In 2025, we have seen many outstanding articles published in BMJ Quality & Safety. The papers highlighted in this blog represent some of the very best from the past year and were selected based on engagement metrics and evaluation by our editorial board. The methodology used to identify these papers is described in a previous blog. This article will focus on the articles ranked first to fifth. You can read about the articles ranked sixth to tenth in a complementary blog.
We would like to express our thanks to the authors of these exceptional papers, as well as to the entire BMJ Quality & Safety team for their support throughout the selection process – choosing amongst so many high-quality contributions was not a simple task!
5 – Re-establishing control limits in statistical process control analyses: the stable shift algorithm
Accompanying editorial available here.
Statistical process control charts are widely used in healthcare quality improvement to monitor performance over time and to distinguish expected variation from meaningful change. They allow organisations to identify when a process has genuinely improved or deteriorated from expected levels of variation using control limits. Control limits define a range of expected performance based on previous data. Whilst establishing baseline control limits is usually relatively straightforward, determining the optimal time to recalculate after a sustained improvement or deterioration is more challenging. Recalculating at the wrong time risks masking future meaningful change.
This study describes the development and evaluation of a novel approach to address this problem, the stable shift algorithm. This algorithm identifies when a process shift has occurred and has become sufficiently stable to justify recalculating control limits. Rather than relying on subjective judgement or fixed time intervals, the approach uses predefined statistical criteria to determine whether a sustained change has occurred. The authors evaluated this algorithm using simulated and real-world healthcare datasets and compared its performance against existing approaches. The stable shift algorithm was able to identify meaningful and sustained change whilst avoiding premature recalculation in response to short term random variation.
Ensuring that control limits accurately reflect current performance is essential if statistical process control charts are to remain useful tools that facilitate organisational learning and decision-making. The stable shift algorithm offers an objective and methodologically rigorous approach to recalibrating control limits, reducing reliance on subjective judgement. Healthcare organisations are increasingly monitoring quantitative aspects of healthcare performance to guide improvement efforts. This process relies on the reliability of methods to support robust data interpretation and approaches that strengthen the interpretability of quantitative data are likely to play an increasingly important role in supporting effective quality improvement.
Patient safety incident investigations are often perceived to be objective and systematic processes that promote organisational learning. However, at the centre of these investigations are individuals that bring their own experience and interpretation of events, and potentially, their own bias. Outcome bias is one example of a cognitive bias that could influence the interpretation and recommendations from incident investigations. It occurs when people judge past decisions more harshly when the decision was associated with a worse outcome, even where decisions were reasonable given the information available at the time.
This study aimed to understand how participants interpreted safety incident investigations where only the outcome differed. A survey was distributed to members of the general public, healthcare staff, and experts, who were all presented with three identical fictional clinical scenarios with outcomes that resulted in either no harm, severe harm, or death. Participants were asked about healthcare professional responsibility, avoidability, and the importance of investigating. They were also asked to select recommendations to prevent such an incident occurring again. The authors found that as the severity of the outcome increased, participants considered healthcare professionals more responsible for causing it, despite the scenarios being otherwise identical. Participants also perceived the investigation to be more important where the outcome was more severe.
These findings demonstrate that outcome bias is likely to influence both the decision to initiate patient safety investigations and the conclusions drawn from them. Incident investigations are conducted by humans and are therefore not immune to broader cognitive biases that influence human judgement. Whilst expertise in patient safety partially mitigated these biases, it did not eliminate them entirely. As healthcare systems strengthen their approach to patient safety, organisations must be mindful of the ways in which cognitive bias can shape investigative processes and subsequent learnings. If decisions to investigate are driven by the severity of outcomes, rather than the likelihood of future harm or the opportunity for system improvement, there is a risk that resources become aligned to outcome severity rather than learning potential.
Exposure to valproate during pregnancy can cause major congenital malformations and neurodevelopmental disorders in exposed children. As a result, medicines regulators internationally have introduced increasingly stringent measures to reduce valproate use amongst women of childbearing potential. However, previous safety warnings issued by the Medicines and Healthcare products Regulatory Agency (MHRA) in the UK have only had modest effects on prescribing behaviour, reflecting the challenge of translating medicines safety policy into meaningful behaviour change. This has prompted further regulatory intervention by the MHRA through a National Patient Safety Alert mandating approval by two specialists before valproate can be prescribed to patients under 55.
This study examined whether stronger regulatory action could succeed where previous warnings had only achieved modest changes in prescribing behaviour. Using population-level prescribing data from England, the authors evaluated the impact of mandating approval by two specialists on valproate use. Following implementation of this policy there was a substantial and sustained reduction in valproate prescribing, suggesting that this more interruptive intervention had a considerably greater impact that earlier information-based safety communications. The findings provide important insights into the potential mechanisms through which regulatory intervention may achieve meaningful and sustained behaviour change in clinical practice.
The implications of these findings extend far beyond valproate prescribing. Healthcare organisations frequently rely on disseminating safety information in the hope that this will lead to the behaviour change, yet this may not necessarily occur. Equally, more interruptive interventions are not without cost. Additional approval steps and prescribing controls may improve safety but can also increase administrative burden, cause delays in treatment and may have negative unintended consequences. Therefore, regulators and healthcare organisations must design interventions that are proportionate to the scale and nature of the risk they are trying to avert.
For many patients, the decision to start a new medication is challenging and it is often shaped by far more than the medication; health literacy, health beliefs, previous experiences, and perceptions of illness all have a role. Non-initiation of medication is common, perhaps more so where medications are used to prevent future complications. Framing this simply as ‘non-compliance’ does not capture the nuance and complexity that surround the realities of decision making. The risks and benefits of medication are usually uncertain, difficult to communicate, and consultations are frequently constrained by time. The challenge for prescribers is not to persuade patients to accept treatment, but to support decisions that are informed, meaningful, and aligned with their individual priorities and values.
The IMA-cRCT explored this challenge through a theory-informed mixed-methods process evaluation embedded within a cluster randomised controlled trial across 24 primary care centres in Spain. The intervention aimed to improve initial adherence to cardiovascular and diabetes medications using a patient-centred approach that combined shared decision-making, decision aids, and multidisciplinary support from general practitioners, nurses, and pharmacists. To understand not only whether the intervention worked, but how and why, the authors drew on a wide range of data including monitoring data, questionnaires, field diaries, interviews and focus groups. Overall, the intervention achieved moderate fidelity and became reasonably well integrated into routine care. Healthcare professionals reported that structured shared decision-making conversations helped to improve patient understanding, although some felt these approaches already reflect aspects of existing practice. Patients valued receiving clearer explanations and being more involved in decisions about their treatment, but preferences for participation varied considerably. Some patients welcomed an active role in decision making, whereas others preferred healthcare professionals to lead. Across sites, limited consultation time, workload pressures and differing baseline levels of health literacy emerged as important barriers to implementation.
This study positions medication adherence as a product of complex interactions between patients, healthcare professionals, and healthcare organisations. Whilst this study was conducted in Spanish primary care centres, many of the barriers identified in will resonate with healthcare professionals globally. Importantly, this study highlights the importance of taking a flexible and individualised approach towards shared-decision making – an approach where healthcare professionals adapt the information communicated and involvement in decision-making based on patient preferences.
Accompanying editorial available here.
Hospital staffing levels are a common source of frustration for staff who frequently work in understaffed environments. For healthcare professionals, the effects of understaffing on both patients and colleagues are often clear to see – delays in care, staff burnout, and less time to communicate with patients. Yet for healthcare organisations, staffing decisions are rarely straightforward, and the benefits must be balanced against the limited resources available. As a result, questions around the affordability of hospital staffing levels are often central to operational decision making. Therefore, there is a need to understand not only whether higher staffing levels improve outcomes, but whether they can be justified as a sustainable investment for healthcare organisations.
This retrospective longitudinal study and economic evaluation examined the cost-effectiveness of eliminating hospital nursing understaffing in English acute hospitals. The authors used routinely collected hospital data to model the impact of increasing staffing levels and evaluated the associated costs and patient outcomes. Their findings suggest that eliminating understaffing is associated with reduced length of stay and lower mortality. Whilst increasing staffing levels inevitably comes at a financial cost, these costs were partially offset through reductions in adverse outcomes and shorter hospital stays. Notably, economic analysis suggested that improving staffing levels represented excellent value for money when considered against commonly used health economic thresholds – in fact, the cost per quality-adjusted life year was approximately ten times lower than the threshold currently used in the UK to inform approval decisions for new medicines.
The results of this study suggest that eliminating nursing understaffing may improve patient outcomes and offers exceptional value for money. Whilst not addressing a staffing gap may appear at the surface to make financial sense in the short-term, this perspective overlooks the potential downstream clinical and economic consequences that could result. These findings provide an important reminder that decisions about staffing are not only a financial decision but are also a decision about risk, and organisations should be mindful of these potential longer term clinical and economic costs associated with understaffing when designing sustainable workforce policies.