Here, we’re featuring the second installment of summaries and comments on the BMJQS Editorial Board’s choice of the top 10 articles from 2019 using data such as citation rates and social media engagement in addition to their own judgement. For the full list of our 20 finalists, click here. Part 1, highighting articles #6-#10 for the year, can be found here. Overall, these results illustrate the breadth of interesting and valuable topics represented in BMJ Quality & Safety.

5. We want to know: patient comfort speaking up about breakdowns in care and patient experience – March 2019

Patients have a unique lens on the healthcare system, yet often are uncomfortable sharing their perspectives on potential safety issues. In this study, Fisher et al. developed a new item within the national Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey for eight different institutions. The goal was to identify patient characteristics associated with experiencing a problem during hospitalisation and to explore whether any discomfort in speaking up about these problems was associated with patient experience metrics. The eight hospitals had a 28% response rate on the HCAHPS survey during the study period. Nearly half of the respondents (n=4958) reported experiencing a problem with their care. Of those, 30.5% (n=1514) expressed some level of discomfort speaking up about the problem, and 13.3% never felt comfortable speaking up under any circumstances. Characteristics correlated with experiencing a problem were younger age, higher educational attainment, certain race/ethnicity demographics, and lower personal ratings of health status. Discomfort speaking up was correlated with older age, not primarily speaking English at home, admission from the ED versus other sources, and lower personal ratings of health and mental health status. Patients experiencing problems also rated several other items on the HCAHPS survey often tracked by hospital systems significantly lower than patients not experiencing them, including nurse communication (composite score 70.5 vs. 87.7), doctor communication (composite score 75.0 vs. 86.7), overall hospital rating (8.2 vs. 9.2), and likelihood to recommend the hospital (61.0% vs 81.0%). Any discomfort speaking up about problems further dropped these same ratings relative to those always feeling comfortable discussing them – nurse communication (47.8 vs 80.4), doctor communication (57.2 vs. 82.6), hospital rating (7.1 vs. 8.7), and likelihood of recommendation (36.7% vs 71.7%). These findings provide important clues about the kinds of patients most likely to experience both problems during hospitalizations and challenges in raising those problems, along with implications for how patients can best be supported.

4. Drivers of potentially avoidable emergency admissions in Ireland: an ecological analysis – June 2019

Potentially avoidable admissions are those that could possibly have been managed in another care setting rather than in the hospital. Suggestions for what drives these admissions have included socioeconomic status, barriers to accessing primary care services, and the overall availability of acute care beds. Here, Lynch et al. reviewed county-level emergency admissions throughout Ireland to determine factors related to admissions for diagnoses thought to be potentially avoidable: non-specific chest pain, COPD, UTI, non-specific abdominal pain, and falls in patients > 74 years of age. In each of their models, markers of economic deprivation (including unemployment levels), being hospitalised for only 1 day (known as short-term length of stay), and possession of a General Medical Services card (allowing those meeting low-income thresholds or other qualifying characteristics to access most health services for free) were correlated with higher levels of potentially avoidable admissions. Having private health insurance was associated with lower rates. However, these same factors were also correlated with other emergency admissions not deemed potentially avoidable. Overall, these population and patient health factors, in total, explained 50% of variation in county-level disparities in these admissions. These findings indicate that much of the variability in potentially avoidable admissions may not be related to the quality of the local primary care services. Questions about the suitability of such admissions as quality indicators should therefore be asked.

3. Characteristics of healthcare organisations struggling to improve quality: results from a systematic review of qualitative studies – January 2019

Organisations performing well on quality metrics have been found to share characteristics, such as a positive culture. However, organisations with lower performance also may share features that would be important to address within improvement plans. Using systematic review methods, Vaughn et al. explored qualitative studies of struggling healthcare organisations to identify factors associated with below-average performance, and summarised them into actionable domains using thematic synthesis. The authors identified 30 studies, including 29 that featured a high-performing comparator, across multiple different quality metrics. Poor performers consistently shared characteristics within the following five 5 domains: 1) poor organisational culture, 2) inadequate infrastructure, 3) lack of a cohesive mission and vision, 4) recent ‘system shocks’, and 5) dysfunctional external relations. Most frequently, struggling organisations exhibited examples of poor culture and inadequate infrastructure such as limited involvement/ownership (90%), poor collaboration (70%), minimal QI systems (67%), insufficient staffing (60%), a hierarchical leadership structure (57%), poor IT resources (57%), disconnected leadership (53%), and general lack of resources (50%). Less frequently noted aspects within the other domains included conflicting missions (40%), executive-level turnover (27%), and antagonism with stakeholders (13%). Overall, most below-average performers were struggling in multiple domains. Half reported a recent system shock (e.g. a scandal) that worsened performance in other domains. These findings are valuable in helping to frame organisational review processes within these domains, which could facilitate more structured and focused analyses about barriers to improvement and the specific interventions that might foster improvement.

2. Effects and costs of implementing predictive risk stratification in primary care: a randomised stepped wedge trial – September 2019

Across the National Health Service (NHS), predictive risk tools are being implemented in general practice to identify patients for case management, in part in hope that these tools might reduce emergency admissions. Here, Snooks et al. performed a stepped wedge trial across 32 practices in one Welsh health board to evaluate the costs and effects of the introduction of an emergency admission risk prediction tool (PRISM). This tool, which updates monthly, identifies the 0.5% of the practice population deemed highest risk for unplanned admission in the following year. The programme provided local education for proactive case management and support for the tool as it was introduced to each practice, but it did not specify how practices should manage care based on algorithm results. Overall, 1.8 participants per practice used the tool 8.1 times over the 16-month study period. The tool had good technical performance for admission prediction, although it generally under-predicted risk of admissions in the higher risk categories. During the intervention period, each of ED attendances (3% higher), outpatient attendances (5%), and bed days (3%) increased per participant per year. The programme was significantly less effective and significantly more costly than usual care. This sobering study highlights the importance of evaluating well-intentioned and apparently well-founded interventions to assess whether they meet their goals. Risk prediction tools require further study, particularly when the aim is to reduce emergency admissions.

1. Nurse staffing, nursing assistants and hospital mortality: retrospective longitudinal cohort study – August 2019

Hospital wards in the NHS rely on a higher average ratio of nursing assistants (NAs) to registered nurses (RNs) than in many other health systems, which may have implications for patient outcomes. In this paper, Griffiths et al. investigated variations in RN vs. NA staffing in one large hospital and how these ratios might be correlated with in-hospital mortality. The authors determined staffing for 138,000+ patients admitted to 32 wards over three years, adjusting for the patients’ diagnoses, comorbidities, admission NEWS scores, and transfers within the hospital. They then calculated the number of RN and NA hours per patient per day for each ward, normalizing these personnel rates for that specific ward across time to reflect different staffing needs across the hospital. During the three years, 4.1% of admitted patients died. Across the whole hospital, 61% of nursing positions were RNs, with a mean of 4.75 RN and 2.99 NA hours per patient per day. For each day a patient was on a ward with RN staffing below that ward’s normalized mean, the hazard of death increased by 3% (adjusted HR 1.03). Similarly, when NA staffing was below the mean, the hazard of death per day increased by 4%. On days admissions per RN exceeded 125% of the mean for the ward, the hazard of death increased by 5%, although a similar effect was not seen with admissions per NA. These findings may have significant implications for both policymakers and local managers. Specifically, there may be an optimal level of nursing staffing, below which there are detrimental effects on patient mortality. Additionally, times of increased workload for higher level nurses may also lead to increased risks to patients. Patient turnover, in particular, may be an important factor for nursing workload. Finally, the consequences of RN shortages are unlikely to be corrected solely by increasing NA staffing.