Tensions in large-scale electronic health record implementations: insights from a meta-synthesis

Abstract

Objective

To synthesize knowledge on tensions characterizing large-scale electronic health record (EHR) implementations.

Materials and Methods

A qualitative meta-synthesis was conducted by searching Scopus, Web of Science, MEDLINE, and CINAHL databases to find studies focusing on large-scale EHR implementations in OECD countries. An extraction table was completed to describe key characteristics of cases, and instances of tensions were extracted within each study based on a conceptual definition.

Results

Twenty-six qualitative studies were included, covering eleven unique large-scale EHR implementation projects. Cases were in Europe (n = 6), North America (n = 4), and Southeast Asia (n = 1). Analysis yielded twenty-one types of tensions associated with five primary objects: people, power, resources, system, and vision. Twelve tensions were found in multiple cases while fifteen were associated with more than one object.

Discussion

Results are aligned with the notion that tensions are inherent to organizational phenomena, showcasing their enduring nature across geographic, temporal, and technological contexts. The diversity of these tensions and their associated object(s) refer to critical, interrelated components of EHR systems implementation that are exacerbated in large-scale projects, and which can affect the implementation across its entire lifecycle.

Conclusion

Stakeholders involved in projects to modernize healthcare through the large-scale implementation of EHRs are prone to experience multiple tensions. Attention to the emergence of the tensions identified in this study helps to understand their impacts on projects and stakeholders. Tensions and their associated objects undergird the sociotechnical nature of these complex projects and the need to manage them effectively.

Introduction

Electronic health record (EHR) systems have become pivotal in optimizing clinical practices and modernizing healthcare systems.¹ These integrated systems offer numerous potential benefits, including better patient care coordination, improved data accuracy, and more streamlined healthcare workflows. However, in practice, EHR systems often struggle to seamlessly integrate into existing workflows, and their anticipated benefits—both individual and organizational—can take time to materialize.^1,2 These issues are particularly relevant in the context of large-scale implementations that have a broad institutional scope, targeting multiple types of healthcare services (e.g. acute care hospital, primary care clinics, and long-term care organizations) across multiple sites, and a broad functional scope, aiming to support complex digital clinical and administrative workflows.

Previous research has identified key success factors of EHR implementations, such as strong financial support, intensive practice assistance, a commitment to collective action, clear objectives, leadership from the physician community, governmental backing, and a community-focused approach.³ At the same time, authors have found that the most significant barriers to large-scale EHR implementations were the substantial capital required and the lack of clear national or regional policies or standards.^4–6 However, less attention has been paid to the tensions that arise between conflicting elements during large-scale EHR initiatives, and which can negatively impact the process as well as the outcome of those initiatives.

These tensions are not merely technical but are deeply intertwined with the social, cultural, and political fabric of healthcare systems. Large-scale EHR projects, for example, often aim to standardize processes across multiple hospitals or medical units, each with its own unique workflows, specialties, and patient demographics. The tension between creating a unified system and allowing for local customization to meet specific needs is a recurring challenge.⁷ Moreover, these projects typically involve a diverse array of stakeholders, including government agencies, healthcare organizations, and vendors.^7,8 Aligning the differing priorities and expectations of these groups can create political tensions over how the EHR system should function. Effectively identifying and navigating the tensions that arise from conflicting demands is crucial to the successful large-scale deployment of EHR systems. This is especially important as governments worldwide continue to invest heavily in such systems. If these tensions remain unaddressed, they may hinder clinicians' adoption of the systems and prevent the realization of their expected benefits.^7,9

Given this context, the primary objective of this meta-synthesis is to deepen our understanding of the tensions observed in large-scale EHR projects. Specifically, this study seeks to answer the following research questions: What are the main tensions characterizing large-scale implementations of EHR systems? How do these tensions manifest themselves during the EHR implementation process? By addressing these questions, this research informs digital health policy and practice, supporting the more successful integration of EHR projects into complex healthcare environments.

Materials and methods

This section reports the steps undertaken to conduct a systematic assessment of literature on tensions in the context of large-scale EHR implementations. The meta-synthesis was conducted in accordance with established methodological guidelines.^10,11 Consistent with those guidelines, this process was highly iterative. The flow diagram is shown in Figure 1.

Figure 1.

Flow diagram.

Search

Owing to the transdisciplinary nature of EHR implementations, a comprehensive search strategy was used (see File S1).¹⁰ The search involved Scopus, Web of Science, CINAHL, and MEDLINE, selected due to their relevance, particularly in health and medical science literature.^12,13 The search was executed on December 1, 2023 and updated on November 6, 2024, retrieving a total of 9,039 unique citation references. The literature review process was organized and managed using Covidence.¹⁴

Screening and study selection

Eligibility criteria for screening and study selection were organized along the five dimensions of the PICOS (Population, Intervention, Context, Outcome, Study characteristics) framework used in Covidence,¹⁵ and detailed in File S1. Population criteria were based on key aspects of large-scale EHR implementations, focusing on studies examining the implementation of systems that have a broad institutional scope, involving multiple types of healthcare services (e.g. acute care hospitals and primary care clinics) across multiple sites where clinicians were key target users. Intervention criteria focused on EHR systems that have a broad functional scope, encompassing diverse clinical and administrative functionalities. Studies focusing on implementations taking place in single facilities or involving the deployment of a single type of functionality^16,17 were thus excluded from the sample. The context was restricted to implementations taking place in Organization for Economic Co-operation and Development (OECD) countries. The outcome criteria for inclusion were based on the diffusion, deployment, or implementation of the EHR system, rather than micro-level elements (e.g. individual-level adoption). Study characteristics criteria focused on qualitative empirical papers, consistent with principles of qualitative meta-synthesis, with language restrictions limited to English and French.

Initially, four team members (AM, GV, MR, and GP) jointly reviewed a subset of 200 randomly selected titles and abstracts to ensure that all team members shared a common understanding of selection criteria, wherein each reference was examined by at least two researchers. During the title and abstract screening phase, each of the remaining 8839 references was independently screened by at least one reviewer. A total of 142 full-text articles were selected and retrieved for further assessment. 118 of those articles were subsequently excluded, and two additional references were identified through manual search of retained articles’ bibliography, bringing the final total to 26 articles.

Quality assessment

The Joanna Briggs Institute Checklist for Qualitative Research¹⁸ was used to evaluate the quality of the studies included in our sample. Two reviewers (GP and MR) met and jointly discussed each of the instrument’s ten items to evaluate three randomly selected studies. They then independently applied the instrument to evaluate the remaining studies. The assessment concorded for 93% of the total number of items. Disagreements on the remaining items were resolved via discussion.

Data extraction and analysis

Data extraction and analysis were performed across seven key phases (see Figure 2; additional details are available in File S2). Three researchers (GP, GV, and MR) performed an initial extraction of key characteristics for each of the implementation initiatives in the 26 articles. Initiatives presented within each article were then treated as a unique case when they covered the implementation of the same technology, in the same setting, and across a similar period, yielding a total of 11 unique implementation cases. All members of the research team then prepared and shared a brief summary memo of 2-3 randomly assigned cases outlining key findings, depending on information available.

Figure 2.

Outline of data extraction and analysis.

In the fourth step, the research team extracted instances of tensions reported within the articles covering the 11 cases based on the conceptual definition of tension as “elements that seem logical individually but inconsistent and even absurd when juxtaposed” (p. 382).¹⁹ Each instance included a description of the tension with extracted quotes describing its inconsistent elements. One researcher (GV) subsequently reviewed all instances, meeting with other researchers to discuss their results, yielding a total of 101 instances of tensions. The same researcher then inductively grouped those instances under 42 unique descriptive labels. The analysis of the labels and their associated instances of tensions led to the inductive identification of 6 overarching elements (e.g. scope, timeline) which were referred to as “objects” of tensions to describe the elements of implementation initiatives affected by each instance. Two researchers (AM, GV) then reviewed the proposed labels and objects, iterating between the conceptual definition of tension and extracted quotes to further group them together. In total, 21 unique descriptive labels and 5 objects were retained to categorize all 101 instances of tensions. Finally, a third researcher (MR) reviewed all labels, objects, and their corresponding associations using extracted quotes. Disagreements were resolved through discussions among the three researchers (AM, GV, and MR), altering the association between 4 labels and their objects. The results were synthesized into a table for discussion with the other team members (GP, LR). To maximize trustworthiness,²⁰ all 5 researchers had access to coded data, meeting at least every other week to openly discuss progress and outstanding issues.

Results

Study characteristics

The 11 cases cover a variety of geographical, institutional, and technological contexts (see Tables 1 and 2), with 7 cases focusing on EHR implementations in Europe (e.g. Denmark,^6,24,36 Finland,⁶ and Sweden,^37,40) 4 in the United States (e.g. in Hawaii,^40,42 and Michigan,³⁴) as well as 1 in Singapore.³⁸ Regarding scale, some projects involve nationwide initiatives while others were restricted to specific regions, sites, or types of healthcare organizations. The case of the implementation of UK’s National Care Record Service (NCRS) is the most frequently studied, with 11 papers reporting on this widely documented case.^{26–31,35,38,41,43} Although Epic is the most frequently cited system, with 5 cases covering its implementation in Europe and in the United States (cases A, B, C, G, and I), cases also report on the implementation of systems such as Cerner,^{21–23,25,38,41,43,44} as well as others with unspecified vendors.⁴⁵ Consistent with the broad reach of EHR systems, these implementations often target large geographical regions or entire countries, affecting multiple sites and types of establishments including primary and secondary care units³⁰ as well as university and teaching hospitals.⁴⁰ Finally, the cases have been published from 2005 to 2024, with focal periods of analysis ranging from 1999⁴² to 2023,^6,25 with some cases focusing on a specific phase of the implementation process³³ and others relating initiatives that span close to ten years.³¹

Table 1.

Overview of studies.

Author(s)	Country	Location(s)	Focusa	Targeted systema	Focal period(s)a	Case(s)
Ahlness et al (2023)21	USA	Spokane	1 medical center as part of Department of Veterans Affairs	Cerner	Oct. 2020-Nov. 2021	J
Anderson et al (2024)22	USA	Spokane	1 medical center as part of Department of Veterans Affairs	Cerner	Sep. 2020-Nov. 2021	J
Ball et al (2024)23	USA	Spokane	1 medical center as part of Department of Veterans Affairs	Cerner	Sep. 2020-Nov. 2021	J
Bansler (2021)24	Denmark	Capital and Zealand regions	12 public hospitals	Epic	May 2016-Jan. 2019	A
Brunner et al (2024)25	USA	Columbus	1 medical center as part of Department of Veterans Affairs	Cerner	Feb. 2022-Apr. 2023	J
Cresswell et al (2011)26	UK	North, Midlands and Eastern regions of England	4 hospitals	National Health Service Care Record Service (Lorenzo)	Feb. 2009-Nov. 2010	F
Cresswell et al (2012)27	UK	England	3 hospitals (2 acute care; 1 community)	EHR system (“Software X”)	Feb. 2009-Nov. 2010	F
Cresswell et al (2012)28	UK	England	3 hospitals (2 acute care; 1 community)	Lorenzo	Feb. 2009-Nov. 2010	F
Cresswell et al (2012)29	UK	England	12 hospitals (8 acute care, 3 mental health, 1 community)	EHR software systems (unspecified)	Sep. 2008-Feb. 2011	F
Currie (2014)30	UK	England	From 28 to 10 GP surgeries and hospital trusts	National Care Record Service (NCRS) as part of NPfIT (unspecified)	2002-2007	F
Eason and Waterson (2013)31	UK	England	2 health communities	Detailed care record system (DCRS) as part of the NPfIT (unspecified)	2002-2011	F
Ellingsen et al (2022)32	Norway	Central Norway	3 regional hospitals, nursing homes, and home-care services	Epic	2022	C
Ellingsen and Hertzum (2020)33	Norway	Central Norway	3 regional hospitals, nursing homes, and home-care services	Epic	2016-2018	C
Gui et al (2020)34	USA	Michigan	1 academic medical center including 3 hospitals, 6 specialty centers, 120+ offices and clinics	Epic	2012	I
Hendy et al (2005)35	UK	England	4 acute hospital trusts	National Health Service Care Record Service	Unspecified	F
Hertzum and Ellingsen (2019)36	Norway	Central Norway	3 hospitals	Epic	2021	C
	UK	Cambridge	Cambridge University hospitals	Epic	2012 (signature)-2014 (go live)	G
	Denmark	Capital and Zealand regions	All hospitals in the two regions	Epic	2013 (signature)-2016 (go live)-2018 (ongoing)	A
Hertzum et al (2022)6	Denmark	Capital and Zealand regions	12 hospitals	Epic	2013 (signature)-2016/2017 (Go live depending on the hospitals)-2021	A
	Finland	Helsinki-Uusimaa region	Hospitals, primary healthcare and social care services	Epic	2016 (signature)-2018 (go live)-2022	B
Janols et al (2014)37	Sweden	Unspecified	1 university hospital. 1 small hospital, 40 primary care centers	EPR system (PAS, eReferral and eMedication)	Post evaluation of a 2.5 year project	E
Klecun et al (2019)38	UK	England	78 public hospitals	National Care Record Service (NCRS) as part of NPfIT (Lorenzo, Cerner)	2009-2011	F
	Singapore	Singapore	All public hospitals and clinics	National Electronic Healthcare Record (unspecified)	2006-2015	D
Mogård and Toussaint (2021)39	Norway	Central Norway region	1 hospital, nursing homes, home pharmacy, ambulatory services, retail pharmacies and homecare	Epic	2019	C
Ovretveit et al (2007)40	USA	Hawaii	1 hospital, 15 clinics	CIS	Unspecified	H
	Sweden	Unspecified	2 teaching hospitals	EMR (unspecified)	2005	E
Robertson et al (2010)41	UK	England	5 acute hospital and mental health trusts	National Care Record Service (NCRS) as part of NPfIT (Lorenzo and Cerner for acute hospitals, RiO for mental health)	2009-2010	F
Scott et al (2005)42	USA	Hawaii	4 primary healthcare teams in 4 clinics, 4 specialty departments in 1 hospital	CIS	1999-2004	H
Sheikh et al (2011)43	UK	England	12 hospitals (8 acute care, 3 mental health, 1 community)	National Care Record Service (NCRS) as part of NPfIT (Lorenzo Regional Care, Cerner, and RiO)	2008-2011	F
Takian et al (2014)44	UK	England	2 hospitals	Cerner	2008-2011	F
Umstead et al (2021)45	USA	Tennessee	12 clinical sites (inpatient, outpatient, and emergency)	Vendor-based EHR (unspecified)	2017-2018	K

^aAs reported in the articles.

Table 2.

Overview of cases.

Case	Description	Article(s)	Costsa
A	Implementation of Epic in two regions in Denmark	Bansler (2021)24 Hertzum and Ellingsen (2019)36 Hertzum et al (2022)6	DKK 2.8 billion (€375m) for the two regions
B	Implementation of Epic in the Helsinki-Uusimaa region in Finland	Hertzum et al (2022)6	€384m
C	Implementation of Epic in the Central Norway region	Ellingsen et al (2022)32 Ellingsen and Hertzum (2020)33 Hertzum and Ellingsen (2019)36 Mogård and Toussaint (2021)39	NOK 2.7bn (270m €)
D	Nationwide EHR implementation in Singapore	Klecun et al (2019)38	Unspecified
E	Deployment of three region-wide EPR system modules in Sweden	Janols et al (2014)37	Unspecified
F	Nationwide implementation of the UK National Care Record Service (NCRS) as part of the NPfIT program	Cresswell et al (2011)26 Cresswell et al (2012)27 Cresswell et al (2012)28 Cresswell et al (2012)29 Currie (2014)30 Eason and Waterson (2013)31 Hendy et al (2005)35 Klecun et al (2019)38 Robertson et al (2010)41 Sheikh et al (2011)43 Takian et al (2014)44	From £12.7bn41 to more than £20bn30
G	Implementation of Epic at Cambridge University hospitals in England	Hertzum and Ellingsen (2019)36	£200m (€275m)
H	Implementation of CIS in Hawaii	Ovretveit et al (2007)40 Scott et al (2005)42	Unspecified
I	Implementation of Epic at Michigan Medicine	Gui et al (2020)34	Unspecified
J	Implementation of Cerner at the Department of Veterans Affairs medical centers	Ahlness et al (2023)21 Anderson et al (2024)22 Ball et al(2024)23 Brunner et al(2024)25	From $39bn21–23 to $50bn25
K	Implementation of EHR in Tennessee	Umstead et al (2021)45	Unspecified

^aAs reported in the articles.

Results of quality assessment

The quality assessment of the 26 papers shows a generally high level of congruity across several key areas. All studies exhibited congruity between their research objectives, methodology, and the methods used for data collection. Most papers also demonstrated congruity between the research methodology and the interpretation of results. However, the congruity between the methodology and the representation and analysis of data was unclear in two papers. The assessment revealed that 14 studies had a statement regarding the researchers’ cultural or theoretical position while only one of the studies addressed the influence of the researcher on the research process. Detailed results of the quality assessment are available in File S3.

Tensions in large-scale EHR implementations

Overview

Inductive analysis revealed 21 different types of tensions across the 11 implemented cases, as presented in Table 3. These tensions relate to important aspects of large-scale EHR implementations associated with their objectives and the means employed to achieve them. For example, the ‘project goal’ tension highlights difficulties related to projects driven and supported by policies and agendas that seek to increase the standardization of healthcare delivery through technological projects. These policies and agendas can run against established practices and technologies that have previously evolved to fit local contexts.²⁴ Tensions related to features and functionalities of EHR systems showcase the reality that configuring these systems is a challenging process that requires balancing stakeholder needs and systems’ capabilities. For example, the promotion of EHR systems as flexible platforms can bring some stakeholders to believe that configuration choices can be reconsidered if needed while in reality, early choices cannot easily be undone as they drive subsequent configuration tasks.³⁶

Table 3.

Tensions in large-scale EHR implementations.

Tension	Description of the Duality involved in Tension
	Element A	Element B
Addressing configuration issues and bugs	Management and/or vendor drives the prioritization of configuration issues and bugs.	System users drive the prioritization of configuration issues and bugs.
Aligning practice and system	Work practices are typically routinized within a given context and do not change frequently.	EHR systems typically change existing ways of performing work.
Benefits realization	The benefits of large-scale EHR implementations are expected to be realized quickly following a short shakedown period.	Large-scale EHR implementations are complex and uncertain, taking time to yield potential benefits, if any.
Communication	Formal chain-of-command, reporting structures and culture define the flow of communication.	Informal communication channels enable rapid feedback during and following the implementation.
Conceptualization of benefits	Financial benefits (e.g. cost savings) are driving project success.	Clinical benefits (e.g. improved patient outcomes) are driving project success.
Costs	Large-scale EHR implementations require significant financial resources.	Large-scale EHR implementations are usually undertaken to standardize processes and reduce costs.
Data accuracy	EHR systems require the input of highly standardized, codified data to support digital workflows.	Clinical care often requires trial and error, and not every part of care can be codified.
Degree of localization	Large-scale EHR implementations aim to offer standardized care across settings.	Local settings (e.g. specialties, geographical regions, establishments) have, over time, optimized their processes to fit their reality.
Features and functionalities	Systems are built, configured and evolved based on pre-established decisions.	Systems are expected to be configurable and evolve based on the emergent needs of the local settings where they are implemented.
Patient record accessibility	Access to patient record is based on the definition of restrictive system roles and privileges.	Access to patient record is based on actual role in providing patient care instead of system roles.
Patient record ownership	Patient records are shared objects managed by multiple stakeholders.	Patient records are managed exclusively by the patient's primary provider.
Professional identity	Healthcare professionals are system users.	Healthcare professionals are experts focused on delivering patient care.
Project conceptualization	EHR implementation projects are technology projects.	EHR implementation projects are clinical transformation projects.
Project goal	Large-scale EHR implementations aim to improve localized care within settings.	Large-scale EHR implementations aim to standardize work practices across settings.
Project management	Local actors are best situated to make and enact project management decisions.	Large-scale EHR systems require formal, global management structures to centralize control over project management decisions.
Project timeline	Large-scale EHR implementations require significant amounts of time.	Financial and human constraints encourage attempts to execute implementation as quickly as possible.
Role of managers	Managers should focus on overseeing the implementation process, focusing on task delegation.	Managers should show support for the project by being actively involved in the implementation process.
Role of mentors	IT mentors aim to assist physician builders in performing configuration work.	IT mentors aim to control the configuration choices proposed by physician builders.
Source of implementation support staff	Internal implementation support staff are best suited to assist the transition to new EHR, but their availability is difficult to secure.	External implementation support staff are less effective in the transition process, but they are easily mobilized.
User participation	Healthcare professionals have demanding schedules.	Extensive user participation and input is crucial to successful EHR implementation.
User training	User training is an important activity that requires time and must take place prior to go-live, when the system is not yet fully configured.	User training is most effective when performed using a system that is already configured and used in the context of prospective users' work.

Analysis

The inductive analysis led to the identification of five overarching objects of tensions in large-scale EHR implementations. Table 4 presents tensions based on those five objects along with their associated case(s), ordered by the number of tensions associated with each object, in decreasing order.

Table 4.

Tensions in large-scale EHR implementations.

Tension	Object					Case(s)
	Peoplea	Powerb	Resourcesc	Systemd	Visione
Professional identity	●					C, F
Communication	●	●				C, F, H
Role of mentors	●	●				A
Role of managers	●	●				E
Degree of localization	●	●		●		B, E, F
Patient record "ownership"	●	●		●		F
Source of implementation support staff	●		●			K
Benefits realization	●		●			A, E, I, J
User participation	●		●			A, C, E, J
User training	●		●			E, I, J
Alignment between practice and system	●			●		A, E, F
Patient record accessibility	●			●		J
Features and functionalities		●	●	●		B, F, A, C, J
Project management		●			●	D, F
Addressing configuration issues and bugs		●		●		A, C, F, J
Costs			●			F
Project timeline			●		●	A, F, G
Conceptualization of benefits					●	B, C
Project conceptualization					●	F
Project goal					●	A, F, C, I, K
Data accuracy				●		F

^a People: tensions that relate to individuals and teams, including their skills, roles, identities and responsibilities in the implementation project;.

^b Power: tensions that relate to the exertion of authority and influence from stakeholders that affect the direction and execution of the project;.

^c Resources: tensions that relate to tangible and intangible assets, including finances, tools, human resources, and time required to execute and support the project;.

^d System: tensions that relate to the technical infrastructure, software features and functionality necessary for the successful operation of the system;.

^e Vision: tensions that relate to the goals, objectives, and overall scope of the project, guiding its direction and desired outcome(s).

People

The most frequently found object was People (12 out of 21 types of tensions). Whether tensions relate to communication (cases F^26,29 and H⁴²) user participation (cases A,²⁴ C,³⁶ E³⁷ and J²²) or benefits realization (e.g. cases I³⁴ and case A^6,36) among others, they reflect the reality that large-scale EHR implementations primarily involve and affect human actors. For example, in the case of Sweden (case E) managers focused on delegating tasks to super-users, which was in tension with the idea that the direct involvement and the support of management in the implementation process increases the chances of success of those projects.³⁷ Similarly, training is a crucial aspect of EHR implementations as it allows future users to become acquainted with the system and its functionalities. However, as illustrated again with the case of Sweden and that of Michigan (case I), the need to perform user training prior to the go-live often means that training sessions remain “generic,”³⁴ going against the need to contextualize user training based on actual expected use cases while giving future users the time they need to familiarize themselves with the new system.

Power

Power is the second object most frequently associated with tensions (8 out of 21). These tensions reinforce the notion that large-scale EHR implementations are often affected by power and politics, like other types of system implementation initiatives.⁴⁶ For instance, in the case of the implementation of the UK’s NCRS (case F), higher-level decision-makers prevented users from sharing knowledge and information with other implementation sites, revealing a tension between the impetus for users to do what is best for the project—share information, and the need to obey rules that prevent such sharing.³⁸ Tensions that relate to the decisions that need to be made regarding the project’s execution and the system’s configuration also reflect important power issues. For example, during the implementation of Epic’s system in Norway (case C), “physician builders” were expected to make thousands of system configuration decisions. At the same time, the implementation took place in a “complex decision structure, sometimes requiring top-level organizational approval,”³³ contradicting the notion that these physicians would have complete authority over configuration decisions.

Resources

Large-scale EHR implementations require the commitment of significant technological, temporal, financial, and human resources. Unfortunately, it is often difficult to ensure that these resources are available at the right time and in sufficient quantities, highlighting an enduring issue that plagues the management of many large-scale projects, technological or otherwise. In the sample, 7 tensions were related to resources. For example, in the case of the implementation of Epic’s system in Denmark (case A), the vendor encouraged the creation of a “pilot project with 11 physicians from various specialties to test the concept of physician builders.”²⁴ However, the reality was that such resources could not be released to participate in this pilot, revealing a tension between the need for participation of physicians to increase the chances of success of EHR implementations, and the actual lack of available physicians to participate in these programs. Similarly, in Sweden (case E), a tension surfaced between management who invited healthcare professionals to be involved in the project and provide critical input to drive the configuration of the new system, and the subsequent dismissal of that input. One respondent noted that “if you ask the health professionals for advice, it is of utmost importance that you ask the right questions and that you listen to their responses; if not, their frustrations will increase.”³⁷

System

The fourth object, System, is associated with 7 out of 21 tensions. These tensions incorporate important elements that relate to the actual EHR system that is being implemented. Accordingly, tensions frequently occur on the topic of the system’s features and functionalities. For instance, in the case of Norway (C), a tension surfaced based on the notion that “Epic is promoted as a flexible platform” which contradicts the fact that early decisions “may “freeze” practices and technology in a way that makes them hard to change later.”³⁶ In the case of the NCRS (F), for instance, tensions were observed when users who reported technical issues that prevented proper use of the system saw that they “often remained unresolved for extended periods”—an issue which was also observed in case J^22,23,25—because of the bureaucracy in place which took care of prioritizing work.²⁶ In the same case, the implementation of centralized, electronic health records meant “out-of-hours doctors could insert diagnoses that the patient’s [general practitioner] GP did not believe were accurate,”,\ which ran against the GPs’ longstanding notion that those records where “their record in the sense that they controlled what went into it and who had access to it.”³¹

Vision

Large-scale EHR implementations are undertaken based on a vision that defines the outcomes to be achieved by the system. Consistent with the importance of this aspect, 5 tensions in 8 out of 11 cases were found. An interesting example of tension associated with this object is the conceptualization of benefits that should result from the implementation. In the implementation of Epic in Finland (case B), it was expected that the implementation would “come with cost savings”. However, this ran against the notion that “EHRs tend to increase documentation requirements—in the interest of improving patient care and research data.”⁶ In other instances, as illustrated previously with tensions related to project goals, different stakeholders within governments and healthcare organizations (e.g. managers and physicians) each have their own vision for the project. However, taken together, these visions often collide with one another. For example, in Michigan (case I), “physician champions” had to reconcile the execution of two concurrent visions: one that required “customizing the new EHR system to meet the local workflow and documentation requirements for their clinical areas”, and one which involved “changing their practices and workflows after go-live to fit the system design.”³⁴

The complexity of tensions in large-scale EHR implementations

The analysis also reveals the complexity of the relationships that exist between each of the five objects identified. Indeed, 15 out of 21 tensions are associated with more than one object. For example, tensions on the degree of localization that should be supported by an EHR relate to both people who are involved in the implementation process and are making decisions that affect the configuration of the system as well as the power of those actors relative to one another. In the case of the NCRS (F), a tension permeated between the roles of individuals to provide input to configure and tailor the system to local needs, and the fact that the project was driven by political decisions that held considerable power over the extent of localization allowed. Accordingly, it has been noted by researchers that “results have also pointed to the importance of considering standards for larger scale interoperability. However, these concerns seemed to be secondary to most organizational stakeholders as local accommodation needed to occur first.”²⁹

Tensions related to user participation as well as user training, among others, are associated with both people and resources. For example, the decision to involve external implementation support staff to help with the configuration of the new system is a decision that involves people—external consultants instead of internal staff, but it also involves the allocation of resources as this decision is driven by constraints related to employee availability and financial costs (Case K).⁴⁵

The analysis further reveals that two tensions associated with the object of power were associated with the vision of the project. These tensions encompass the notions that the goals, objectives, and the scope of a large-scale EHR implementation influence the power structure in charge of the implementation. In the case of Singapore (D), the system was “planned, designed, and managed, top-down” while during the implementation itself, chief medical information officers worked hard to ensure that the system’s design would “meet doctors’ requirements,” working against the power exerted by the authorities who had established the project’s initial vision.³⁸ In the implementation of Cerner’s system at the Veterans Affair in the USA (case J), a tension on patient record accessibility associated with both the people and system objects was observed. CPRS, the previous system, relied heavily on user training and system audits to enforce appropriate access to patient data. In contrast, the new system is described as “roles-heavy”, with the creation user roles that differed significantly from the previous system, leading to situations where “individuals were granted access to a limited tasks which precluded their usual work.”²¹

Discussion

Principal findings

The analysis of tensions and their associated objects yields four key observations. First, many tensions surface across contexts that exhibit different characteristics. For instance, tensions related to project goal were found in 5 out of 11 implementation cases, including in Denmark,⁶ Norway,³² the United Kingdom,⁴³ and the United States.⁴⁵ Overall, this highlights the enduring nature of tensions as an inherent component of large-scale EHR implementations that must be managed, irrelevant of time, location, system, or geopolitical environment, consistent with prior research in project management.⁴⁷

Second, there exists a wide variety of tensions that can affect large-scale EHR implementations. These tensions relate to both the goals of the implementation as well as the means dedicated to its achievement. Although it is expected that some of those tensions exist regardless of the scale or the scope of the EHR implementation project, they are exacerbated in the context of large-scale implementations. For instance, in our context, tensions associated with power and vision involve other stakeholders who may be less visible in small-scale projects (e.g. government agencies leaning toward standardization of care). Another example is the ability for stakeholders to participate and to perceive that their involvement has a significant impact on the project’s outcome given its broad scope.⁴⁸

Third, although it is important to consider each of the five objects—people, power, resources, vision, system—in its own merit, it is not possible to reduce all tensions to a single object as this undermines their inherent complexity. The findings of this study therefore contribute to highlight the importance of studying the relationships between context, technology, and people as a sociotechnical system to understand the unfolding and the outcomes of large-scale EHR implementations.^49–51 Large-scale EHR implementation projects involve human stakeholders, who each have their own interests and objectives, working together to implement some form of configurable technology that will digitize, or alter, well-established routines and processes.⁵² It is of the utmost importance to understand the scope and the depth of these changes beyond their instantiation into a technological artefact.^52,53 As reported in the case of the implementation of Cerner’s system in the UK (case F), the chances of success for these large projects increase when organizations execute them based on the idea that they are “a multidimensional and sociotechnical journey.”⁴⁴

Fourth, evidence from the cases suggests that the relevance of tensions may depend on the “moment” of the implementation process. Research breaks down the execution of system implementation projects across four main phases.⁵⁴ During project chartering, key parameters that will drive the project’s business case are defined, including financial and technical requirements. It is during this first phase that the initial vision for the project is built. This will subsequently drive much of the execution of the implementation, as evidenced by tensions that fall under the ‘vision’ object (e.g. tensions related to the degree of localization). During project configuration and rollout, the system is built and/or configured, eventually going live. It is during this phase that important activities involving prospective users are carried out and where tensions associated with the “people” and “resources” objects can be expected to surface. For example, tensions related to user training as well as user participation, which are critical to system implementation projects,⁵⁵ will likely be more salient during this phase. During the shakedown phase, the newly implemented system is stabilized, with the objective of routinizing system use. It is during this phase that tensions related to the “system” and “power” objects may be expected to surface. For example, a tension can surface regarding the prioritization of bug fixes if it depends on the relative authority of stakeholders.⁴¹ Finally, during the onward and upward phase, which takes place after the go-live, the system is going through routine maintenance and upgrades. During this phase, tensions that initially drove the execution of the project may resurface. For example, in the case of the NCRS, the initial vision of the project changed to allow for a more pragmatic vision that accommodated localized adaptations through the implementation of different, interoperable systems.³¹ In addition, some tensions are susceptible to surface at any point in time during the implementation process. For example, setting up centralized, formal means of communication for the entire project can run against the need for actors to quickly share information using informal channels.²⁶

Implications for practice

The findings discussed above have implications for stakeholders involved in the governance, planning, realization and evaluation of large-scale EHR implementation projects. These findings inform the need to be aware of and engage with the tensions that are more likely to emerge in context. However, to do this and to manage those tensions first requires that they be clearly identified and acknowledged as an “opportunity” to be seized by policymakers and project decision-makers rather than as a “problem” to be avoided.⁵⁶ For instance, tensions on the degree of localization highlight the importance of ensuring that a vision to implement highly standardized processes using an EHR is not imposed on those who are directly involved in the provision of care and the performance of clinical workflows without careful consideration of their needs.

The tensions identified in this study and their associated objects also serve as a reminder that large-scale EHR implementations are highly complex. Although prior research has highlighted the importance of the high degree of complexity associated with technology implementation in healthcare,^9,57 the tensions identified here offer a pragmatic outlook on specific aspects that require attention. These findings also help to stress the importance of considering that both elements involved in a tension are sensible on their own; it is their juxtaposition that creates a contradiction. Doing so helps to view how addressing tensions may involve strategies that seek to achieve ambidexterity,¹⁹ for example, by tending to their constitutive elements simultaneously, or by alternating between elements at different points in time, thereby engaging with the dynamic nature of tensions (ie, including aspects such as emergence, acceptance, accommodation, and resolution) during the implementation process.

The classification offered here may be used by policymakers and practitioners to identify tensions that are likely to surface, and ascertain whether policies, implementation strategies and project management practices in place will enable or facilitate their resolution. Doing so can inform the prevention of unintended, undesirable consequences associated with the large-scale implementation of EHRs^58,59 to maximize their chances of success.¹

Additionally, this classification could constitute the foundation for a diagnostic tool to recognize and apprehend the complexity of tensions inherent to large-scale EHR implementations⁶⁰ by suggesting appropriate (context-sensitive) management strategies. For instance, in implementations where a gradual rollout of the system across multiple sites is carried over time, enabling knowledge transfer between stakeholders across multiple sites can help to alleviate some of the tensions related to communication observed in our sample.³⁸

Limitations and avenues for future research

There are certain limitations that are intrinsic to the qualitative meta-synthesis method applied in this study, owing to authors’ presentation and interpretation of case data. First, it was often impossible to ascertain at what point of the implementation process a tension initially emerged, preventing the adoption of a diachronic perspective of tensions. Second, data on the implementation approach (e.g. big bang vs. gradual rollout) as well as the main stakeholder in charge of the implementation process (e.g. provider vs. target organization) were not always clearly identified. While this observation prevented comparisons across cases on these aspects, it provides a call for future research to provide more contextual information that can help understand the complexity of the sociotechnical context where the implementation takes place.^49,51 Third, cases provided little evidence regarding the responses implemented to address tensions. While this is consistent with the objectives of each paper within the selected sample, it compels an invitation for future research to further explore this important aspect. Future studies on large-scale EHR implementations are also encouraged to use alternative theoretical lenses to further our understanding of this phenomenon. Moreover, longitudinal case study designs, preferably using a participant observation or action research method, would allow both researchers and practitioners to further apprehend tension dynamics.

Author contributions

Gregory Vial (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review and editing); Aude Motulsky (Conceptualization, Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review and editing); Mickaël Ringeval (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review and editing); Louis Raymond (Conceptualization, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review and editing); Guy Paré (Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Resources, Writing – original draft, Writing – review and editing).

Supplementary material

Supplementary material is available at Journal of the American Medical Informatics Association online.

Funding

Funding was provided by the Quebec Ministry of Health and Social Services (#11979).

Conflicts of interest

None declared.

Data availability

The data upon which this article is based are available in the article and in supplementary material files.