Introduction

UK General Practitioners (GPs) manage a high and rising workload of increasingly complex patient care with many competing demands to attend to within time-limited consultations [1]. This, and ongoing recruitment and retention challenges, has led to a GP workforce ‘crisis’ [2,3,4,5]. The COVID-19 pandemic has introduced further pressures on general practice, with associated back-logs of consultations, diagnoses, and referrals [6,7,8,9]; GP workload therefore continues to be an increasingly pressing issue for health professionals, patients and policy makers.

Clinical decision support (CDS) tools are used by health professionals to assist with clinical decision making in relation to screening, diagnosis and management of a range of health conditions [10,11,12,13,14]. Many CDS tools exist for use in primary care and more recently are being embedded in electronic form (eCDS) within practice IT systems, drawing directly on data within patients’ electronic medical records (EMR) for their operation [11, 15, 16]. Many Clinical Commissioning Groups and Primary Care Networks have supported the introduction of eCDS tools that facilitate diagnosis and expedite referral for certain conditions, such as cancer, particularly since the COVID-19 pandemic [17]. For the purpose of this article, an eCDS tool is defined as any electronic or computerised tool which provides an output pertaining to a possible diagnosis and/or management of a health condition, using patient-specific information.

The workload implications of GPs using eCDS tools during consultations are unclear. One way of examining GP workload is to evaluate the duration of consultations [18], although that is only a single element of GP work, not including time taken for managing referrals, investigations, results, and general administration, undertaking training, and supervising colleagues [19, 20]. The duration of consultations and the ‘flow’ of patients through consulting sessions, however, provide key ways of measuring workload as these have an impact upon GPs’ levels of stress throughout the working day [21,22,23]. Understanding whether using eCDS tools impacts consultation duration and patient ‘flow’ through consulting sessions may help facilitate the implementation of eCDS tools into practice.

Here we aimed to establish if there is pre-existing evidence on potential workload, including impact on consultation durations, associated with the use of eCDS tools by health professionals in general practice and primary care. The objective of this literature review therefore was to identify the available evidence on using eCDS tools and analyse their impact on workload.

Methods

A systematic scoping review was undertaken to identify literature using the stages set out in the Arksey and O’Malley methodological framework, enhanced by more recent recommendations [24, 25]. This method enables examination of the extent, range and nature of research activity with an aim of identifying all existing relevant literature.

A broad research question was used: What is known from the existing literature about the use of eCDS tools by health professionals in general practice/primary care and the associated impact on workload and patient ‘flow’ through consulting sessions?

An initial scoping search was conducted using the databases: MEDLINE (Ovid), HMIC (Ovid) and Web of Science (TR). Keywords from titles and abstracts identified by this search, and index terms used to describe these articles, were identified (see Fig. 1). A second search across the same databases was then undertaken using the identified keywords and index terms, and studies collated for abstract and title screening to identify relevant full-text articles to be reviewed. The searches were conducted in September 2019 and updated in August 2021. The review extensively targeted articles in written English, and published in a ten-year period prior to the initial search date. This time period was selected in order to identify research on eCDS in the context of today’s general practice and primary care, and to manage review in context of available resources. A comprehensive search strategy and set of search terms used is provided in Fig. 1.

Fig. 1
figure 1

Search terms

Full size image

The review aimed to identify research studies, reports and articles, including literature reviews, investigating the use of eCDS tools by all health professionals in relation to their impact on workload, such as consultation duration. The focus on ‘health professionals’ in primary care, not just on GPs, was intentional – we sought to identify all relevant contextual research. Therefore, studies concerning any type of health condition, eCDS tool, healthcare context within primary care or health professional were eligible. Both quantitative and qualitative evidence were included. Systematic reviews were included as studies in their own right, and thereafter the references of studies included in those reviews were screened for eligibility and relevance. Eligible and relevant references within a systematic review were then included in addition to those primary studies identified by the original searches. Studies relating specifically to the design or development of eCDS tools, and those focussing on clinical factors associated with specific conditions, were excluded. Protocol articles were excluded if the published results article of the same study were available.

Study selection was guided by: (i) an initial team meeting to discuss inclusion and exclusion criteria, (ii) all abstracts and full text articles were independently reviewed by two reviewers, and (iii) team meetings were held throughout the process to discuss and resolve conflicts of agreement. The following key information was gathered from the included studies: author(s), year of publication, study origin, study aims, type of eCDS tool in study, study population/context, methods, and outcome measures. EF, a health services researcher, classified the key findings into categories, defined as consultation duration-related (‘perceived’ or ‘objectively-measured’), or ‘other’ workload-related. The articles were organised using Covidence review software, then collated in a descriptive format using Microsoft Excel, and reviewed to summarise the key findings.

Results

The database search yielded 5694 publications (4007 after removal of duplicates, Fig. 2). After screening titles and abstracts, 211 publications were selected for full-text screening. Of these, 120 were excluded for not meeting the inclusion criteria, resulting in 91 publications being included in the scoping review. Four of these articles were systematic reviews; screening of eligibility and relevance of references included in those reviews led to the inclusion of a further four articles. The total 95 included articles referred to 87 research studies.

Fig. 2
figure 2

Summary of the screening process

Full size image

Description of included articles

All studies were conducted in high-income countries, with the exception of one from Sub-Saharan Africa. A third of the articles from the studies originated in the USA (36), with UK and Australian articles comprising another third (21 and 11 respectively). A further 18 publications originated in Canada and mainland Europe, with the remaining studies conducted in New Zealand (2), South Africa (2) and Malaysia (1). For most articles workload was not the main focus, with only 16 examining it either as a main focus or as one of the aims.

The most common clinical areas of focus among the eCDS tools studied were cancer risk assessment (15 articles), cardiovascular disease (11), and prescribing for various conditions (10). Other common clinical areas included: blood-borne viruses (3 articles), and various other long-term conditions (14 articles, including those on diabetes, chronic kidney disease, asthma, Chronic Obstructive Pulmonary Disease, and hypertension). Smaller numbers focussed on tools for other conditions including: transient ischaemic attack and stroke, abdominal aortic aneurysm, respiratory infections, psychiatric disorders, skin conditions, hearing loss, and familial conditions (one or two publications on each). Some tools were also designed to support general delivery of care across a range of domains such as maternal and child health, occupational health, behavioural health, and geriatric home care.

A third of articles (31) utilised purely qualitative methods, almost all of which included interviews and/or focus groups with health professionals. One exception reported conversation analysis of audio- and video-recorded consultations and another study reported observations of consultations. Twenty-eight articles reported quantitative methods; 23 involved a survey of health professionals and/or analysis of EMR data or usage data from the investigated tool. The other quantitative articles included three randomised controlled trials and two observational studies. The remaining 28 articles utilised mixed methodologies. The majority of these involved either a survey of health professionals plus qualitative interviews/focus groups (n = 12) or an analysis of EMR/tool usage data in addition to qualitative interviews, focus groups and/or observations (n = 15). Four further articles were systematic reviews, two involving qualitative synthesis and one being a mixed-methods narrative review. All included articles are summarised in the data extraction table (Table 1).

Table 1 Data extraction table
Full size table

Workload-related findings

The scoping review had the broad aim of identifying evidence regarding impacts on workload and workflow; evidence most frequently reported these issues in terms of time and consultation durations. Findings from articles relating to perceived and objectively-measured impacts on either the time spent interacting with an eCDS tool or on whole consultation durations are summarised first (also in Table 2). Findings from articles that reported other workload-relevant results are summarised after.

Table 2 Summary of key findings from qualitative and quantitative evidence
Full size table

Perceived impacts on consultation duration

Seventy-two articles described impacts on consultation duration. These were gathered from qualitative interviews or focus groups with health professionals, often with the aim of identifying barriers and facilitators to implementing eCDS tools in practice. In spite of the wide range of contexts and functionalities of eCDS tools encompassed within this review, the majority of articles indicated that using an eCDS tool was thought to be associated with an increase in consultation duration (n = 36). Some showed a mix of views among health professionals (n = 20). Six articles reported an overall impression that an eCDS tool reduced or ‘saved’ time within the consultation. The remaining articles either indicated no perceived impact on consultation duration (n = 4) or made no explicit conclusion (n = 7).

Perceived increase in consultation duration

Among the 36 articles that indicated a perceived increase in consultation duration, the most commonly highlighted concerns related to existing time pressures and lack of time during a consultation for clinicians to interact with eCDS tools and/or to carry out resultant recommended actions [16, 27, 28, 30, 36, 41, 47, 76, 93, 96, 97, 100, 102, 110, 111, 115, 123]. A prevalent view was that workload was ‘already heavy’ and that using eCDS tools would inevitably add burden [31, 44, 49, 60, 70, 78, 89, 102, 111, 115, 123]. In the case of one tool to support delivery of preventive care through review of patients’ lifestyle factors, the sense of lack of time for preventive care in general drove the view towards the tool increasing consultation duration [31]. Hirsch et al. (2012), however, highlighted that even though the majority of physicians in their study subjectively appraised consultation duration as being extended (85%), there were more of these physicians who felt that the time extension was ‘acceptable’ than those who judged it to be ‘unacceptable’ [52].

The usual flow of tasks to complete during a consultation (often referred to as ‘workflow’ [29, 33, 35, 39, 40, 42, 47, 58, 66, 67, 72, 74, 75, 84, 86, 91]) was commonly expected to be disrupted by eCDS tools, causing an increase in consultation duration [30, 34, 35, 47, 83, 93, 102, 104]. Specific time-consuming functions of tools, such as reading text, additional data-entry and using tools which were stand-alone from the EMR [16, 34, 70, 92, 102, 107, 117, 123], as well as perceptions of poor- or slow-functioning software [35, 37, 60, 104] were also highlighted. A potential for negative impact of eCDS tools on the trajectory of the conversation with patients was expressed by some health professionals. Some expressed concerns that introducing unexpected discussion, such as addressing the risk of cancer, would overtake the allotted consultation time and cause clinics to run late [15, 33, 53, 92].

Among these 36 articles, a wide range of eCDS tools with varying features and functionality were described (some overlapping). Thirteen involved tools which could interrupt the consultation, by presenting an on-screen alert containing risk or safety information, triggered by opening the EMR or by inputting diagnosis or prescription details [27, 28, 36, 53, 59, 78, 83, 89, 92, 93, 100, 104, 117, 131]. In addition, ten of these articles specifically highlight the issue of the tool directing the clinician’s attention towards a condition or matter that was not the reason for the encounter [27, 28, 33, 36, 53, 60, 63, 83, 100, 102, 104]. This was seen as necessitating additional time and/or workload, as a result of requiring prolonged discussion with the patient, serving as a distraction, and adding more tasks to already busy consultations. An eCDS tool flagging an issue that did not match the reason for the encounter could be unhelpful if seen as an ‘unwelcome intrusion’ [105], or if undermining a clinician’s professional expertise (particularly if there are doubts regarding the tool’s accuracy [51]) [105]. Such perceptions would be barriers to using or responding to such tools [51, 53, 60, 104]. Arranging a follow-up consultation in order to allow time for additional discussion and tasks was cited as an option for overcoming such barriers [27, 33].

Thirteen articles presented non-interruptive eCDS tools, accessed by a clinician at any time, used to obtain information, decision support or risk calculation, either for individual patients or as an audit tool used across the practice population [15, 30,31,32,33,34,35, 37, 41, 44, 49, 97, 123]. Eight articles described systems that were standalone from the EMR such as web-based eCDS tools [15, 16, 31, 32, 34, 37, 44, 49].

Perceived decrease in consultation duration

The six articles that reported a perceived decrease in consultation duration suggested explanations which included reduced need for data entry [62, 81, 98], synchronisation with the usual workflow of decision-making [118] and saving time when discussing risk management of specific conditions during the consultation [67, 68]. In terms of the purpose, feature and functionality of the studied eCDS tools, the articles referred mainly to tools that were seen to improve efficiency, four of which featured a tool designed to support clinicians in the management of conditions, rather than on their diagnosis. All of the tools described were either embedded within the EMR system or linked/interacted with the EMR in some way. Two included an interruptive component among other functions [67, 68] and two were entirely user-accessed [62, 81].

No perceived impact on consultation duration

No specific causal factors were suggested by the articles that reported an overall perception of no impact on consultation duration. One study of a cardiovascular risk assessment tool highlighted that consultation duration was perceived to be increased in cases where the GP did not expect the patient’s risk to be high, however the number of such instances was low [63]. A study involving both a survey and interviews with US physicians about a family history data collection tool showed that none reported an adverse impact on their workflow [127]. In terms of the studied eCDS tools’ purpose, features and functionality, the tools described included one with an interruptive component (cardiovascular risk score alert [63]) and two that were non-interruptive: a tool pre-populated by clinic staff that generated an email to the physician one week ahead of a patient’s visit to prioritise Chronic Kidney Disease care [42], and a computerised Family Health History CDS tool which included risk stratification [127].

Objectively-measured impacts on consultation duration

Twenty-six articles reported an objective measure of time. These included: (i) time spent using or interacting with an eCDS tool (ranging from three seconds [73] to between 0.5–13 min [35, 50, 54, 84, 90, 116, 120, 121]) and/or (ii) consultation duration [30, 38, 40, 45, 48, 57, 73, 79, 95, 103, 108, 109, 113, 114, 119, 122], including one which measured time from triage to final disposition decision [128].

Increase in consultation duration

Overall, three articles suggested that consultation duration increased, although none measured consultation duration directly. Two of these articles reported that the time taken to use the eCDS tool was ‘too long’ for a typical ten-minute consultation (four minutes [50] and 13 min [54]), implying that consultation durations would increase as a consequence. One of these two articles highlighted the low rates of usage of the eCDS tool as an important consideration alongside the authors’ conclusion [49]. The third study also did not directly measure time, but instead reported ‘visit type’ as a proxy measure of consultation duration; clinicians more often used the eCDS tool in the longer, annual medical review visits (usually allotted 40 min in that study) than in the shorter, acute care visits [38].

No particular purpose, features, or functionality were shared by the eCDS tools described in these articles. In addition, none were highlighted as potential explanatory factors for the concluded increase in consultation duration.

Decrease in consultation duration

Four articles suggested that consultation duration decreased, noting that the eCDS tools helped clinicians to undertake specific tasks more quickly. Two found that calculating cardiovascular risk scores and making clinical decisions, when assisted by an eCDS tool, was faster [116, 120], and another found a 7.3-min reduction in time within an asthma chart review consultation [121]. The fourth reported consultations to be 3.41 min shorter on average when using an eCDS tool to support diagnosis and management of hypertension [122]. All of the tools featured in these articles supported clinicians in the management of long-term conditions by design, or included an element of management support, as opposed to solely supporting initial risk assessment and/or diagnosis. All bar one described tools that were embedded with the EMR system, with only one of these having an interruptive component [120].

No impact on consultation duration

Nine articles concluded that eCDS tools neither extended nor saved time in consultations. Having compared an intervention and control group or a set of baseline and intervention consultations, five articles reported no significant difference in consultation duration [40, 103, 108, 109, 128]. Lafata et al. (2016) found no association between use of a range of eCDS tools with the consultation duration. [57] The remaining articles reported that their measure of duration when using various eCDS tools (9.05 min [48] and 10 min [95]) was ‘similar’ in length to a standard consultation, concluding that the tools did not prolong consultations [119]. The remaining articles did not make any stated conclusion regarding duration or the conclusion was unclear [79, 84, 90, 114, 130].

A common explanation for lack of impact on consultation duration, or where perceptions of such impacts were mixed, was low rates of tool usage by clinicians in studies. Suggested reasons for non-use included perceived or actual difficulties in the tool’s functionality, slow-functioning software [30, 35, 37, 61], disruption to the usual workflow in a consultation [30, 83, 93] or requiring additional data entry to what would normally be inputted to the EMR, particularly where eCDS tools operated as a standalone system [34, 71].

In terms of purpose, features, and functionality of the tools described by these articles, while one article discussed only a stand-alone system from the EMR [95], the other articles reported either a tool embedded in the EMR system or described a range of both embedded and stand-alone systems. None of the described tools had an interruptive component. Most were guiding or supporting either prescribing tasks or decision making during consultations with a focus on patient management.

Conflict between perceived and objectively-measured impacts on consultation duration

Seven articles reported both perceived and objectively-measured impacts on consultation duration of using eCDS tools. Two found that both their perceived and objective measures suggested increased duration [50, 114]. However, five indicated a conflict between the perceived and objectively-measured impacts [30, 35, 45, 73, 108]. The common perception was that consultation duration was (or would be) increased, but there was actually no measurable difference in duration found. All of the tools described by these five articles were embedded with the EMR system, and did not include an interruptive alert feature or pertain to conditions or tasks likely to be irrelevant to the consultation.

Trafton et al. (2010) described physicians’ perceptions that eCDS for prescribing opioid therapy was ‘too time-consuming’, with insufficient time available during a 15-min consultation to use it [73]. However, the measured time spent using the tool ranged from 3 s to 10 min, and the study concluded that clinicians had ‘a reasonable amount of time’ to use the system. Curry & Reed (2011) reported that physicians felt the time taken for an eCDS system to interact with the EMR was ‘too slow’ despite the captured duration for this interaction being less than one second, although it is unclear whether this reflects physicians’ views of the overall interaction time rather than data processing time specifically [35]. Bauer et al. (2013) reported that although primary care clinic staff felt that a paediatric visit eCDS system slowed down clinics, an “informal” time study did not show any significant delays [30].

Porat et al. (2017) reported that 13 GPs (38%) felt their consultations took longer when using an eCDS system. They felt that inputting free text into the EMR instead was faster, and these same GPs did indeed have longer consultations when using the tool (an average of 15.45 min compared with their baseline 13.53 min average consultations). However, this was the case only for the GPs who expressed concern about time, and not for the GP sample as a whole where no significant difference in consultation duration was observed.

Further, a study by Gregory et al. (2017) found that the perception of physicians regarding the time available to manage eCDS alerts (termed ‘subjective workload’) was not correlated with actual hours spent managing alerts based on physicians’ self-report (‘objective workload’) [46]. When the authors examined whether these ‘subjective’ or ‘objective’ workload measures predicted physician burnout, only the ‘subjective’ measure was predictive. This suggests that the perception of eCDS alert burden in the context of existing high workload is more problematic than the measure of actual time spent managing alerts.

Methods utilised to measure consultation duration

A range of methods was utilised to measure objectively consultation duration or the time spent using an eCDS tool. In five articles, clinicians provided a self-report of time spent, using either a paper or electronic case report form [45, 95, 114, 121, 122]. A member of the research team manually timed the duration of study consultations or scenarios in four articles. [40, 54, 57, 73] Five articles reported time data captured electronically from log files within the eCDS tool itself, including clinician time spent using particular elements of the tool or completing certain activities [35, 50, 73, 84, 90]. Three articles described using specialist software, operating in the background, designed to record users’ interactions with the eCDS tool during consultations [116, 119, 120]. Specific software included Morae Recorder and Camtasia, both TechSmith Corporation products. Three studies used video- or audio-recordings to capture consultation durations in addition to other elements of the consultation they aimed to observe [48, 80, 113]. Two articles that referred to the same core UK study, described capturing duration data from the practice IT system (Vision), based on the opening and closing of the EMR [108, 109]. One USA study estimated consultation duration based on the reasons patients were attending – either for a ‘shorter’ visit, such as for acute care or follow-up, or for a ‘longer’ visit, such as for a general medical examination [38], and two articles provided insufficient details of the methods used [30, 117].

Other workload-related findings

Twenty-seven articles included additional workload-related findings. Twenty-three of these reported the impact on ‘workflow’, regarding how eCDS tools altered the usual order in which patient-related tasks were carried out [33, 35, 39, 40, 47, 58, 66, 74, 75, 83, 84, 87, 91, 93, 94, 100, 102,103,104, 111, 113, 118, 119, 127]. Five referred to the impact of using eCDS tools on the trajectory of dialogue with patients, to the extent that follow-up appointments were arranged to avoid consultations running late [15, 39, 75, 94, 100]. One of these mentioned clinicians’ concerns about ‘taking time away’ from other waiting patients, expressed as a barrier to the implementation of eCDS systems [26]. Many of the tools in these articles were clearly described as having an interruptive alert component [33, 58, 83, 84, 86, 91, 93, 100, 104, 111, 118, 119].

Some articles (n = 10) mentioned ‘alert fatigue’ indicating that eCDS tools designed to support health professionals can increase the number of on-screen alerts, leading to a high chance of them being missed or ignored [15, 36, 40, 42, 51, 74, 100, 105, 111, 117]. None of these articles reported a decrease in consultation duration.

Cognitive workload was referred to in three articles. Qualitative interview data suggested that clinicians felt an eCDS tool for prescribing tuberculosis preventive therapy decreased their cognitive workload during consultations. [98] This was perceived as advantageous as it reduced the amount of time spent documenting medications and their contraindications. However, in two articles, eCDS tools were noted to increase cognitive workload. A systematic review that examined factors influencing the appropriateness of interruptive alerts found such alerts increased cognitive weariness, and that an ‘overload’ of alerts increased mental workload [117]. A study of an eCDS tool for assessing cardiovascular risk also highlighted clinicians’ concerns about the cognitive burden of changing to a new way of calculating risk compared with the conventional method they had used until that point [124].

One study reported workload expressed as the number of follow-up consultations needed. This study examined eCDS tools for patients with upper respiratory tract infections, and found no significant difference in the proportion of follow-ups needed between the intervention and control arms [82].

Discussion

This scoping review identified 95 articles that examined the use of eCDS tools by health professionals in primary care and reported findings that included impacts on workload and workflow. While the scoping review had the broad aim of identifying evidence regarding these issues, they were most frequently reported in terms of time and consultation durations.. A large proportion of the research was qualitative and exploratory in nature. The majority of articles reported health professionals’ subjective perceptions of time spent using eCDS tools and/or the impact on consultation duration and there was a smaller evidence base which objectively-measured impact of using eCDS tools on workload, specifically in relation to consultation duration and the flow of consulting sessions.

The reviewed literature reflected that although a small number of articles suggested that using certain types of eCDS tool decreased consultation duration, a strong perception exists among health professionals that consultation duration was increased when eCDS tools were used. It is worth noting that eCDS tools designed to support management of health conditions and tools supporting diagnosis and associated risk assessment may have different impacts on consultation workload and duration; the small number of reviewed articles that indicated a time saving mostly featured tools designed to support patient management. It is also notable that many of the articles describing tools that introduced a condition or issue that was outside of the patient’s or clinician’s agenda for the consultation, frequently reported clinicians’ perceptions that workload and/or consultation duration increased.

The perception that consultation duration was increased is not necessarily backed by studies that objectively measured actual durations of consultations. Although many of the quantitative articles reported the time taken to use various eCDS tools within consultations, fewer studies captured the duration of entire consultations and/or made a comparison between an intervention and non-intervention group. Interestingly, those that did showed no significant difference in consultation duration when using eCDS tools compared with not using them [40, 103, 108, 109, 128]. Various methods were used to capture consultation durations, with no one method that seemed most practical or accurate. For instance, while the manual (stopwatch) timing of consultations by a researcher [54, 73] might arguably capture consultation durations more accurately than clinicians’ self-report, this method could be seen as intrusive to the consultation. Capturing time stamp data in an automated way, for example from EMR systems [108, 109], might address this issue and provide a practical solution, but errors may be introduced by this method if patient records are left open after the end of a consultation, or some part of the consultation takes place when records are closed.

The reviewed literature highlighted that low usage rates of eCDS tools by clinicians in studies (for varying reasons) may be responsible for a lack of observable impact on workload or consultation duration. Conversely, a tool that fits easily within the usual workflow of a consultation might explain the lack of increased duration. The experience of ‘alert fatigue’ was frequently mentioned; a large number of different on-screen alerts during consultations can desensitise clinicians to alerts, and an alert generated by a new tool may be missed or ignored [27, 28, 50]. Ignoring an alert or not utilising an eCDS tool might indicate clinician’s preference to rely on their own clinical judgment, or doubts as to an alert’s accuracy or relevance, which is particularly highlighted within the alert fatigue literature [36, 107, 132,133,134]. It might equally be the case that a clinician did indeed utilise or respond to the eCDS tool, but arranged a follow-up appointment to allow for more time to discuss the clinical issues raised [26, 28, 33], thereby not impacting the duration of the current consultation. Whether use of eCDS tools had an impact on the duration of the healthcare ‘episode’ as a whole (i.e. the index consultation plus the number and duration of any subsequent follow-up consultations) was unclear from the reviewed articles.

Reviewing articles that included both a subjective measure of health professionals’ perceptions and an objective measure of consultation duration provided an opportunity to observe if the perceptions were borne out in reality. These articles most commonly reported that health professionals felt consultations were (or would be) prolonged by using eCDS tools, but objective measures did not consistently back this up [30, 35, 73]. However, the evidence base for actual consultation durations associated with using eCDS tools remains a lot smaller than that of the perceived impacts on consultation durations. One should note that the perception or expectation of health professionals in relation to consultation workload and duration is very important. Firstly, perceptions and expectations may well determine how often eCDS tools are used. Secondly, ‘subjective’ workload (clinicians’ reported amount of time available to manage alerts), rather than ‘objective’ workload (the number of hours actually spent managing alerts), has been found to be predictive of physician burnout [45]. It is worth also noting, however, that a perception or an objective measure of increased workload or duration may not always be viewed negatively; for example, it may not matter how much consultation duration is increased (if it is) if diagnosis and/or management is improved [52].

Strengths and limitations

This study benefits from undertaking a comprehensive literature review addressing a key area of primary care service provision, namely the interface between technologically enhanced service provision in the form of eCDS, and clinical workload and workflow. We successfully identified and summarised a large number of articles published from a variety of international settings.

The review may have been affected by the inclusion of names of specific eCDS tools within the search terms. This reflects research team members’ awareness of existing systems in UK primary care; tools not known to the authors may have been missed from the review. We identified a number of studies through systematic reviews that were not found through our initial searches, this suggests that our initial searches may have missed some relevant work. Inclusion of articles published in the last ten years, since 2009, may also have omitted potentially-relevant research on eCDS since its inception in the 1960’s, however we aimed to identify evidence from research articles based in modern-day primary care settings In addition, although the vast majority of international scientific literature is currently published in English, our exclusion of foreign language articles may have prevented fuller coverage of non-UK primary care contexts with different standards of consultation lengths, workload or workforce challenges, and policy expectations. The review also included a large number of qualitative articles, but time and resource issues prevented a full qualitative synthesis of these articles.

The two independent reviewers who undertook screening were not always the same two reviewers due to resource constraints, however EF undertook all stages of the review and had regular discussions with the small group of four ‘second’ reviewers. Only EF undertook data extraction and so details from included articles may have been affected by selection bias.

Conclusion

This scoping review identified over 90 articles that explored the use of eCDS tools in primary care by health professionals in relation to aspects of workload, including consultation duration. Whilst the qualitative literature showed a strong perception among health professionals that eCDS tools increased workload and consultation duration, a smaller number of studies captured quantitative measures, which neither disputed nor supported this view.

eCDS tools designed to support GPs will continue to be introduced within primary care with the aim of assisting clinicians to diagnose and manage patients effectively. Despite the absence of strong objective evidence that using eCDS tools necessarily leads to increased (or decreased) consultation durations, the perceptions of additional time being taken within consultations, additional workload being generated, and workflow being disrupted, are barriers to implementation and routine use of eCDS tools, irrespective of their potential benefit in the diagnosis or management of patients.

Further quantitative evidence measuring actual consultation duration and GP workload is needed to confirm whether the reported concerns are justifiable, particularly in the time-constrained setting of primary care. Future efforts to implement potentially valuable eCDS tools need to take account of the context of increasing GP workload, workforce shortages and associated pressures, and the ongoing challenges generated in the wake of COVID-19.