers to come up with a stepwise approach to achieve their targets slowly but steadily. By targeting fewer complex functionalities of CDSS, such as order sets (predefined collections of clinical orders) and documentation templates, as the “low-hanging fruit,” managers can break down the challenges to smaller and readily resolvable issues (Ash et al., 2012). However, some functions are fundamental or “must haves” and pave the way for reaching aspirational functionalities (Mozaffar et al., 2016a). The advantage of this would be enabling managers to learn all the pros and cons of CDSS and in the future being more capable of handling complex challenges that might arise from the full development of CDSS functionalities across hospitals. Additionally, as the changes are not extensive, managers are able to consider changing their approach towards the type (i.e. modular towards integrated) and source (i.e. home-grown towards commercial) of CDSS which they will adopt (Mozaffar et al., 2016a). This is captured in the following quote:

Further, this approach can prevent future resistance by users as the system was not initiated as a whole and the possibility of alterations in approach and type of CDSS are considered. Thus, users’ opinions, which often have been neglected at the adoption phase, should be taken into consideration for the next phases of the IT journey.

In order to design a stepwise plan, a situational analysis (including evaluating inter-disciplinary relationships, alignment of organisational norms and policies with CDSS technology) will give managers a good grasp of the organisations’ readiness against future IT aspirations (Liberati et al., 2017; Mozaffar et al., 2016a).

The literature showed that there is a tendency towards rushed adoption of CDSS while the majority of prerequisites are not in place (Ash et al., 2015; Cresswell et al., 2015, 2017; Lee et al., 2016; Mozaffar et al., 2016b). Strong financial incentives driven by politicians’ ambitions through digital transition initiatives were perceived as coerced and a wrong motivator by managers (Mozaffar et al., 2016a; Simon et al., 2013). This is further illustrated in the following quote:

Another factor motivating hospitals to adopt CDSS technology was the progression of the healthcare community at large towards digitalisation (Baysari et al., 2023; Joshi et al., 2022). This exponential transition was part of the reason that hospital managers took the evidence on the effectiveness of CDSS for granted and embarked on this journey. It was deemed that CDSS adoption is becoming an inevitable part of clinical care and there is no point in further evaluation of this technology. This notion is demonstrated as follows:

Conditioning the central capital funding to adopting a CDSS potentially exposes hospitals to rushing the adoption followed by opportunistic behaviours (Cresswell et al., 2017; Lee et al., 2016). For instance, managers are deemed to search for popular and complex CDSS rather than CDSS which focuses on quality of care (Ash et al., 2015). A vendor representative articulated this wrong motivation as follows:

This wrong impression of urgency was particularly found to be a problem in an emerging CDSS industry. For example, in the UK, the rushed adoption of CDSS in an inexperienced market led suppliers to grow their commercial aspirations while their lack of capacity was evident (Ash et al., 2012). As such, Cresswell et al. (2015) noticed hard-selling tactics or ‘cherry picking’ of more promising and digitally mature hospitals. This was also noted by Ash et al. (2012) as follows:

The literature identified several signs of an immature market, including: (1) lack of knowledge and heterogeneity of users, leading to difficulties for vendors modifying their products, (2) diversity of offerings and opportunistic behaviours on the vendor side and (3) absence of integrated and transparent policies and control mechanisms on the government side (Mozaffar et al., 2016a). This is of significant concern because there should be a close compatibility between CDSS, and hospitals’ needs and capacities which seems too far from ideal at this point.

Trust in vendors and their recommendations is a significant driver of the selection process (Baysari et al., 2023; Joshi et al., 2022). The misalignment between the vendors’ and hospitals’ priorities and goals led to faulty contracts. Lee et al. (2016) referred to this as “incomplete contract” situations, leaving hospitals vulnerable to vendors’ opportunistic behaviours such as charging hospitals for unprecedented maintenance that should have been embedded in the contract. This is highlighted by Lee et al. as follows:

Usage of output-based specifications (clarifying expectations from CDSS) can serve managers as a tool to refrain from technical jargon and complexities to pinpoint actual needs. However, the wide range of CDSS solutions available requires unique IT structures and preparation, hence making the right decision can be challenging for managers. Following this imbalance in the distribution of knowledge between the provider of the solution and the buyer, the proposal put forward by the hospital authorities are not thorough and practical to be acted upon by vendors. This is highlighted by Cresswell et al. (2015) in the following statement:

Requiring adaptations to organisational workflows is a usual demand by hospitals at the adoption phase. It was evident from the literature that vendors’ marketing strategies considered different extents of pro-activeness and enthusiasm incorporating user demands. This has caused an ongoing dispute between vendors and hospitals that revolves around what hospitals refer to as “configuration” and what vendors consider as overly specific “customization”. From the hospital's point-of-view, system configurability is a necessity; however, vendors, particularly those with a lower market share, were reluctant to embed the required changes in the generic CDSS design (Mozaffar et al., 2016b). On the other hand, from the vendors’ point-of-view, this in part is due to the fact that vendors’ past experiences show that the lack of knowledge and uncertainties on the customer side causes gradual demand for alterations in the system as their CDSS knowledge expands along the way (Lee et al., 2016).

Admitting the importance of government presence in the CDSS market, studies concluded that governments must help in addressing the power balance in the market through centralisation of procurement and tendering processes. In the context of emerging markets, hospitals can benefit from a flexible framework navigating managers for selection, assessment, negotiation and contracting purposes. The presence of government in the CDSS market can ensure a thriving and competitive market (Cresswell et al., 2015). Along with government support, the existence of well-defined HIT standards can not only guide and protect hospitals during the adoption and implementation of CDSS but also guarantee robust and shared health information exchange at a wide scale (Ash et al., 2012). This is emphasised in the following quote:

In this respect, governments are able to allocate funds, not only to the adoption stage but also to the results and outcomes emerging from auditing and accreditation of standards (Bailey et al., 2020; Cresswell et al., 2017; Joshi et al., 2022). There were also few studies that noted benchmarking and being motivated by inferior hospital outcomes compared to other hospitals which established CDSS solutions (Bailey et al., 2020; Joshi et al., 2022). However, central governance can lead to frustration and loss of opportunities for a successful rollout by enforcing a wrong choice and a one-size-fits-all approach adopted by governments. Consequently, the incongruency and inadaptability of CDSS workflows with hospital routines were reported as a shortfall amongst a wide range of commercial products (Lee et al., 2016; Mozaffar et al., 2016b), as highlighted in the following:

To the best of authors’ knowledge, this is the first systematic review that investigates the technical, organisational and environmental factors that influence the adoption of CDSS in hospital settings. We have investigated the possible remedies revealed by the literature to strengthen and unify CDSS adoption guidelines while also considering the local context. This study reaffirms the capacity of the TOE theoretical framework to explain the influential factors regarding firms’ adoption behaviour and provides an in-depth understanding of the phenomenon.

Foremost, we found the absolute scarcity of quality studies focusing on the adoption process of CDSS technology. Those studies that focus on the adoption stage do so only peripherally. Literature has disproportionately focused on the other phases of CDSS establishments such as “implementation” and “go-live” phases (Cassidy et al., 2019; Liberati et al., 2017; Mølgaard et al., 2022; Pontefract et al., 2018; Vandenberg et al., 2016; Yilmaz and Ozdemir, 2017) rather than the “adoption” and “procurement” phases and on CDSS users rather than managers (Ackermann et al., 2022; Kawamoto et al., 2005; Khajouei and Jaspers, 2010; Pierce et al., 2022; Zare et al., 2022; Zha et al., 2022; Zhai et al., 2022). The vast majority of studies investigated commercial CDSS as opposed to home-grown CDSS which can be, in part, due to the popularity of commercial CDSS in the healthcare market and lack of appropriate skill sets to design and develop home-grown CDSS.

The result of this review aligns with previous literature highlighting that generic workflow models used in the off-the-shelf CDSS solutions, predominantly from the US market, due to their low compatibility and adaptability to the destination countries’ context has turned into a concern amongst hospital managers (Abdekhoda et al., 2019; Cresswell et al., 2017). In the case of the UK, around one-half of its hospitals use CDSS products developed overseas that have caused issues at the implementation phase (i.e. process misalignment between CDSS and destination context) (Aljarboa and Miah, 2019; Mozaffar et al., 2014).

The compatibility of CDSS is also driven by the internal IT infrastructure as they define the scope and pace of technological advancements that organisations can uptake (Abekah-Nkrumah et al., 2022; Baker, 2012). This will be a foundation for the technology innovation trajectory in accordance with the extensiveness of existing technologies in the market. Managers should take note that depending on the hospital’s infrastructure, the type of innovation varies (Abdekhoda et al., 2019; Ahmadi et al., 2015; Steinhauser et al., 2020). Hospitals with an existing EMR might find the adoption of CDSS rather an incremental innovation because they already onboarded the transition from paper to electronic format. In contrast, less technologically advanced hospitals with no EMR (the majority of hospitals in developing countries) will face a discontinuous innovation adoption. In doing so, not only an EMR platform must be established but also there is a need for a radical infrastructural and practice change in transitioning to CDSS.

Our study showed that the cost of adopting CDSS was also a major influencing factor which is in agreement with previous studies (Aljarboa and Miah, 2019; Sutton et al., 2020). In addition to the initial cost associated with system procurement, set-up and integration there will be ongoing expenses related to system adjustments, hiring and training of staff and software upgrades.

Top management plays a remarkable role in the adoption of IT interventions in organisations. IT knowledge and experience of managers are critical for the successful design, development and adoption of CDSS technologies (Abdekhoda et al., 2019; Abekah-Nkrumah et al., 2022; Cresswell et al., 2017; Maroufkhani et al., 2022; Miller et al., 2015). These results are similar to those reported by Ingebrigtsen et al. (2014), confirming that healthcare leaders with IT skills and experience make better selection of IT solutions for their organisations and stronger long-term commitment to implement them. Bialas et al. (2023) similarly posited that a knowledgeable and motivated workforce and supportive managers are so crucial that it can even offset the limited financial resources impact on technology adoption (Bialas et al., 2023).

Another measure that plays a significant role in the adoption process is gauging the organisation’s capacity to embed CDSS technologies (Abdekhoda et al., 2019). Several factors such as degrees of integration, interoperability, time, budget and adaptability with hospital practices must be taken into consideration before contracting (Mozaffar et al., 2014; Wang et al., 2022). To be able to choose from the wide range of IT solutions in the market, hospital leaders are to construct a digital strategic vision which aligns with a professional, managerial and administrative consensus of all stakeholders (Ash et al., 2012; Bates et al., 2003; Klecun et al., 2019).

The result of this review confirms communication as a key to the successful adoption of IT solutions (Pu et al., 2020). The knowledge imbalance that leads to inappropriate choices on users’ side and opportunistic behaviours on suppliers’ side stems from lack of effective and sustainable communication paths between vendors and hospitals (Cresswell and Aziz, 2013). Thus, clear communication of the strengths and weaknesses of the hospital and technology is a must before embarking on a contract (Ash et al., 2012; Cresswell et al., 2020; Maroufkhani et al., 2022; Mozaffar et al., 2017).

Literature suggests that new technology may not always be the remedy for existing problems and be ill-fitted with organisations’ infrastructure (Cresswell and Aziz, 2013). Institutional theory, similar to TOE, recognises this internal initiative that is stimulated by external forces and refers to this phenomenon as isomorphism. Isomorphism is the act of onboarding a new technology without questioning its legitimacy and claims as mimetic pressure driven from uncertainty. Thus, organisations face uncertainty towards a new technology, hence, imitating successful organisations’ actions (Abdekhoda et al., 2019; Klecun et al., 2019). Mimetic isomorphism compels vendors as well. Our study showed that proactive vendors undertook measures to address the dynamic demands of the market by designing configurable software packages which contain a basic set of organisational functionalities besides built-in templates. These templates cater for the most encountered workflows in that context, though they contain limited scenarios and flexibility to be tailored to the functions which were not pre-programmed (Mozaffar et al., 2017). Maroufkhani et al. (2022) in the study of big data analytics adoption amongst small and medium size firms showed that the high rate of adoption attracts vendors urging them to increase their products' compatibility with adopters' demands and reprioritised their marketing strategies.

Coercive isomorphism in IT adoption triggered by the external environment was also demonstrated by previous CDSS research (Klecun et al., 2019). For example, the literature highlights that the coercive pressure exerted by governments that interfered with the firms existing digitalisation strategy which affects the predisposition of organizations towards the adoption of a system (Ahmadi et al., 2015; Currie, 2012). Further, Ahmadi et al. (2018) confirmed the role of accreditation and standardisation in healthcare environments as a driver of Malaysian hospitals adopt HIT (Ahmadi et al., 2018).

Likewise, literature recognises governments’ tendency towards rushed adoption of CDSS solutions (Cresswell and Aziz, 2013, 2020; Mozaffar et al., 2017; Sutton et al., 2020). Mozaffar et al. (2014) similar to our findings, traced this rushed adoption back to governments’ IT aspiration to showcase digital success and neglecting the time needed to yield positive outcomes. That is part of the reason of an increase in the adoption of CDSS by hospitals across the world despite the low uptake by clinicians and scarcity of studies showing positive patient outcomes (Ronan et al., 2022).

According to the results of this study, consideration of all the key factors for a transition through pre-implementation phase can be significantly challenging and requires years of planning and dedication of resources. These challenges can be more severe in developing countries context due to inadequate infrastructure, cost of implementation and maintenance of CDSS, lack of financial motivation for adopters, lack of IT literacy and technical personnel, inadequate Internet bandwidth, poor electricity supply, Internet connectivity deficit and lack of governments’ capacity to motivate and empower organisations (Ahmadi et al., 2015; Damanpour and Schneider, 2006; Essuman et al., 2020; Lin and Kofi Kujabi, 2022). This can be even more heightened in the case of public hospitals being micro-managed by governments (Ahmadi et al., 2015; Bialas et al., 2023).

This study confirms the capacity of the TOE theoretical framework in explaining the adoption behaviours at a firm level and provides an in-depth understanding of the phenomenon. The flexibility that the TOE framework brings to each research context enables researchers to grasp and fit contextual differences while being attuned to the three main constructs of the theory (Baker, 2012). However, there are factors that were recognised in this research that can strengthen the comprehensiveness of this TOE framework for utilisation in the context of CDSS and healthcare. While management support was noted in this literature review as well as in previous TOE literature, the IT knowledge and previous experience with the adoption of IT solutions were declared as a crucial factor in effective adoption decision-making. This is even more so in the event of home-grown CDSS that requires far more knowledge, IT skills and support.

The other surprising and contradictory outcome of this study is the lack of importance in technological aspects of the intervention. Further, the capacity and complexity of the system are the most, if not only, aspects that are considered by hospital management teams and the reason is its direct association with current and future expenses of system adoption and implementation. The evidence-based approach adopted by a number of studies included in this review was noted as an important organisational trait crucial in choosing the best CDSS for the organisation. Therefore, system observability (requiring tangible outcomes by managers) was not mentioned in the literature which pertains to a lack of knowledge among adopters and choosing the system for the wrong reasons (i.e. receiving incentive packages from the government) which entirely contradicts other industries as they focus on the tangible outcomes of IT solutions (Nguyen et al., 2022).

Our study demonstrates that the dynamics of the relationships between a hospital and government, between a hospital and a vendor and between a vendor and government are really complex. The context of healthcare that inherently embeds a strong government presence in the market intervening HIT transition (subsidising, incentivising, etc.) changes the nature of influential factors that affect the CDSS adoption decisions. For example, our study showed that the regulatory and financial incentives announced by governments were a dominant influential factor for CDSS adoption leading to suboptimal adoption. Although necessary to regulate and support, the extent of centralisation and authority of governments related to CDSS adoption must go under further scrutiny. Another positive driver of CDSS adoption was the progression of digitalisation in healthcare which made CDSS an inevitable change from the adopters’ point-of-view.

This review has lessons for managers and politicians. To be able to choose from the wide range of IT solutions in the market, hospital leaders should take a strategic view to develop a deep understanding of digitalisation that aligns with professional, managerial and administrative consensus of all stakeholders (Ash et al., 2012; Bates et al., 2003; Klecun et al., 2019). Klecun et al. (2019) suggest managers seek modular solutions in comparison to ambitious and comprehensive solutions as their success is bound to sustained stakeholders’ commitment and managers can envisage their IT journey through a stepwise approach. This along with the situational analysis will assist hospitals to opt for the best approach to achieve their IT aspirations (Cresswell and Aziz, 2013; Mozaffar et al., 2017). It is vital for top managers to comprehend that the results of the situational analysis might establish inadequate capacity for adopting a new technology, therefore, recommending maintenance of the status quo (Pu et al., 2020).

There is a vast range of products and vendors available to choose from which makes the selection and procurement process intensely convoluted. There are not only technical and strategic factors that must be considered by decision-makers but also the future national IT vision is also a factor that can affect the success of adopted CDSS technologies. The rapid cycle of change in the CDSS market, entering numerous architecturally conflicting products into the market and the lack of clear guidelines for adopting hospitals attests to the role of government policy in guiding, or misguiding, the CDSS market (Mozaffar et al., 2017).

Due to the imbalance of knowledge distribution and high chance of incomplete contracting between hospitals and CDSS vendors, authorities must stay vigilant of the pros and cons of different CDSS solutions and oversee the market through soft controlling mechanisms (i.e. accreditation standards), development of strong pressures through imposing international standards (Health Level Seven) and development of suitable procurement guidelines for locally chosen IT pathways (Ahmadi et al., 2015; Bell et al., 2019; Cresswell and Aziz, 2013; Mozaffar et al., 2014)

This meta-synthesis is prone to some limitations. Firstly, the majority of the studies captured in this review were conducted on commercial CDSS while the challenges that might occur in the homegrown CDSS context and the way to address them can be widely varied and exclusive. Secondly, the results of this review are limited to developed countries context with a lengthy background in CDSS technology which, to some extent, excludes the implications to dissimilar contexts in which IT maturity and budget can impose challenges. Thirdly, although all the processes were mitigated through double reviewing by multiple authors, the selection of studies may be subject to selection bias. Additionally, compliance with PRISMA guidelines limited such bias. Finally, studies in English were included in this review thus, future reviews may consider publications in other languages as well.

Future work can bridge the abovementioned gaps through, first, abiding by qualitative research guidelines to increase the rigor and trustworthiness of the results. Considering the significant contextual differences between developed and developing countries (Mozaffar et al., 2014, 2017), there is an urgent need for qualitative studies investigating the establishment of CDSS technologies with a distinct focus on each phase of establishment (i.e. adoption, pre-implementation, etc.). A deeper contextual understanding of CDSS establishment would provide tailored recommendations which will warrant a more efficient investment in the target context.

1. Exp Decision Making, Computer-Assisted/

2. Exp Decision Support Systems, Clinical/

3. Decision Support Techniques/

4. cdss.ti,ab

5. CCDSS.ti,ab.

6. Letter/

7. Editorial/

8. News/

9. Exp historical article/

10. Anecdotes as Topic/

11. Comment/

12. Case report/

13. (Letter or comment*).ti.

14. OR/6–13

15. Animals/not humans/

16. Animals, Laboratory/

17. Exp animal experiment/

18. Exp animal model/

19. Exp Rodentia/

20. (Rat or rats or mouse or mice).ti.

21. OR/15–20

22. 14 or 21

23. 1 or 2 or 3 or 4 or 5

24. Exp Hospitals/

25. Exp Secondary Care Centers/

26. Exp Secondary Care/

27. Exp Tertiary Healthcare/

28. (Process or success or factor or mediat* or moderat* or failure or challenge* or lesson* or effect* or capabilit* or acquisition or implement* or maintenance or establish* or change or barrier or facilitat* or enabler or hurdle).ti,ab.

29. 24 or 25 or 26 or 27

30. 23 and 28 and 29

31. 30 not 22

32. Limit 31 to yr = “2010 -Current”

Source(s): Author’s own work

Table A1