Understanding “Alert Fatigue” in Primary Care: Qualitative Systematic Review of General Practitioners Attitudes and Experiences of Clinical Alerts, Prompts, and Reminders

Introduction

Primary care is facing unprecedented pressure across the globe, caught between growing patient demand, issues of workforce recruitment and retention, and increasingly complex options for treatment [1,2]. To improve the consistency and quality of care in this dynamic and highly pressured environment, health systems worldwide have introduced guidelines for care informed by the latest clinical evidence [3,4]. These guidelines are promoted and supported in high-income countries through a range of clinical reminders (CRs) that are presented to general practitioners (GPs) via their clinical management systems and consist of “alerts” that remind them of the consequences of making a certain decision, and “prompts” that remind the user to perform a task that promotes desirable clinical behaviors [5-10]. These alerts can relate to potential diagnoses, drug-drug interactions, and allergies, and prompts linked to patient-specific needs, further information, and recommendations for high-quality care (see Figure 1; see also [11]).

‎

The growing sophistication of medical knowledge and the electronic systems that manage patient data and inform clinical decision-making have led to a steady increase in the activation of CRs at the point of care and the cognitive load placed on GPs [12-15]. Not all CRs are acted upon, and some may justifiably be considered irrelevant or inapplicable [16,17], but increasingly GPs are disregarding pertinent or urgent CRs [18-20], a chronic negligence commonly referred to as “alert fatigue” [21].

Despite the implications of alert fatigue for the safety and quality of care, GP burnout, and the medico-legal consequences for individuals and their organizations, the proliferation of CRs in primary care is reliant on evidence limited to their effects on a specific condition [22,23]. What remains underexplored is how their use might be optimized in the context of the cumulative effect of multiple CRs on a single health care provider [19,20,22,24]. There is then a need to better understand GP attitudes toward the deployment, design, and overall effectiveness of CRs, as well as broader implications for their impact on patient-clinician communication [22,25]. The systematic review presented here aims to address this by synthesizing current qualitative research on GPs’ attitudes toward CRs describing the findings within the Non-adoption, Abandonment, Scale-up, Spread, and Sustainability (NASSS) framework [26]. This framework was designed to study unfolding digital health technology programs and so provides the opportunity for offering structured insight into the future design, use, and positioning of CRs in primary care.

Methods

Applying the NASSS Framework to Digital Health Implementation

The review used best practices in systematic review methodology and directed content analysis [27,28]. We describe the results through the lens of the NASSS framework developed to support, guide, and monitor the implementation of digital technologies in health and social care [26,29,30]. The framework uses 7 domains to describe various aspects of the technology, end users, organizational structure, and broader policy-driven context [26]. The domains are defined alongside their potential impact on implementation in Table 1.

Search Strategy

The databases searched were the Health Technology Assessment Database, MEDLINE, MEDLINE in Process, Embase, CINAHL, Conference Proceedings Citation Index, PsycINFO, and OpenGrey, from January 1, 1960, to December 31, 2023, a date range chosen to reflect the introduction of electronic or digitally supported health care. The search was conducted by the author IG in March 2024 and no language or location limits were applied. Medical Subject Headings (MeSH) terms and free-text words were used against the Sample, Phenomenon of Interest, Design, Evaluation, Research type framework [53] to collect relevant studies from the perspective of GPs in a primary care setting (see Multimedia Appendix 1). The search strategy is described in the PROSPERO registration and the previously published protocol [54]. The EndNote (Clarivate Plc) bibliographic software was used to organize all retrieved studies and remove any duplicates.

Eligibility Criteria

Studies included were primary qualitative research conducted in primary care settings and published in English. The research had to have a focus on general or family practice and be qualitative in nature because we were specifically interested in GPs’ experience with and perspectives of CRs (see Multimedia Appendix 2 for detailed eligibility criteria).

Study Selection and Data Extraction

We used the recommended full dual review approach, that is, 2 independent reviewers (IG and AP) screened the title, abstract, and then the full texts, as this has been shown to increase the number of eligible papers identified and more broadly minimize error in the selection process. The data were managed using an Excel spreadsheet (Microsoft Corporation), with any disagreements on the inclusion of a study consensually resolved through discussion with a third reviewer (IL). A data extraction form was designed to summarize study characteristics. The data extracted included information on study outcomes, study design, participants, and key findings.

Quality Appraisal

The quality and bias of included papers were determined by the Critical Appraisal Skills Programme checklist for qualitative studies and the Mixed Methods Appraisal Tool for mixed method studies [55]. Although the Mixed Methods Appraisal Tool cautions against the approach of labeling studies at low, medium, and high risk of bias, we used a scoring system to determine this labeling, for ease of interpretation between the 2 different tools. The quality appraisal was undertaken independently by 2 authors (IL and IG) who between them agreed on the final value.

Data Synthesis

The data were coded and managed via NVivo12 (Lumivero, LLC). We used a directed content analysis to populate the NASSS framework, based on the unconstrained matrix approach. This allowed the development and inclusion of constructs directly relevant to emergent themes within the established framework [56,57]. The analysis was conducted by IG and IL who both independently analyzed a sample of the same 3 papers. They then met to discuss their analysis and any discrepancies in the allocation of data before IL analyzed the remainder of the studies, with the final analysis and data allocation consensually agreed upon by all authors. In completing the analysis, we merged the domains of “Wider system” and “Embedding over time” due to the concurrence that arose in the data.

Results

Study Characteristics

The screening and study selection are outlined in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram (Figure 2; also see Multimedia Appendix 3). The study selection and data extraction templates can be found in Multimedia Appendix 4. Full-text studies were excluded because of duplication, or lack of detail on the care provider, setting, or digital tool (see Multimedia Appendix 5 for a detailed description). A total of 9 studies were included, 5 from the United Kingdom [58-62] and 1 each from Australia [63], Belgium [64], the Netherlands [65], and Norway [66]. Seven of these were considered high quality [58-62,65,66], 1 medium [64], and 1 low [63]. The papers were published between 2003 and 2021. The key characteristics of included studies and findings are summarized in Table 2. This includes the quality assessment (see Multimedia Appendix 6 for a detailed description). The directed content analysis enabled the mapping of the data within the NASSS framework. Within each domain, we determined a series of emergent constructs relevant to the implementation of CRs in primary care. These findings are summarized in Table 3 and presented below within each domain of the framework alongside exemplar quotes taken from the studies identified.

‎

Condition

The studies identified described how CRs were used to optimize care and manage patient safety in 3 key areas: medication [58,59,62-64,66]; diagnosis, both broadly [64,65] and for specific conditions, namely, atrial fibrillation and stroke [60], and cancer [61]. CRs were also used in the ongoing management of chronically ill patients, providing reminders of reviews or tests [64].

Technology

Challenges in Technological Aspects of CRs: Alerts, Design, and Infrastructure

Seven studies described issues relating to various technological aspects of CRs, specifically the frequency and content of alerts, their design and integration with existing interfaces, and the broader digital infrastructure and capability including the interoperability of data.

Frequency and Content of CRs

Four studies described how the inability to process a large number of alerts was a concern among GPs [59,61,63,64]. As one GP described:

There are so many things (...) so many things popping up, so many things prompting you. You don’t probably respond to all the prompts, because there’s a box here, a box there, a box everywhere, and you don’t see everything (...) It’s such a busy screen you don’t respond to everything...
[61]

The value of presenting accompanying patient-centered information alongside the clinical data of the CR was described in one study. As a GP elaborated:

It [the CR] actually comes up with a nice diagram that it can use with your patients, that’s got smiley faces of different colours to illustrate risk (...) it’s just a different way to help augment that communication with them. For things like that it’s quite helpful because patients are wary of tablets, quite rightly.
[59]

Design and Integration of CRs

Three studies described how pop-up CRs that required the checking of drop-down lists or otherwise had to be actioned before any next step were overly disruptive and actually inhibited their engagement with the CRs content [59,62,63]:

[when they appear]...you can’t go looking in the notes, you can’t input anything else, this is now taking priority, so you either suspend it and then re-open it and suspend it and re-open it.
[59]

Digital Infrastructure

The capacity of practice information technology (IT) systems and hardware was also a consideration, for example, some CRs slowed the operating speed of practice computers [59,60,62,66].

Our problem here is our computers are rubbish. They work really, really slowly (...) it takes a long time to come through so we scroll down an option, a pick list - it can pick the wrong thing! Which is very frustrating...
[62]

Value Proposition

GPs’ Perspectives on CR Quality, Safety, and Educational Impact

GPs described their perspectives on the quality and safety of CRs, their potential to support continuing education and training, and their impact on self-efficacy.

Overall Safety and Quality

In 5 studies GPs reported that CRs were capable of improving patient safety by promoting reflection on their decision [58-61,63]:

...even if you immediately dismiss it, at least that millisecond you’ve thought about it, and I think that is going to be useful at some point, but for how many people I don’t know
[61]

GPs also appreciated that CRs could improve the quality of care by providing useful reminders of tasks or actions [59,60]:

...in a patient with diabetes where there are 17 things to do and I’ve just missed out two or three of them (...) it’s really nice to have “somebody” say “Oh, don’t forget, there’s the urine test, and you haven’t yet sent the patient for the eye test that you thought somebody else was doing, and it looks like nobody’s done after all so maybe get that done?”
[59]

GPs in one study reported how CRs could improve how they practice in the long term by their ability to promote and reinforce appropriate prescribing behaviors.

...and it’s actually educational, because a lot of them you know the warning before it comes up because you’ve seen it so many times so you’re already aware of it. So I mean some of the things I know now are because of it, so there is an educational element to it [...]
[58]

Self-Efficacy

There were conflicting accounts regarding the extent to which CRs enhance GPs’ confidence in practicing safely. In one study, GPs described how the presence of CRs was a source of reassurance, supporting their ability to provide safe care in the increasingly complex environment of modern primary care [61,63].

...work in general practice is getting bigger and bigger, and more work’s going to come to general practice. So, although it [the CR] can cause irritation, the flip side of that is you can relax a little bit more and that you don’t have to remember absolutely everything (...)
[63]

In another study, a GP reflected upon how having to choose a predetermined response outside of their original decision led them to question their ability [59]:

...in order to dismiss it you have to select what action you’ve taken such as, well have you called an ambulance, have you sent them into hospital...you’re then panicking thinking gosh I don’t want, I don’t want to lie, no I haven’t sent them into hospital but is that a judgement, is that them saying because I haven’t done that I’m a rubbish doctor?
[59]

Adopter

Factors Influencing GPs’ Engagement With CRs: Digital Literacy and Consultation Impact

From the perspective of those adopting CRs, GPs described the factors affecting their engagement with CRs, the influence of digital literacy, and the impact of CRs on their consultation.

Engagement and Compliance in CRs

In one study, GPs expressed hesitation in relying on a computer-generated algorithm [59]. Concerns surfaced in 3 studies over the accuracy and relevance of the recommendations associated with CRs, with examples including unrecognized contraindications [58,63,65]:

[drug-drug interactions] were mentioned in printed textbooks but were not picked up by the interaction checking facility of their prescribing software
[63]

One study described a preference for greater transparency between the clinical evidence and the presentation of the CR. As one GP described:

[We’re] inherently distrustful of anything that hasn’t got somebody saying “Oh! I’ve done this trial, and this trial, and we’ve used it!” and I think that’s what we’ve been taught at medical school - to be sceptical, until we’ve got the evidence.
[59]

Digital Literacy and Acceptability

The belief emerged that the new generation of GPs more familiar with digital technology in everyday life would be more accepting of computer-generated reminders [59].

I think to a certain extent there is a bit of a cultural change in the sense that the newer generation of doctors coming through are much more used to computers telling them things in their own lives and so they’re much more accepting of the idea that the computer might give them helpful information.
[59]

GPs in 3 studies described how the reliance on CRs risked undermining the accrual of contextual knowledge of a particular patient [59,61,67].

A lot of the patients who I kind of see regularly with chronic problems, I kind of know if they’re well or not because I know the patient and I just can tell if they’re themselves or if they’re not because you have a constant, you know, you lose that with a computer.
[59]

Related to this was the feeling that CRs were overly prescriptive and one GP felt that the value of their input was diminished relative to the diktats of the CR:

I think it feels frustrating...when the system is so process-driven that there’s no sort of autonomy of the GP to, you know, change it or, kind of, put in what’s relevant.
[59]

Patient-Clinician Relationship

Two studies reported that GPs understood the importance of focusing on patients during consultations, acknowledging that referral to, and interaction with, CRs can be an impediment to patient-centered interaction [61,63]:

I don’t use any separate clinical decision support tool in the consultation because of time problems, losing eye contact with patients
[63]

Organization

Organizational Influences on CR Adoption: Training and System Capacity

In considering organizational influences, evidence emerged of the importance of training and the impact of existing practice capacity and systems.

Training and Purpose

GPs in 5 studies reflected on the importance of training not only to understand the purpose of CRs but also their functionality [58-61,63]. One GP bemoaned the typical lack of training before the introduction of digital tools:

...so, yeah things just happen [. . .] They often will appear before you have any training or knowledge of it whatsoever. Then if you’re lucky maybe there’s a launch event six months after you’ve already started using it.
[58]

Practice Capacity and Workflow

GPs noted how the pressure on their time restricted their ability to interact with CRs that take time to resolve [58,60]. As one GP explained:

We might have six or seven of them [reminders], and once you’ve got (...) worked through about three or four, and then there’s this anticoagulation...you think, well that’s another 5 or 10 minutes of consultation (...) so you leave it.
[60]

In another example, one GP likened the appearance of a CR to a colleague bursting through the door during a consultation:

It really is like an interruption, you know, no GP wants somebody to just burst in with the door opening or the phone ringing. In the same vein no GP really wants a big thing to just pop up on the screen that they didn’t call up.
[59]

Wider System and Embedding Over Time

Enhancing CR Implementation: Systemwide Collaboration and Adaptation Strategies

The potential impact of improved systemwide collaboration was described, and a number of suggestions were made as to how CRs may be adapted to ensure more sustained implementation.

Systemwide Collaboration

It was suggested in 2 studies that more closely integrated data sets across settings would support the quality of the CRs [63,66].

I want a dynamic electronic health communication with the possibility of a written dialog and forwarding missing information.
[66]

The importance of sharing best practices and recommendations for certain CR systems between peers was also described:

I think there is a degree of sharing these kind of protocols and algorithms amongst GP practises and we certainly, you know, GPs talk to each other and say “oh yeah we’ve got this thing that alerts for that particular problem”, “oh, have you? Right great can we, can we bring it in?”
[59]

Adaptation Over Time

Three studies described how GPs would have liked greater engagement with the software developers in the design of the CRs [58,63,65]. Suggestions for ways in which CRs might be improved included the acquisition of patient-specific alerts, better integration into practice workflows, and customizability informed by GP feedback [59-61,63]. In terms of their design and content, there were also suggestions for the inclusion of additional details on dosage or color-grading CRs according to severity [63].

Discussion

General Findings

A total of 9 studies published since 2003 were identified, reporting research conducted in Europe, predominantly the United Kingdom, and Australia, 7 of which were judged to be of high quality. The findings from these studies were effectively categorized within the domains of the NASSS framework. Regarding Condition and Value Proposition, the use of CRs in diagnosis, prescribing, and the ongoing management of chronic conditions was seen as an effective means of maintaining or improving safe and high-quality care, albeit with contrasting effects on clinician self-efficacy. In terms of Technology attributes, the efficacy of CRs was found to be directly linked to their frequency, the presentation and accuracy of their content, and the speed and reliability of IT infrastructure and hardware. With respect to Adopter attitudes, GPs raised concerns about the accuracy of CRs and the potential loss of benefits derived from their clinical experience and consideration of individual patient contexts, in favor of algorithm-based decisions. From an Organization perspective, the importance of training and the ability of CRs to integrate seamlessly with existing workflows were emphasized. Finally, in the context of the Wider System, several suggestions were made to improve the usability of CRs, including the sharing of best practices and involving GPs earlier in the design process to ensure successful embedding over time.

Specific Findings and Comparison With Existing Literature

Condition/Value Proposition

Managing the medications of a growing number of patients with multiple long-term conditions is a major contributor to GP workload in the United Kingdom and other countries [68,69]. The studies we identified suggest that CRs can play a positive role in supporting this management [58,59,62-64,66], and it is widely recognized that CRs present a valuable opportunity to reduce adverse drug reactions [70].

The potential of CRs to support diagnosis was highlighted in 4 of the identified studies [60,61,64,65]. This ability has been previously acknowledged as a significant advantage in primary care, where diagnostic errors are a critical concern, particularly for diseases with low incidence but high health risk [71-73].

Primary care is expected to take on an increasing role in managing chronic conditions across various countries and health systems, including the United Kingdom [74,75]. However, this capacity is often hindered by limited consultation times, high patient volumes, and inadequate communication with specialists [75,76]. In such contexts, CRs that draw on patient-specific data to remind GPs to test, review, or conduct periodic interventions appear highly valuable, though only 1 study explicitly examined this functionality [64].

The ability of CRs to prompt reflection on clinical decision-making was viewed positively in 5 of the studies [58-61,63]. The benefits of reflective practice on clinical decisions are well established, particularly when the feedback on decisions, as with CRs, is immediate [77].

Technology

The studies we identified confirmed that GPs felt overwhelmed by the number, complexity, and often poor functionality of CRs [59,61,63,64]. This cognitive overload caused by multiple appearances of CRs reflects previous research suggesting it can hinder clinician interaction with clinical decision support systems [18-20,78]. The cognitive burden associated with CRs also stems from the complexity of their content [59,62,63]. Therefore, CRs should be designed to accurately convey risk and relevant information in a clear, easily interpretable format that adheres to best practices in digital interface design [79-82]. Additionally, they should be better integrated into existing software and workflows [83,84].

Artificial intelligence has begun to be utilized in clinical decision support systems for individual conditions, and it is reasonable to assume it will play a role in refining the specificity of CRs in primary care [85]. However, efforts to minimize the negative impact of CRs, such as improving specificity [18] or allowing clinicians the autonomy to disable certain CRs [83], must remain mindful of patient safety implications [86].

The ability of GPs to respond promptly to CRs was also influenced by the speed and reliability of practice IT infrastructure and hardware [59,60,62,66]. Ensuring that existing IT systems can handle the growing prevalence and complexity of digital health care is widely acknowledged as a critical need in both high- and low- and middle-income countries. However, this challenge has yet to be universally addressed [87,88].

Adopter

GP concerns about CRs that enforce process-driven decisions at the expense of context and experience were described in 4 of the studies we identified [58,59,63,65]. Notably, lasting cynicism expressed by GPs who had experienced prior failures with CRs (described in 2 of the identified studies [59,63]) has been previously recognized in primary care, where digital interventions have failed to meet expectations [89].

A framework for technology developers has emerged that aims to enhance clinician confidence in clinical decision support systems by adhering to a predetermined set of criteria, focusing on understanding and sustaining trust among end users [90,91]. Evidence from one study we identified suggested that resistance to digital interventions may be lower among the generation of clinicians who grew up alongside the proliferation of digital technologies [59]. It is hoped that trust in CRs can be fostered through greater and earlier engagement with clinician end users by software developers and senior decision makers [92-94]. This aligns with long-standing recommendations for increased user involvement in the development of digital health tools [67,89].

The negative disruption of patient consultations by CRs was also highlighted [61,63]. It is already known that technology can interrupt the flow of conversation, reduce nonverbal communication cues, and compromise the overall quality of patient interactions [95,96]. Concerns expressed by GP participants in the identified studies reflect existing evidence that clinicians highly value uninterrupted in-person consultations, as they build patient trust and foster a constructive provider-patient relationship [97,98].

Organization

The lack of training emerged as a barrier to engagement and compliance with CRs in the studies we identified [58-61,63] and has previously been highlighted as central to their effective use, including sharing with GPs the processes behind their generation [99,100]. However, targeted training to enhance engagement with CRs has yet to be systematically implemented in the high-income countries where they are most prevalent [101].

The pressure on GPs to address increasingly complex care needs and rising patient expectations within shortening consultation times is widely acknowledged [102]. GPs in the identified studies reported that managing multiple CRs was unfeasible within the constraints of limited consultation times [58,60]. Some studies suggest that specific CRs, such as those supporting preventive care, could be delivered to patients before the consultation to alleviate the time pressure on GPs [103].

Wider System and Embedding Over Time

The need for systemwide coordination and collaboration in introducing CRs into primary care, as described in the review’s findings [59,63,66], has also been recognized in other health care settings [104-106]. Similarly, the calls for earlier clinician engagement in the design and development of CRs identified in this review [58,63,65] have been previously acknowledged [107-109], alongside the input of clinical informaticists [110]. While there are recognized challenges in engaging time-pressured GPs during the early stages of software development, sustainable adoption of health technology requires understanding its social and organizational contexts. This can only be achieved by involving end users throughout its design and development [111,112].

Strengths and Limitations

The recognized systematic review methodology was applied to identify relevant papers [27,113], including the dual-review process [114]. The comprehensive NASSS framework successfully accommodated all of the data identified [26] and its use demonstrated both the transparency and rigor of our approach but also allowed us to structure and contextualize our data with existing research to avoid distorting our conclusions [28]. We acknowledge that there were a number of potential limitations. First, all the studies identified originated in high-income countries and were presented in English; however, we would suggest that the lessons learned have the potential for broader applicability for digital interventions in health systems of low- and middle-income countries that in reality share many of the same, if accentuated, barriers to technology-enabled care [111,112]. Second, it is possible that the use of an a priori framework might limit the analysis, with some data unable to be allocated within its confines; however, taking the unconstrained matrix approach suggested by Elo and Kyngäs [56] allowed the development and inclusion of emergent themes within the established framework. This meant we could maintain alignment with our established objectives and provide a systematic description of novel findings.

Conclusions

If utilized correctly CRs can enhance patient safety, workflow management, preventative care, and reduce cognitive overload. Based on the findings of the review several recommendations can be made for the future design and implementation of CRs (see Table 4). These include improved graphical designs better suited to the purpose of the CR, more nuanced and patient-specific content, greater clinician autonomy regards if and when CRs are seen, and earlier and more consistent engagement with the workforce in their design and development.