Integration of a nationally procured electronic health record system into user work practices

We conducted a longitudinal comparative case study undertaking field work between February 2009 and November 2010. The majority of data were collected in three "early adopter" National Health Service (NHS) specialist care providers (henceforth referred to as 'sites') implementing early releases of Lorenzo in England. We conceptualised these as case studies to take into account local contingencies [29, 30, 32‐35]. Our focus was on exploring the micro-environment of implementation and adoption, but without neglecting the wider implementation context (i.e. the national implementation environment), which was found to impact heavily on local developments [27, 28].

We purposefully sampled sites from among the first few to implement Lorenzo software. Sampling criteria included the range of different settings (e.g. mental health, community, acute) and the timeframes of implementation (i.e. it had to begin during our study period).

Within each case study site, implementation team members were initially approached for interviews and observations through the local head/director of IT. We then used snowball sampling, asking individual participants to recommend further interviewees, actively seeking a range of different viewpoints. Individual users were initially recruited with the help of IT managers, ward managers and heads of service (again asking for recommendations of further interviewees). In doing so, we sampled a range of staff groups including both junior and senior nurses and doctors, allied health professionals and administrative staff.

As work practices are heavily context dependent, we focused on collecting qualitative data through interviews, observations and documents in the three selected sites. These data collection approaches helped us to gain an insight into local contingencies, as well as more general processes across settings and over time.

We conducted semi-structured interviews with users (i.e. those who were expected to use the technology within their everyday work including doctors, nurses, allied health professionals and administrative staff) and other hospital staff (e.g. managers, IT staff and clinical leads) in order to gain an insight into their experiences of using the new system. Particular issues explored included views and attitudes relating to the system and implementation strategy (if relevant over time), perceived challenges and changes to individual and collaborative working, and potential suggestions for improvements (see Table 2 for an indicative topic guide). The wide range of interviewees sampled allowed us to gain an insight into the consequences for individuals as well as for wider organisational functioning, which is dependent on effective collaborative working. The first round of interviews (Time 1) was conducted during early use, when sites had implemented the technology for between three months and one year. Where possible, interviews were conducted longitudinally, speaking to the same stakeholders around six months later (Time 2) in order to trace potential changes in attitudes and investigate the effects of more embedded use of the system. This temporal component also helped us to assess the local consequences of the continuously changing political and economic climate. This was particularly important in the context of the national implementation, as changes in national arrangements often had significant consequences for hospital staff. These included, for example, delays in implementation timelines (or conversely pressures to implement), and changes in strategic direction such as reduced software functionality [27].

In order to gain an insight into the more informal work practices of users, interviews were complemented by a mixture of object/activity/person-oriented observations of the software in use with on-site questioning, where appropriate and convenient. This involved "following the thing" (i.e. the software) and the various ways in which it was used and talked about in the clinical setting (e.g. by observing activities that involved the software) [36].

Hospital-specific documents, including project initiation documents, 'lessons learned' documents and local evaluation reports, were also collected in order to investigate the more formal (i.e. official or desired by management) changes to work practices. These included documents drafted before the system was implemented, outlining planned changes to work practices. We used these as a comparator to actual observed changes.

In order to examine concurrent developments in the wider political and economic landscape, we conducted additional semi-structured interviews with stakeholders outside sites. Interviewees here included governmental stakeholders, developers and representatives from the independent sector.

Transcribed interviews, observations, researcher field notes and documents were coded with the help of NVivo8 software [37], drawing on both deductive and inductive coding approaches [38]. We began by building a coding framework based on the existing literature surrounding technology implementation and associated challenges with integration into user work practices [6]. This formed the deductive part of our analysis and included dimensions of integration of the system with professional responsibilities, existing roles and routines, and changes in individual and collaborative working. The inductive part of the analysis involved recording of new themes emerging from the data. Evolving categories from these themes were fed back into subsequent data collection. In investigating how exactly the software shaped both individual and collective work practices (i.e. the micro-context), drawing on ANT proved particularly useful [39, 40]. In doing so, our focus was on the active impact of the new technology on work practices of users, as it was beginning to replace established paper processes in the organisations. These insights were then used to draw conclusions about the wider environment in which these practices were situated by a constant process of "zooming in" and "zooming out" [39, 40]. This approach helped us to build on the literature and gain a holistic picture of the implementation and adoption landscape whilst still allowing new themes to emerge. As the focus of the study was on exploring changes to user work practices, this category was investigated in more detail first within and then across cases and data sources as well as time points [41]. Triangulation was facilitated by using the query function in NVivo8, which allowed extracting data from different data sources and collection times, and drawing up thematic tables recording emerging issues in each data source, time-point, group of interviewee, and individual location.

We discussed findings in regular team meetings and kept a research journal capturing our emerging understanding of the cases. During the analysis process, we paid particular attention to reflecting on our own assumptions and the way we had reached our interpretations, seeking potential alternative explanations. This was achieved by testing whether these better explained the data, ensuring that findings did not occur due to chance, searching for evidence that may refute explanations, and by following up unexpected findings [41].

The observational component of this study received ethical approval from the East London and the City Research Ethics Committee on the 2nd of April 2009. The interview component submitted for ethical review to the same ethics committee but was classed as a service evaluation on the 9th of October 2008 (08/H0703/112). To protect the anonymity of participants and Trusts, we have removed potential identifiers wherever possible.

The limited ability to customise the software meant that it was difficult to adapt to suit local needs, which often resulted in significant increases in users' workloads.

The nature of Lorenzo also meant that it had to be implemented in phased releases as increasing functionality became available. Consequently, paper and electronic systems had to be run in parallel initially in order to compensate for the limited functionality of the early releases. Acknowledging that this temporary arrangement may have increased workload for end-users yet further, local implementation teams suggested possible workarounds for users to cope with these parallel systems. These, for example, included the:

We in addition observed a number of more unintended workarounds and coping strategies employed by users struggling to accommodate software that was perceived to be of poor usability. Some users, particularly those with more autonomy such as senior consultants, resisted use altogether by insisting on using paper records.

Others, notably those who could not avoid using the system, devised various ways to compensate for the increasing demands on their time and concerns about the shortcomings of the technology. The most commonly employed techniques here included using other systems to compensate (e.g. typing letters in Microsoft Word as the spell-check was perceived to be much faster than that embedded in the software and/or reverting to paper systems); partially using the system (e.g. not recording certain activities if they were not viewed as important); and using the system in ways other than intended by management (e.g. getting around compulsory boxes by cross-referencing).

Over time, some of these difficulties of integrating Lorenzo within users' everyday work practices attenuated to some extent as they became more familiar with the system. This was particularly true in smaller-scale deployments in sites that had invested significant time and resources to adapt the software to fit with their everyday practices. In addition to more effective integration with work practices as a consequence of increased familiarity, stakeholders reported at follow-up interviews that, as they received system upgrades, technical performance had improved significantly, particularly in relation to speed.

Coping strategies employed by users had several direct and indirect knock-on effects. These are considered below in relation to collaborative working, patients, paper records, managerial outputs and recording activity.

ANT helped us to conceptualise changes in constellations between various network components including humans, paper and the technology. This is illustrated in Figure 1, which depicts existing work practices before the introduction of Lorenzo, planned changes in work practices with the introduction of Lorenzo, and actual changes in work practices resulting from the introduction of Lorenzo.

The decision to nationally procure EHR systems, coupled with the deliberate decision to limit customisability, resulted in significant changes to user work practices. Users had to adapt to a system that they did not believe was fit for purpose. If they could not resist use, they tended to employ workarounds which were often unanticipated by management. These had several knock-on effects on other factors within the hospital, including hierarchical structures and patterns of communication within the multi-disciplinary healthcare team, the time and quality of interactions with patients, paper records (which became more distributed), the timeliness of the record itself, and managerial outputs (which became less predictable).

Drawing on a range of data sources has allowed us to triangulate our findings and gain an insight into both formal and informal work practices associated with both intended and unintended workarounds [42]. ANT proved useful in guiding our data collection by placing the technology at the heart of the investigation, helping to appreciate the fluidity of sociotechnical processes, and the active role of the technology on other factors around it, both human and inanimate. It however proved less useful in explaining changes in the technology resulting from the way it was received by users and organisations. Other theoretical lenses may help to address this issue, such as for example the Strong Structuration Theory and the Social Shaping of Technology [40, 43], which emphasise the mutually shaping relationship between structures and agents.

Our longitudinal design meant that we were able to explore certain changes and developments over time, but this was still, however, limited to early implementation and adoption. We therefore observed certain ways in which users accommodated the system over time, but despite spending many months in the field we did not have the opportunity to assess the more embedded use of the software. Given our limited funding, we were also unable to develop a detailed understanding of the way healthcare workers used paper systems by examining work practices before the introduction of the new system, although this has often been advocated as an important first step towards understanding how electronic systems are likely to affect individual and collaborative work [42, 44‐46].

In addition, it has to be kept in mind that: 1) work practices are not static, but constantly changing/adapting (which complicates planning) [47]; and 2) work processes are contingent upon situations [48]. It may be, therefore, that over time and in different contexts users may have developed more (or indeed less) effective ways of using the technology. Consequently, despite our results being likely to be transferable to similar settings (i.e. sites implementing early releases of the same software), they may not apply to all care activities and situations.

Our results have shed light on the complex consequences and knock-on effects of a nationally procured EHR system for the work practices of a range of users. We illustrate these diagrammatically in Figure 2.

Usability of the system was perceived to be lacking, as the design of the "one-size-fits-all" technology limited tailoring to user needs. It is commonly acknowledged that increased customisation can facilitate the introduction of technology within user work practices [49‐53], so the finding that the relatively inflexible nature of Lorenzo resulted in unintended workarounds is probably not surprising [21‐23, 54].

Users needed to change their work processes to fit in with the technology; and this could be expected with the introduction of a new system as new routines need to be developed and existing work practices inevitably change to some extent [14]. However, issues with poor system usability resulted in clinicians employing informal workarounds to accommodate a system that was viewed as lacking fitness for purpose and increasing workloads and stress levels, as well as taking time away from patient care. When examining the literature on 'workarounds', most of the responses users in our study demonstrated can be thought of as "essential hindrance workarounds". That is, they were used to get around perceived problems in systems that were seen to impact on the more direct task of care provision. Most of the workarounds we identified were also viewed as essential as they were designed to save time that was needed for adequate patient care [22]. However, paradoxically these workarounds, designed to reduce the workload of users, also made the use of the system very specific to users' own contexts, thereby hampering the possibility of meaningful communication between interoperable systems being used in different contexts [55].

We have outlined several knock-on effects of these workarounds. Changes to existing professional responsibilities, hierarchical structures and ways of collaborative working of the multi-disciplinary healthcare team have been found in previous studies investigating the impact of new IT systems in healthcare settings [16, 56‐59]. Similarly, the finding that new technology can adversely impact on the interactions with patients is reinforcing existing evidence [60]. However, the effects on other factors outside of the immediate environment of use (e.g. organisations and paper records) are indeed new findings, possibly because they are context dependent. These included paper (which became more distributed), managerial outputs (which were perceived to become less accurate), and the timeliness of the record (as data entry was reported as being increasingly delayed). Although previous studies have investigated the potential impact of 'workarounds' on the safety and quality of care, the mechanisms by which effects are observed are likely to vary across settings and may not be transferable to other IT applications [24, 61].

Our findings are worrying, considering the potential knock-on implications for the quality and safety of patient care (e.g. if an important record is not available when needed). It is therefore vital that these wider factors are considered when mapping anticipated changes to work practices, which needs to be routinely and regularly done by organisations preparing for changes associated with IT applications in healthcare [62, 63].

Considering the range of staff and contexts of IT use in healthcare settings, mapping of work practices needs to be local by definition. Likewise, emerging 'workarounds' are likely to reflect the idiosyncrasies of the individual setting and are therefore highly context-dependent. This is in itself at odds with a national implementation with limited system customisability. Here, work practices may be effectively mapped, but the reality of systems-in-use is likely to vary from planned patterns as informal and emergent work practices that characterise the dynamic and ever-changing hospital environment are often not taken into account [42, 64]. This, in turn, may result in a lack of attention being paid to consequences for other stakeholders. The national nature of the implementation means that the resulting mitigating actions are far too slow to be visible in a timely manner to users on the ground, limiting the potential for organisations to flexibly address unanticipated issues [61]. This is supported by the relative successes of small-scale EHR implementations characterised by extensive customisation of systems to suit local needs [65].

In relation to England's national IT strategy, a more participatory approach to development and implementation has repeatedly been advocated [6, 66, 67]. For example, Catwell and Sheikh have suggested that this should be characterised by early user involvement and development of a shared vision, formative evaluations and testing of prototypes to assess whether the system is perceived as usable, potential re-design so that it fits with users' needs, summative evaluation and benefits identification once implementation completed, and incorporation of issues identified along the way [67].

Engagement and user input in design is clearly important [6, 21, 66, 68‐71], but in line with our findings, it is likely to be difficult to realise in the context of a centrally procured system that is intended for implementation on a national scale. This was found to be the case in relation to Lorenzo, which was intended to be "co-creational" by involving users in system design. However, the result of the focus on interoperability was limited local input and a resulting application that did in many instances not fulfill user and organisational needs. In addition, the organic development of engagement was hampered from the start as organisations and users were mandated to implement and use Lorenzo. Our experiences and the wider evidence-base suggest that a potential way forward may involve implementing a combination of standardised core solutions and customised elements, corresponding to particular users and environments (also known as 'configurational technology') [72, 73].

Our findings provide important insights relating to the impact of nationally procured systems on both formal and informal work practices of users and collaborative working. There is clearly a need to understand better how coordination is achieved without IT in order to produce a system that has the ability to facilitate this coordination or indeed revolutionise it by making processes more effective [42]. This is likely to vary across different settings and will warrant detailed small-scale local studies over longer periods of time in order to trace how work is structured with paper systems, as well as the initial, continuing and more embedded use of new EHR systems. As part of this work, it may be helpful to explore these issues from a range of theoretical angles. These may include deliberations surrounding the classification of an EHR system as a medical device and associated considerations surrounding usability testing including assessments of potentially negative consequences; more specific discussion surrounding effective leadership in organisations that have a high degree of autonomy; and exploration of ways in which evidence-based practice can be effectively promoted and integrated within individual workflows and within political strategies.

With regards to the impact on the quality and safety of care, there is a need to assess outcomes identified in our study quantitatively, in order to determine whether the mechanisms identified are indeed contributing to adverse patient outcomes. We do, however, acknowledge that such studies will be challenging to design as outcome measures may not necessarily be traceable to the system and comparators may be difficult to identify [74, 75].

The implications for practice emerging from our research are two-fold: firstly, there is a need to allow local work process mapping and responses to emergent practices and perceived inadequacies in the technological design; and secondly, it is vital not only to assess workarounds employed by local users, but the reasons behind their adoption in order to be in a position to reflect on other downstream consequences.