Workload, Burnout and Safety in Clinical Teams: Four Reviews Read Against One Trial

Four of the sources reviewed here examine how health care organizations change the conditions of clinical work. They cover interventions for clinician burnout ^[2], artificial intelligence (AI) in nursing ^[3], quality and safety recommendations for nursing practice ^[4], and the World Health Organization Surgical Safety Checklist (WHO SSC) ^[5]. Read together, they repeatedly return to workload, but the strength and reporting of their underlying evidence differ ^[2][3][4][5]. A fifth source, a phase 3 randomized trial of the cardiac drug aficamten, addresses an unrelated clinical question. It is used here only as a reference point for what a prespecified randomized design reports ^[1].

The five sources and their designs

The designs differ more than the topics do, and those differences affect how much weight each finding can carry ^{[1][2][3][4][5]}.

• Burnout and well-being interventions. This is a systematic review, meaning a structured search and appraisal of published studies against stated eligibility criteria ^[2]. It covered studies since 2015 of physicians, nurses and allied healthcare professionals that reported quantifiable preintervention and postintervention outcomes using validated well-being measures ^[2]. The authors searched Medline, Embase, Emcare, CINAHL, PsycInfo and Google Scholar in May to October 2022, screened 1663 articles and included 33 ^[2]. Two researchers assessed quality using the Medical Education Research Study Quality Instrument (MERSQI) ^[2]. A meta-analysis was not possible because study designs and outcomes varied ^[2].
• AI in nursing. This is an integrative review that used thematic synthesis and is indexed as a systematic review ^[3]. It included 25 studies and was guided by PRISMA 2020 and the SPIDER framework ^[3]. Study quality was assessed with the Mixed Methods Appraisal Tool (MMAT), and risk of bias with ROBINS-I ^[3]. The authors judged 21 of the 25 studies to be at moderate risk of bias ^[3].
• Quality and safety in nursing. A triservice military nursing team searched five databases and included 249 articles ^[4]. From these the team derived 94 practice recommendations and eight focus areas ^[4]. The work was intended to inform the quality and safety component of a nursing professional practice model for military hospitals ^[4].
• WHO Surgical Safety Checklist. This is a systematic review combining qualitative thematic analysis with meta-meta-analysis, meaning analysis of findings already reported in existing systematic reviews ^[5]. It drew on 20 systematic reviews ^[5]. Those reviews contained 24 unique observational cohort studies reporting pre-post data, that is, outcomes before and after checklist introduction, on 18 clinical outcomes ^[5].
• Aficamten. This was a phase 3, double-blind, randomized controlled trial in 282 adults with symptomatic obstructive hypertrophic cardiomyopathy (HCM), funded by Cytokinetics ^[1]. The authors describe elevated intracardiac pressure from left ventricular outflow tract obstruction as one of the major determinants of exercise intolerance in obstructive HCM ^[1]. Patients were assigned to aficamten (142) or placebo (140) for 24 weeks ^[1].

Where the sources agree: workload as a shared concern

Workload appears in all four workforce and safety sources, but each treats it in a different role ^[2][3][4][5]. In the burnout review, workload reduction is one of the organisational interventions, alongside job crafting and peer networks ^[2].

In the AI review, workload is a target for automation ^[3]. AI systems that automated routine follow-up tasks and generated predictive workload models were reported to free nurses from repetitive, non-clinical duties and to inform staffing decisions ^[3]. These efficiencies were associated with reductions in burnout and improved workplace morale ^[3]. The review reports this as an association drawn from studies mostly judged at moderate risk of bias, not as a demonstrated effect ^[3].

In the nursing quality and safety review, staffing, workload and work environment form one of the eight focus areas ^[4]. The other seven are communication, adverse events, leadership, patient experience, quality improvement, safety culture and committees, and technology and the electronic health record ^[4].

In the checklist meta-meta-analysis, workload appears as an aspect of using the tool rather than as its target ^[5]. One of six themes concerned the efficiency and workload involved in using the checklist ^[5]. The authors note that this theme is associated with checklist use but was not a focal area when the checklist was developed ^[5].

Taken together, the sources treat workload in three ways: as a target for intervention ^[2][3], as a domain for practice recommendations ^[4], and as a theme in the experience of using a safety tool ^[5]. None of the abstracts reports a common workload measure, so the agreement is thematic rather than quantitative ^[2][3][4][5].

Where the sources pull in different directions

Breadth of benefit. The burnout review reports effective outcomes in 29 of 33 studies ^[2]. Those outcomes spanned improvements in well-being, work engagement, quality of life and resilience, and reductions in burnout, perceived stress, anxiety and depression ^[2]. The checklist meta-meta-analysis reaches a narrower conclusion ^[5]. It finds that the WHO SSC positively affects what it was explicitly designed to address and does not positively affect what it was not designed for ^[5]. The two findings concern different interventions and populations, so they do not contradict each other directly; one source reports benefit across many outcomes, while the other reports benefit confined to designed aims ^[2][5].

Clinical outcomes within the checklist evidence. In the pooled pre-post studies, nine outcomes favored use of the WHO SSC: mortality, morbidity, surgical site infection, pneumonia, unplanned return to the operating room, urinary tract infection, blood loss requiring transfusion, unplanned intubation and sepsis ^[5]. Deep vein thrombosis was the only postoperative outcome assessed that did not favor the checklist ^[5]. Because the underlying studies were observational, these results indicate that checklist use was associated with better outcomes; they do not establish that it caused them ^[5].

Benefits and costs of AI within one review. The AI review itself contains opposing findings ^[3]. AI-powered simulations were linked to deeper student engagement, better case-management performance and higher satisfaction scores ^[3]. Learners also reported greater cognitive load and heightened stress during these activities ^[3]. AI alert algorithms and wearable sensors were reported to detect signs of patient deterioration and fever earlier than conventional methods ^[3]. Critical-care staff, however, stressed that automated insights must be balanced with professional judgment to avoid overreliance ^[3]. Nurses also raised ethical concerns about patient data privacy, algorithmic bias and the preservation of human-centered care ^[3].

Where the evidence sits versus where recommendations point. In the burnout review, 30 of 33 studies used individually focused interventions and only three were organisationally focused ^[2]. Similarly, 31 studies used secondary-level interventions, which manage stress in individuals, and only two used primary-level interventions, which eliminate the causes of stress ^[2]. Mindfulness-based practices were used in 20 studies ^[2]. The nursing quality and safety review, by contrast, places organisational domains such as leadership, safety culture and staffing among its eight focus areas ^[4]. The burnout evidence base is therefore concentrated on individual coping, while practice-model recommendations extend to the organisational level, where the burnout review found few studies ^[2][4].

How strong the evidence is

The aficamten trial reports the elements of a prespecified randomized design ^[1]. Its primary end point was change in peak oxygen uptake from baseline to week 24, measured by cardiopulmonary exercise testing ^[1]. Its 10 secondary end points were prespecified and tested hierarchically ^[1]. Peak oxygen uptake rose by a mean of 1.8 ml per kilogram per minute (95% CI, 1.2 to 2.3) with aficamten and by 0.0 ml per kilogram per minute (95% CI, -0.5 to 0.5) with placebo ^[1]. The least-squares mean between-group difference was 1.7 ml per kilogram per minute (95% CI, 1.0 to 2.4; P<0.001) ^[1]. All 10 secondary end points were significantly improved with aficamten compared with placebo ^[1]. The incidence of adverse events appeared similar between groups ^[1]. The trial population had a mean age of 59.1 years, 59.2% were men, and the baseline mean resting outflow tract gradient was 55.1 mm Hg ^[1].

The four workforce and safety abstracts report their evidence in less specific terms, and each in a different way ^[2][3][4][5]. The checklist analysis is a meta-meta-analysis, but its clinical outcome data come from observational pre-post cohort studies, so its clinical findings rest on observational comparisons ^[5].

The burnout review states that the outcomes of numerous included studies were affected by design limitations, specifically no control or a waitlist control, and no post-intervention follow-up ^[2]. The abstract does not say how many of the 29 studies reporting effective outcomes were among those with these limitations ^[2].

The AI review judged most of its studies at moderate risk of bias ^[3]. Its authors still conclude that the evidence suggests improvements in critical thinking, learner engagement and clinical satisfaction ^[3]. Its finding on burnout is reported as an association ^[3].

The nursing quality and safety abstract reports the number of included articles and recommendations ^[4]. It does not report the designs of the included articles or how the evidence behind each recommendation was graded ^[4]. Readers therefore cannot judge the strength of individual recommendations from the abstract alone ^[4].

What remains unknown

The sources leave several questions open ^[2][3][4][5].

• Whether organisational, primary-level interventions perform better or worse than individual ones cannot be judged from the burnout review, which found only three organisationally focused studies ^[2].
• How long reported well-being gains last is unclear, because numerous burnout studies lacked post-intervention follow-up ^[2].
• The cost-effectiveness of AI in nursing is not established in the review abstract; its proposed Nursing AI Integration Roadmap (NAIIR) instead emphasizes rigorous economic evaluation ^[3].
• Whether the checklist associations would hold in other designs is not addressed, since the pooled clinical data are pre-post observational ^[5].
• How well recommendations developed for military hospitals transfer elsewhere is untested; the authors suggest the practices could be adapted to many healthcare delivery systems ^[4].
• Whether reducing workload through AI or organisational change improves patient safety outcomes, as distinct from staff well-being, is not reported in any of the abstracts ^[2][3][4].

Limitations

This review is based on published abstracts, not full texts, so details of effect sizes, subgroup results and risk-of-bias judgments may be incomplete ^{[1][2][3][4][5]}. The five sources were not identified through a systematic search and cover different populations, settings and interventions, so their agreement is thematic and was not tested statistically ^{[1][2][3][4][5]}. The aficamten trial concerns a cardiac drug in patients with obstructive HCM and is included only as a methodological comparison, not as evidence about workforce interventions ^[1]. That trial was funded by Cytokinetics ^[1], and the nursing quality and safety review and the checklist analysis are both indexed as having non-U.S. government research support ^[4][5]. The burnout review restricted inclusion to English-language full texts ^[2].