Understanding variation in departmental adoption of Electronic Health Records: an embedded case study

Background: Electronic Health Records (EHR) are integrated software applications used by healthcare providers to create, share, retrieve, and archive patients’ health status information. Especially for large healthcare organizations, implementing Electronic Health Records organization-wide is a complex endeavour. The EHR literature generally suggests that contextual factors play a major role in adoption. We demonstrate how the work context influences adoption at the departmental level in a situation where each department has its own medical specialty or patient stream, clinical authority, and accountability. Here, the achievement of full adoption by all departments is not self-evident. Drawing on EHR implementation in a Dutch hospital, this study explores how the clinical departments’ work context characteristics contribute to their pre-implementation intended adoption and their post-implementation EHR uptake. Methods: This embedded case study allowed us to examine the EHR adoption of eight diverse clinical departments in terms of their work and socio-political context. Data collection entailed semi-structured interviews, observing meetings, document analysis, and feedback sessions to check our interpretations. We examined the context and adoption intentions before implementation and the adoption level approximately half a year after the Go Live. The comparative case analysis iterated between holistic department-level descriptions and structured data displays based on inductive and deductive coding. Results: We identified three departmental types that varied both in adoption intention and post-implementation uptake: (1) departments oriented towards the organization with an enthusiastic or compliant adoption; (2) internal-oriented

departments with a selective or conditional adoption; and (3) externally oriented departments with no or low adoption with workarounds.
Conclusions: We conclude that work context characteristics contribute to individual departments' adoption of an EHR. By acknowledging departmental types that will vary in adoption intention, and especially the underlying explanatory mechanisms, we recommend that implementers acknowledge these departmental types and differentiate their strategies towards clinical departments accordingly based on their observable work context characteristics. Based on these findings, we develop propositions that contribute to the development of a department-level EHR adoption theory.
background Healthcare is in an era of digital transformation in the quest to deliver better quality of care at lower costs. For hospitals, having a well-functioning Electronic Health Record (EHR) is key to this transformation. The main promises of these systems are integrated applications, a uniform data model, and shared patient data that result in improved patient service, quality, and healthcare safety [1] and contribute to cost effectiveness due to efficient patient data flows [2]. The EHR systems are repositories in which digitalized healthcare process information is organized in a way that enables clinical departments to register, process, retrieve, archive and, importantly, share patients' data on both the individual patient level within the daily workflow as well as on the aggregated level (e.g. for quality monitoring, teaching, and research purposes). As such, EHRs were also meant to facilitate clinicians to base their decision-making and research on an allencompassing view of patient data. In practice, the promises are not fully met due to reported low adoption or fragmented use caused by low interoperability and unintended adverse consequences such as information overload, incessant box checking, laborious documentation, and designs based on overly rationalist, linear models [3][4][5][6][7][8]. This paper explores the contextual factors underlying differences in realized adoption between clinical departments in a hospital.

The adoption of Electronic Health Records
For hospitals in particular, the successful adoption of an EHR depends on the compliance of all its clinical departments. In the embedded case study reported in the current paper, one surgeon stated: "we [the department] are very much in favor of this new integrated health record, we are looking forward to it", whereas a specialist from another department argued, "we don't see many advantages for our department, rather we foresee problems". However, only through widespread adoption by all departments, an EHR can reap its benefits and fulfil the high performance expectations in terms of empowering health professionals to efficiently and transparently provide better continuity and higher quality care [9]. As clinical departments are heterogeneous and powerful units, each with its own goals, medical specialty developments, and interests, the achievement of any hospitalwide adoption is a complex endeavor [10][11][12][13]. In this embedded case study, we explore how clinical departments' diverse contexts may contribute to differences in their adoption of an hospital-wide Electronic Health Record. Thus far, the departmental or clinical subunit level has barely been taken into account in EHR adoption studies. Such studies have mostly focused on either the individual [e.g. 14] or the organizational level [e.g . 15].
Still, some results suggest that hospital-wide EHRs suit some clinical departments better than they do others [e.g . 4]. A meta-review [10] concluded that small EHRs may be more efficient and effective than larger ones. A reason for this finding could be that the large diversity in clinical processes across specialties and their specific information requirements prevent standardization, and that the extent of customization required renders a large EHR inefficient. A qualitative study also showed how paper-and EHR-based practices within one network organization had different priorities for implementing a new EHR [16]. On their turn, Park et al. [17] compared the adoption of EHR by Canadian ophthalmologists with physicians from other surgical specialties. Compared with other surgeons, the ophthalmologists felt that EHRs were not suitable for their practice, and too time consuming. This was attributed to unique features of their field, including a heavy reliance on handdrawn figures in documentation and high patient volumes. Park et al. consequently called for more research to explain differences in EHR adoption and identify implementation strategies to mitigate discipline and clinical department-related barriers. This call is in line with an earlier review of the impact of emergency department features on EHR adoption, which led to a call for examining diversity in needs and expectations across departments [18].

Research aim
We took a focus on the clinical department level for three interlocking reasons.
First, differences in views on adoption among clinical departments are critical for EHR implementers and technicians who require their cooperation in aligning the system and departments' clinical processes [17]. Clinical departments have their own distinct specialist responsibilities, and legal and budgetary accountabilities, and this grants their boards a level of discretion when it comes to adopting an EHR [19]. Second, the staff of a clinical department share working goals and practices with which the information system will have to be integrated if it is to be effectively used [20]. For nationwide EHR systems in primary care, whether the EHR enables or constrains micro-level care delivery depends on the realized alignment of goals and resources [21]. The same may be true for hospitals, making the department a highly relevant level of analysis [22,23]. Third, evidence suggests that healthcare organizations that are divided into semi-autonomous departments, with decentralized decision-making structures, assimilate innovations more readily [24].
However, given that EHRs are aimed at organization-wide integration of data, applications and work processes across departments, it is possible that contextual differences among these semi-autonomous clinical departments complicate this readiness. The literature, however, offers little insight into the nature of the relevant clinical departments' contextual characteristics. Such insights would be highly valuable for implementers negotiating EHR adoption and adaptation.
The stake in these negotiations is high since adoption outcomes such as non-use, partial use with workarounds [4,25,26], and fragmented or neglectful use, even if by only one clinical department, could jeopardize effective use by other departments and for the hospital as a whole. In such situations, the EHR will fail to deliver the promised data overview and seamless integration of operations [25].
The paper is organized as follows. First, we describe the methods employed in our embedded case study of eight departments within a hospital and present the results. These departments are compared in terms of both work and socio-political context, pre-implementation adoption intention, and realized adoption. Based on within-case analyses, we first develop a categorized overview of the contextual characteristics that were perceived by department board members and leading representatives to affect the departmental intention to adopt the EHR. A postimplementation analysis demonstrates that the departments' realized adoption was in line with their earlier adoption intentions. The cross-case analysis shows three departmental adoption types that are directly related to differences in their work context and less so to their socio-political context. Based on these outcomes, we develop a model and propositions for further research. Implementers could use the knowledge of a department's work context at an early stage of the implementation process to guide the required collaboration with the department members.

Research site and department selection
To explore how the particular contexts of clinical departments affect the adoption of organization-wide EHR, we conducted an embedded case study within a large hospital in the Netherlands. A case-study approach allowed us to study departmental adoption in its natural setting and so understand its nature and the complexities involved [27, p. 78]. The Dutch healthcare system is in transition from a supply-side oriented, government-regulated approach towards managed competition [28]. The Netherlands has 77 hospitals [29]. Since 2006, hospitals have increasingly been allowed to negotiate with health insurers over treatment prices and quantities. The number of negotiable treatments has risen from 20% in 2008 to 70% in 2018.
Increasing pressure from insurance companies to lower prices has resulted in competition between hospitals [30]. In order to lower prices while maintaining quality, hospitals are increasingly specializing, which requires negotiations between hospitals and specialties. Partly as a consequence of this specialization, hospitals are still expected to collaborate with each other in networks to maintain accessibility and increase the quality of care [31]. This development is enabled through demands for transparency in terms of price and quality imposed by government, insurers, and patients, a process that represents a strong push towards EHR use. In many hospitals in the Netherlands, clinical departments enjoy a fair degree of autonomy if only because the medical specialty is professionally accountable for the patient care delivered. Often, medical specialties are organized in legally separate entities that negotiate with the hospital management. The focal hospital of our research had initiated a program directed at purchasing and then implementing an organization-wide EHR to improve cross-departmental workflows, patient safety, and efficiency of care. Top management had decided to implement, through a big-bang approach, a uniform system, with hospital-wide functionality, that left little room for customization. The focus was on internal hospital processes and reimbursement for services rather than on information exchange with external actors. The system was expected to replace numerous stand-alone applications and integrate all the existing paper-based and electronic records.
The case study in this paper covers the pre-implementation stage that lasted 2.5 years (Autumn 2012 -Spring 2015) and the subsequent uptake between December 2017 and Spring 2018, complemented with a snapshot of the realized adoption for each department a year later. The pre-implementation stage was divided into two phases, in each of which we gathered data. Phase 1 focused on project preparation, requirements specification, the tendering procedure, and vendor selection. Phase 2 involved the subsequent system design, system integration, training, and testing. The implementation strategy from December 2017 onwards was plateau-oriented and is still in progress. The level of customization will be increased at each successive plateau. The analysis of the realized adoption concerned the implementation of the first plateau.
Given our study's comparative nature [32], eight departments were selected to allow theoretical replication [33]. The departments (coded A -H) were selected in consultation with the EHR's project management to ensure variation on three selection criteria: (1) the department's existing patient administration and handling (largely digitalized versus largely paper-based); (2) department staff level (small:< 75fte, medium:75-150fte, large >150fte); and (3) the grouping logic underlying the medical specialty (see Table 1). given a preliminary score on a number of characteristics to support our sampling.
During the pre-implementation period, after having gained formal approval from the department boards, we conducted 44 core interviews: 36 within the sampled departments and 8 with project managers. The authors were involved in the interview protocol design and in conducting most of the interviews, the latter either individually or as a dyad, often assisted by students. Interviews were taped and transcribed. During Phase 1 of the pre-implementation, in 2013, 24 interviews were conducted with department board members and leading user representatives (i.e. physicians, nurses, and administrators) from each department [34]. At the end of  To contextualize the data and ensure construct validity [35,36], we held a meeting every six weeks during the pre-implementation stage with between two and four project management representatives. Given our aim to examine the department level, we also held four feedback sessions with interviewees and some of their colleagues to share and discuss interpretations. We ensured that two of the authors were present at all these meetings and minutes were taken. The meetings and additional data sources, including newsletters, written reports, and policy plans, enabled us to better contextualize and verify the relevant departmental characteristics and how these characteristics impacted adoption [32].

Data reading, coding, and analysis
Since each department was different in terms of the clustering of its main context characteristics, they were each treated as a separate embedded case [37]. The following six steps were conducted and the results from each step were discussed in the research team as a whole.
First, for each department, we developed a holistic storyline, based on our reading of the available materials, to enable an initial understanding of their voiced adoption intentions. In discussing these initial storylines, and tentatively mapping emerging differences between them, we noted that, in all the clinical departments surveyed, the views of the board members and leading representatives on EHR adoption were built around their department's specific work context. The specifics of the work context were prominent in interviewees' reasoning about future (non-)adoption of the EHR.
Second, following this preliminary finding, we provisionally split up all the relevant quotes in the data as relating to either work context or socio-political context. To more in-depth unravel the dominant work context characteristics, we returned to the literature and selected a framework that distinguishes three elements in the work context: Technology, Organization, and Environment (TOE) [38]. Two authors then individually coded the pre-implementation data using deductive TOE-based and complementary inductive codes. These codings were compared and combined, systematizing and refining the context characteristics' codes as required.
Third, for each department, the same two authors discussed all the quotes that directly referred to departmental adoption intentions, leading to a short description of each department varying between firmly declared intentions to use and stated strong likelihood of non-use. On a few occasions, we had to go back to the transcripts for clarification and to minutes and notes of meetings to verify the convergence in the interviewees' views. We concentrated on the expressed views regarding the likelihood that the department would show compliance in using the system on a regular basis after the 'organization-wide' go-live. In our conceptualization and coding, we kept the departmental intention to use well apart from the more general departmental support for or resistance towards the implementation process [39].
Fourth, we went back to the transcripts, and earlier generated data displays, to deepen our understanding of the different reasons behind the departmental adoption intentions and again further refined the code definitions for the contextual characteristics (see data structure in additional file 1 and codebook in additional file 2). These iterations led to a department-based list of voiced reasons behind the adoption intention. Each voiced reason could be grouped into either work context or socio-political context characteristics.
Fifth, we subsequently cross-analyzed these lists to arrive at a tree of perceived contextual characteristics and we carefully identified the direction of each characteristic's influence based on the interviewees' accounts (see below in Table   3, 2 nd column; see also the data structure in Additional file 1). The elaborated tree with context characteristics, whose preparation again included additional data checks, enabled us to distinguish, for each department, between dominant characteristics and complementary characteristics that further strengthened or weakened the department's adoption intention. After comparing these eight departmental patterns, we arrived at three department types in terms of adoption intentions and the contextual characteristics underlying these intentions.
Finally, we examined what happened in the period between 0.5 to 1.5 years after the EHR go-live, and compared the realized adoption after approximately one year with the departments' earlier voiced adoption intentions.
results  in Department G, the EHR was not adopted. In this section, we present the crosscase analysis leading to these distinct types.

Context characteristics and related adoption
The context characteristics that interviewees used to explain their department's adoption intention are provided in the second column of Regarding the socio-political context, the interviewees' narratives did not reveal any dominant role of these characteristics: as Table 3 indicates, our findings provided no indications of a direct relationship between these characteristics and a departments' adoption intentions.
Pattern identification: three department types By examining the interplay between the contextual characteristics and then relating the identified patterns to the department's adoption intention, we were able to determine the dominant factor(s) for each department (shown in Table 3 by the department letters in bold). For example, Table 3 shows that the material technical capabilities of Departments C and E, critical resources for delivering their specialty's patient care, had the most impact on their adoption intention. Although these departments were supportive of the change direction, they had developed conditional, or limited, adoption intentions (Table 4; see also In contrast, the low technical capabilities of Departments B and G had a positive influence on their adoption intentions. Department B's interviewees mentioned that greater integration between applications and a long-desired connection with a specific technology were expected to be realized through the EHR implementation.
However, their serious concern was that this positive influence would be outweighed In the end, these departments had nevertheless to implement Plateau 1. In Department C, a collective workaround was determined to secure their own throughput efficiency until a customized module was provided. In Department E, many of the analytical functionalities of their highly customized patient data management system were lost. This was, however, partly compensated for by a medical specialist who managed to acquire high EHR expertise and so build customized flowcharts for their clinical processes and specific order sets, and also generate the required data analyses "deep within" the EHR itself. Finally, halfway through 2018, both departments were provided with a 'physician builder' -a physician responsible for maintaining and updating their own EHR configuration.
Representing the third type, with an external environment orientation, were Departments B and G that had significant extra-organizational dependencies through regional or (inter-)national collaborations, or through laws and insurance regulations. These departments were highly supportive of implementing an EHR, but said they could only adopt the new system to the extent that it would not hinder them in fulfilling their external requirements. Their fears in this respect showed in pre-implementation reservations. The post implementation data show that it was negotiated that Department G, facing externally imposed production registration requirements that went against the EHR logic, did not adopt the system. A customized module was developed for Department B to enable it to meet its external demands. Moreover, a third system that the physicians worked with, alongside the two applications replaced by the EHR, would be integrated in the EHR for Department B. The latter customization was, however, ultimately postponed to Plateau 2 by the supplier. Consequently, both a medical specialist and the project manager acknowledged that the outcome was perceived as complicated to work with, leading to individual physicians using workarounds.
In the hospital studied, the EHR was initiated in a top-down manner, with the implication that management expected the system to be fully adopted. However, our data suggest that certain work context characteristics decrease the likelihood that a department will adopt the system as anticipated. In particular, some departments have specialized, local technologies that the EHR cannot be integrated or connected with, others have an advanced and vulnerable workflow design that would need to be adapted, and others have binding requirements imposed by their extra-organizational environment. Other influential work and socio-political contextual characteristics were also identified, but played a more modest role in the view of the interviewees.
Our finding that departments' workflow and resource dependencies influence EHR adoption accords with classic sociological research that recognizes these factors as critical contingencies for work design at the department level [40,41]. The patterns we reveal add to a systematic eHealth review [42], which shows that a crucial fail factor is not understanding the workflow(s) that the eHealth will support and the Departments with a hospital orientation (Type I) experience their work processes as having reciprocal dependencies with other departments that require intensive and reliable two-way information flows (compare [40]). Since EHR targets the hospitalwide integration of data and work processes, Type I departments expect an EHR to be especially useful [compare 43], leading to intended and realized adoption of the hospital-wide EHR. In our study, they showed themselves also relatively resilient to negative socio-political contextual issues. Finally, their adoption intention was strengthened when the department negatively evaluated the fit between their workflow design and the existing IT systems that supported them before the implementation. These arguments lead to the first two propositions for further research. In contrast, a department orientation (Type II) reflects a department that operates relatively independently, the orientation is inward, and management may in this respect even feel or portray it is 'isolated'. In terms of Thompson [40] the dominant work dependencies are sequential. Given that these departments tend to focus on optimizing their internal processes, they may have developed or acquired localized, highly customized, and advanced applications, hardware, and other digital equipment. The Type II departments in our study had also developed their own IT vision and knowledge. We saw that, when such departments are satisfied with their tailored IT systems (realized or under development), their adoption intention is lower since the improved integration with other departments that the EHR offers is of limited value. Consequently, these departments only intended to adopt the EHR if it was compatible with their current IT vision, systems, and other equipment. In other words, their adoption was conditional. Further, their adoption intention tended to be selective in that only the prospect of previously unavailable or improved functionality that would render their operations more efficient would lead to the EHR being embraced. As Greenhalgh et al. [10] suggested that small EHRs may be more efficient than larger ones; for Type II departments an organization-wide EHR may not be able to beat their local dedicated digital systems. Our theoretical interpretation of these observations leads to three propositions. Externally oriented departments (Type III) tend to have externally imposed data requirements from stakeholders they are reliant on for their legitimation or resources [41], such as government or financers, that lead to more decentral decision-making [44]. They often have or require customized IT-supported interfaces with the outside world, such as suppliers, customers, and partner organizations.
These departments feel a need to optimize their external relationships through dedicated technological interfaces, or at least compatible systems that allow (future) digital data exchange. Further, our study suggests that a department's evaluation of its current task-technology fit may affect the influence of extraorganizational dependencies on a department's adoption intention. If the current fit is experienced as low, and hampers performance, departments seem more open to the new EHR. Therefore, we propose the following: The model in Figure 1 summarizes the propositions developed above based on our interpretation of the results summarized in Table 4.

Implications for research
Our study and the propositions that follow from our findings build on the socialpsychological nature of behavioral intentions [45]. For the departments studied, we find that the interviewees' perceptions of the dominant work context characteristics affected their adoption intentions [37,46]. Second, our proposed model (Figure 1) includes fit variables related to the existing technology and organization of the clinical departmental work. That is, the nature of an EHR embraces the logic that intra-organizational dependencies will be served by the standardization of the internal data model and information flows. This logic aligns with the interests of those departments that have intra-organizational work dependencies, but is less relevant for departments that operate relatively independently and for those whose critical operations are interwoven with external partners, whose output is tied up with external networks, or who have to conform with medical professional associations. This 'fit' aspect of the proposed relationships accords with the explanatory mechanisms in adoption studies on the organizational level [47] and especially with enterprise system alignment studies [48,49]. Indeed, a system can prescribe the mode, the content, and the temporal aspects of clinical work [19], and these prescriptions may not align with critical characteristics of a department's work context.
Our study focused on contextual differences between departments, and we expected differences in the socio-political context, existing or arising during the organizational change process [e.g. 50,51], to operate at the department level. We, therefore, looked at these characteristics separately, and in combination with work context characteristics, for alternative explanations for departmental adoption intentions and subsequent EHR uptake, but failed to identify any. For example, in socio-political terms, Departments C and H both saw themselves as having no or little influence although they were actively supportive of the change process, yet both their intended and realized adoption differed. Departments C and G showed that such subunits can actively support an organizational implementation process without having a high adoption intention. Our reading of the socio-political context and process data is thus complementary to, rather than competing with, the developed propositions. For example, we saw how a clinician acting as a local project champion (a change process characteristic) can strengthen a department's adoption intention and how frustrations about a lack of communication (another change process characteristic) reinforced reservations about the system.
We argue that contextual characteristics are a reality that has to be dealt with. That is, although socio-political and organizational change process factors may influence how contextual characteristics are dealt with, and the weight they are given, the former are unlikely to change the direction of the contextual influences. It follows that these should be analyzed separately, and therefore that any quantitative follow-up research aimed at testing the proposed model should control for sociopolitical and change process factors.
Another avenue for future research would be to study the extent to which the constellation of department types within an organization influences successful adoption at the organization level. To the extent that department management has a formal say in the local adoption decision, and that a 'big bang' is not imposed, the order in which clinical departments adopt an EHR over time is also relevant. Rogers [52] showed that early adopters pave the way for subsequent adoption by other

Limitations
We argue that, within the time horizon of a project, the characteristics of the work context are more stable and accessible than those of the socio-political context and, for this reason, are a relevant focus for implementers. Nevertheless, one might still ask how permanent work context characteristics are across the adoption stages in long-term projects. It has even been shown that the implementation process can affect a department's characteristics [53], which may need to change because deploying an organization-wide system does not in itself create integrated and seamless processes [48]. Moreover, different contextual characteristics might be relevant in different stages of departmental adoption, as has been shown to be the case at the individual level [54].
Due to our focus on the clinical department level, we acknowledge that we may have overlooked factors on the hospital and individual levels that might codetermine departmental adoption intention. However, direct influences from these other levels did not come to the fore during this study's interviews and meetings.

Acknowledgments
We are hugely indebted to the project managers and hospital employees for their trust and willingness to participate. We highly value their cooperation and time invested. We are grateful to the students that assisted us in our efforts at different points in time during the data collection.

Authors' contributions
MO, AB and JV jointly developed the study design and its underlying conceptual framing, They equally contributed to the data gathering and analysis. The three authors read and approved the final manuscript.

Funding
There was no funding for this research project.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
As this study did not involve research on patients or human subjects, no Medical voluntarily orally agreed to participate in this study. They allowed us and to use the data they provided, including the use of quotes, under the condition of confidentiality.

Consent for publication
The management of the hospital granted access to interviewees, committee meetings and documents, and consented publication under the condition of confidentiality of the hospital, the departments involved and the individual interviewees. Those who attended meetings were informed about the research and consented presence during meetings. All interviewees participated on a voluntary basis and were granted confidentiality. Further, in feedback sessions to the project and to single departments, our reported data and interpretations were checked by interviewees and project managers. We obtained oral permission from interviewees and from the organization's management to use the data, including quotations from the interviews, for scientific publication purposes.    Additional Files Data Structure-Codebook-Case descr.docx