Understanding the relationship between care hours per patient day and summary hospital-level mortality indicator in English hospitals: A time series approach
Abstract
Background
The Summary Hospital-level Mortality Indicator (SHMI) offers insight into hospital mortality. Prior evidence suggests hospital staffing, particularly of registered nurses (RNs), influences hospital mortality. However, the dynamic relationship between delivered Care Hours Per Patient Day (CHPPD) and SHMI over time remains underexplored.
Methods
Retrospective analysis of National Health Service (NHS) Trust-level data from December 2020 to May 2024. Care hours data for RNs and other healthcare workers were extracted alongside SHMI values, comprising 302,406 data points from 122 Trusts. Time series including cross-correlation analysis were applied to examine temporal dependencies between CHPPD provided by different staff groups and SHMI.
Results
There appears to be a relationship between CHPPD and SHMI in English Hospitals. Significant associations were observed between CHPPD and SHMI over time. RNs, median CHPPD was highest in trusts with SHMI below control limits (Median = 5.65), followed by trusts within control limits (Median = 4.87). In trusts with SHMI above control limits, median RN CHPPD was lower (Median = 4.59). For other healthcare workers, median CHPPD in trusts with SHMI above control limits was 3.53, slightly higher than in trusts within control limits (3.51). In trusts with SHMI below control limits, the median CHPPD for other healthcare workers was lowest (3.19).
Conclusion
This study provides insight into the relationship CHPPD and SHMI across NHS Trusts. Higher RN CHPPD is consistently associated with lower mortality, Trusts with higher-than-expected mortality were asscoaited with lower RN CHPPD.
What is already known on this topic
Previous studies have shown that workforce constituencies and in particular higher numbers of registered nurses (RNs) are associated with lower hospital mortality, but have primarily relied on cross-sectional regression models, though an emerging body of work has also applied within hospital and multiyear data to examine these relationships. However, such approaches still often limit the ability to capture dynamic, long-term trends.
What this study adds
This study applies time series analysis to examine the relationship between Care Hours per Patient per Day (CHPPD) distribution provided by different types of healthcare professionals and the relationship with a nationally agreed mortality index-the Summary Hospital-level Mortality Indicator (SHMI) over time. The analysis demonstrates an association between RN staffing levels and SHMI, with higher RN CHPPD associated with SHMI within or below control limits. In severely understaffed hospitals, other healthcare worker groups may also contribute to reducing SHMI, though their impact appears more limited. The findings also show that hospitals operating below SHMI control limits experience the greatest improvements in SHMI when RN CHPPD increases, highlighting the importance of optimal staffing levels.
How this study might affect research, practice or policy
Hospital administrators and policymakers should carefully consider how staffing changes, such as abscence of RNs or replacing RNs with other workers, including support workers, leads to increases in mortality. It may also call into question any scenario where CHPPD falls for example due to the addition of extra beds. Workforce planning should prioritise evidence-based staffing models to ensure patient safety and quality of care.
Introduction
Understanding mortality and its variation across healthcare settings is a fundamental approach to assessing healthcare system performance and population health.1 In the UK, national reports on hospital mortality have consistently highlighted significant variations between hospitals, even after adjusting for case-mix, patient demographics, and regional healthcare differences.2,3 These variations raise important questions about the underlying factors influencing hospital mortality, including healthcare staffing levels, resource allocation, inequalities and system-wide policies, some of which could be improved.
The SHMI is a key metric used by the National Health Service (NHS) in England to track mortality at a point in time across hospitals in England. SHMI provides a rolling 1-year ratio between the actual number of patients who die following admission to hospital or within 30 days of discharge and the number that would be expected to die based on average England figures for a similar case-mix.4 From a national perspective SHMI levels fall within, above, or below expected control limits, risk adjusted for specific variables. Previous reports on SHMI data have shown that some hospitals consistently exhibit a SHMI above control limits (excess deaths), while others report values below control limits (lower-than-expected deaths), suggesting potential differences in care quality, efficiency, inequalities and patient safety measures.4
In April 2008, the Healthcare Commission launched an investigation into what it regarded as a concerning reaction by Mid Staffordshire Hospital to mortality statistics and the number of complaints about poor care. This led to a number of investigations into why issues of care quality and higher-than-expected hospital mortality occurred.5 The Francis Report (2013) which investigated the Mid Staffordshire NHS Foundation Trust revealed severe failings in care quality, with higher-than-expected mortality rates linked to chronic RN understaffing, poor leadership, and inadequate patient safety monitoring.6 Among its key recommendations, the Francis Report emphasised the critical role of adequate RN staffing in ensuring patient safety. This reinforced the need for systematic workforce planning, better staff-to-patient ratios, and improved monitoring of hospital mortality as a safety measure.6
Previous studies have shown that hospital staffing levels, particularly the availability of RNs, impacts mortality. Existing evidence indicates that increased RN staffing correlates with lower mortality rates, with studies from the UK, US, and Europe demonstrating that hospitals with higher nurse-to-patient ratios tend to have improved patient outcomes.7–10 A recent study found for every increase of one patient per RN; there was a 7% increase in the likelihood of in-hospital mortality.11 In some studies, not only do RNs improve outcomes, but so does a complex multidisciplinary approach which nurses often co-ordinate.12
An approach by some health systems to overcome ongoing skilled staffing shortages and financial pressures is the taskification of care and shifting of tasks or substitution of healthcare professionals for other types of workers. In such circumstances, RNs have been replaced by nursing support workers in hospital wards and other units. However, the impact of these workforce changes on patient outcomes remains unclear.13 Given the central role of RNs in delivering safe and effective care, it is essential to assess how these substitutions influence key indicators such as patient safety, mortality ratio, and overall quality of care.
Despite this growing body of research, key gaps remain in understanding of how staffing variations interact with hospital mortality. The CHPPD metric, introduced by NHS England, across NHS Trusts which are groups of NHS hospitals, is the national principal measure of nursing, midwifery, and healthcare support staff deployment on inpatient wards, essentially showing how many hours of care are provided to each patient per day.14 It provides a measure of comparison of workforce deployment across ward/unit/Trust in terms of care hours delivered rather than other workforce attributes such as experience, which can support quality and performance indicators such as quality of care, patient outcomes, people productivity, and financial sustainability.13 However, little is known about its relationship with SHMI and whether CHPPD can serve as a predictive measure for patient safety outcomes.
This study aims to examine the relationship between CHPPD and SHMI, across NHS hospitals in England.
Methods
Study design
A retrospective cohort study was conducted, analysing NHS Trust-level data using Time Series Analysis. Data was analysed in Python 3.10 software using the following libraries: Scipy, NumPy, Matplotlib, Pandas, and OS libraries. The data for this research were obtained from NHS Digital and is in the public domain. The dataset includes detailed records on NHS England's CHPPD and the Summary Hospital-SHMI from December 2020 to May 2024 for 122 acute trusts.
Key variables
The time series variables analysed were the following:
•
CHPPD for RNs refers to the care hours per patient per day provided by RNs.
•
CHPPD for other healthcare groups: captured by a combined grouping of aggregate care hours of the following professionals (Healthcare Support Workers, Registered Nursing Associates, Unregistered Trainee/Student Nursing Associates, Registered Allied Health Professionals (AHPs), and Unregistered AHPs.
•
SHMI: the SHMI is a nationally published measure used by NHS England to compare mortality outcomes across non-specialist acute trusts. It is defined as the ratio between the actual number of patients who die following hospitalisation at a trust and the number that would be expected to die based on average England figures for patients with similar characteristics. SHMI includes all deaths of patients admitted to non-specialist acute trusts in England, whether they occur in hospital or within 30 days of discharge. Bandings are reported to indicate whether a trust's mortality is ‘higher than expected’, ‘as expected’, or ‘lower than expected’. The expected number of deaths is estimated using statistical models that adjust for patient characteristics, including the primary diagnosis, comorbidities, age, sex, method and month of admission, and, for perinatal groups, birthweight. SHMI does not deliberately adjust for other factors such as socioeconomic deprivation, ethnicity, or clinical acuity. Including such controls could inadvertently legitimise poorer outcomes in more deprived or ethnically diverse areas. These factors are excluded to avoid masking differences in quality of care between Trusts that serve demographically distinct populations. Additional variables such as observed and expected deaths and upper/lower control limits were also examined.
It is important to highlight that the CHPPD data were available at a monthly frequency, whereas the SHMI data were smoothed using a 12-month rolling window. Each SHMI value reflects a 12-month lookback period and is labelled according to the final month in that window. This smoothing was an intrinsic characteristic of the dataset and was not a methodological choice made for this study. The dataset CHPPD does not include medical staffing data.
Data preprocessing
The data obtained from NHS Digital comprises 42 monthly CHPPD datasets at the ward level, 42 monthly CHPPD datasets at the trust level, and one SHMI dataset, totalling 302,406 data points across trusts and wards over time. For this study, the 42 monthly CHPPD datasets at the trust level were merged with the SHMI dataset by date and trust code, resulting in a single dataset containing 4700 data points at the trust level.
To ensure a comprehensive understanding of the dataset, the initial step involved examining its structure, including identifying missing values, inconsistencies, or anomalies that could potentially influence the analysis. As there were no missing values, imputation was not required in this step. However, a pipeline was developed to extract, transform, and load the data from the 85 monthly CSV files, addressing inconsistencies in column names, category names, and trust names. Once all data was cleaned, it was consolidated and saved into a single, well-structured CSV file for subsequent analysis.
Following this, a feature engineering process was undertaken. The CHPPD values and the pre-smoothed 12-month rolling SHMI data were mean-centered by subtracting their respective means. This transformation facilitated the calculation of cross-correlation by ensuring that correlations reflected variations rather than absolute values.
Exploratory data and cross-correlation analysis
Subsequently, an exploratory data analysis was conducted, incorporating summary statistics. This was followed by a Cross-correlation analysis to explore the temporal relationship between CHPPD and SHMI across different Trusts. The raw cross-correlation was computed using NumPy's correlate function. This function calculates the cross-correlation between two time series as a function of the lag (time shift) applied to one of the series. Specifically, it computes the dot product between the two series at different lags, effectively measuring the similarity between the series as one is shifted relative to the other. The output is a sequence of correlation values corresponding to different lag values, which helps identify potential lead–lag relationships between the variables.
A normalisation was applied through each trust cross-correlation ensuring comparability across trusts and facilitating interpretation. A significance threshold was defined as , corresponding to the 95% confidence interval. Cross-correlation values exceeding this threshold were deemed statistically significant.
The analysis focused on negative lags – where staffing levels precede SHMI – in order to assess whether fluctuations in CHPPD staffing levels are associated with variations in the mortality indicator. Given SHMI's 12-month rolling construction, a change in staffing (CHPPD) can influence SHMI gradually across the subsequent months that enter the rolling window. To capture both immediate and delayed associations created by this smoothing, the analysis focused on negative lags from 0 to −18 months (staffing leading SHMI by up to 18 months). This window covers the full 12-month SHMI accumulation period and allows for spill-over beyond a single annual cycle due to overlapping windows and operational delays (e.g. sustained staffing shifts, case-mix drift). Because each SHMI value overlaps with the preceding 11 months, a single change in staffing can continue to influence multiple consecutive SHMI values, which justifies examining up to −18 months rather than limiting the analysis to −12. Lags beyond −18 months and positive lags (SHMI leading staffing) were considered non-interpretable for this research question and are not used in any reported results.
For each trust, the normalised cross-correlation was plotted against the lag values to visualise the temporal relationship between CHPPD and SHMI. The analysis and accompanying plots provided valuable insights into the temporal dependencies and potential lead–lag relationships between the two variables within each trust.
Cross-correlation analysis was conducted across three trust groups, categorised by SHMI classification (above, within, or below expected), as defined in NHS England's SHMI publication using 95% over-dispersion-adjusted Poisson control limits. The analysis was first performed within each trust to examine temporal relationships between staffing levels (CHPPD) and SHMI. Cross-correlation results were then compared across the SHMI categories to assess cross-sectional differences in the association between staffing levels and mortality.
A fuller explanation can be found in the supplementary material entitled ‘Cross-Correlation Analysis Step-by-Step Methodology’.
Results
Standardised Hospital Mortality Index across English National Health Service Hospitals
A total of 122 Trusts were reviewed, with 22% presenting outlier mortality results. The scatter plot in Figure 1 illustrates the relationship between the SHMI and the expected number of deaths for hospitals across England from June 2023 to May 2024.
The majority of Trusts fall within the control limits, suggesting their mortality outcomes align with expected levels adjusted for specific variables as per national SHMI methodology.4 However, 11% of trusts exceed the upper control limit, indicating higher-than-expected mortality that may require further investigation into potential contributing factors. Conversely, 11% of trusts fall below the lower control limit.
CHPPD by SHMI cohorts
Overall trends
From the 122 Trusts, over the three cohorts, notable differences were observed in CHPPD distribution between RNs and other healthcare workers, as well as across trusts with SHMI above, within, and below control limits (Figure 2). For RNs, the median CHPPD was lowest in trusts with SHMI above control limits (Median = 4.59), higher in trusts within control limits (Median = 4.87), and highest in trusts with SHMI below control limits (Median = 5.65). This indicates fewer RNs’ CHPPD are provided in trusts with SHMI above control limits. This association of higher RN CHPPD being associated with lower-than-expected mortality is consistent with the hypothesis, both in this study and examined in the literature, that higher RN staffing availability is associated with lower mortality. The association of lower CHPPD for other health workers with SHMI below control limits (and higher in trusts above the control limit) may reflect an attempt to substitute for RNs with other health workers, raising the hours overall even where RN hours are lower.
For other healthcare workers, a different pattern was observed. The median CHPPD was highest in trusts with SHMI above control limits (Median=3.53), slightly lower in trusts within control limits (Median=3.51), and lowest in trusts with SHMI below control limits (Median=3.19). The RN CHPPD difference across the three cohorts (1 hour) contrasts with the difference in the other workers across the three cohorts (34 min). This may also have significance and requires further investigation.
These findings could indicate differences in resource distribution and nursing workforce configurations among trusts with varying SHMI. The highest median CHPPD for RNs in trusts with SHMI below the control limits suggests an association with maintaining care quality and achieving more favourable clinical outcomes. The lowest median CHPPD for other healthcare workers is in trusts with SHMI below control limits. This raises important questions about whether these variations in types of staffing translate into differences in mortality outcomes at the hospital level. To explore this, the association between CHPPD for types of healthcare professionals and SHMI was examined.
SHMI and CHPPD groupings across hospitals
One hundred and twenty-two Trusts were categorised into three groups: those within the control limits (n = 97), those with SHMI below control limits (n = 13), and those with SHMI above control limits (n = 12). For each cohort of Trusts, both negative (higher staffing associated with lower mortality and lower staffing with higher mortality) and positive correlations (higher staffing associated with higher mortality and lower staffing with lower mortality) were identified, along with instances of no significant correlation. Only those Trusts with statistically significant cross-correlations were included under the negative and positive cross-correlations.
Association of registered nurse staffing hours provided with SHMI
In 44% to 50% of hospitals, depending on SHMI category, an increase in the number of RNs CHPPD was significantly associated with a reduction in SHMI, whereas a decrease in RNs CHPPD was significantly associated with an increase in SHMI over time, indicating registered nursing levels are associated with hospital mortality outcomes.
This effect was most pronounced in hospitals with SHMI above control limits, where the greatest improvements were observed (50%). However, even in the hospitals with SHMI below control limits, increasing RN staffing continued to yield SHMI improvements (46%), and the same applies vice versa.
Association of other healthcare workers with SHMI
The impact of other healthcare workers (e.g. healthcare assistants, nursing associates and Allied Healthcare Professionals) on SHMI is evident but not as pronounced as that of RNs. Among hospitals with SHMI below control limits, only 38% showed a reduction in mortality when the number of these staff members increased or an increase in mortality when their staff members decreased.
In contrast, in 50% of hospitals where the SHMI exceeds control limits, increasing the number of non-registered healthcare workers was associated with a reduction in SHMI, while reducing staff numbers was linked to an increase in SHMI. This suggests that in hospitals where SHMI is well above control limits, increasing the number of workers of any kind reduced excess mortality.
Association of total staffing levels with SHMI
Between 42% and 62% of hospitals, depending on SHMI category, showed a significant association between increased overall staffing levels and a decrease in SHMI. Hospitals in which the SHMI was below control limits had the highest impact, with 62% showing reduced SHMI when staffing levels increased, or increased SHMI when staffing levels decreased. This reinforces the notion that even hospitals with SHMI below control limits will still benefit from additional staff.
Positive cross-correlations, ranging from 8% to 40% of hospitals, were identified depending on the Staffing and SHMI category. This effect was weakest (8%) in just one hospital with SHMI above control limits, suggesting that in these settings, increasing or decreasing staffing generally did not lead to SHMI above the control limits. These hospitals might have difficulty achieving substantial improvements despite staffing increases, likely due to insufficient staffing increase for impact to be seen or combined with other systemic issues beyond CHPPD alone. The presence of these positive correlations likely reflects confounding factors such as operational disruptions during the COVID-19 pandemic, increased patient acuity, workforce skill level mismatches, and other structural challenges.
Discussion
This study investigated the relationship between CHPPD and SHMI using time series analysis. Using CHPPD as a reflection of staffing levels, particularly of RNs, appears to play a crucial and consistent role in influencing hospital mortality. One of the most significant findings in this study is the association between RN staffing levels and reductions in SHMI. This effect was most pronounced in hospitals with above SHMI control limits, but an effect in SHMI was also observed in hospitals where CHPPD was below control limits. This suggests that RNs continue to contribute positively to patient outcomes, even in institutions with lower SHMI. These findings align with the broader literature, which consistently demonstrates the impact of RNs on hospital safety, patient survival, and quality of care.8,11
Importantly, hospitals with SHMI below control limits also exhibited an effect on SHMI over time, indicating that these institutions are not necessarily overstaffed. Instead, they appear to be operating at nearer optimal CHPPD levels, reinforcing the argument that investing in RN staffing yields ongoing benefits regardless of initial SHMI.
While RNs had the most significant association with SHMI mortality ratio, in some hospitals other healthcare professionals also contributed to reductions in SHMI, particularly in hospitals where SHMI was above expected limits. In these settings, where likely severe understaffing occurs, workforce expansion – regardless of professional category – can have impact.
Although the number of organisations was smaller below the control limit/lower-than-expected mortality, those organisations with SHMI below control limits the impact of other staff appeared more limited. This supports concerns raised in previous research that replacing RNs with other healthcare workers who are not qualified RNs may not yield the same benefits in terms of patient safety and mortality.7
This concern is particularly relevant as hospitals increasingly move towards workforce substitution models, replacing RNs with other workers to reduce costs or fill workforce gaps. A previous study found that wards with higher levels of nursing assistants had higher rates of failure to rescue, urinary tract infections, and falls with injury, despite a decrease in mortality.15 While cost-saving is often the primary motivation for these staffing changes, previous research has not demonstrated their cost-effectiveness, as poorer patient outcomes may ultimately lead to higher long-term healthcare costs. In contrast, increasing the proportion of RNs in medical and surgical wards has been associated with improved patient outcomes and potentially lower overall costs.16
This aligns with the study observation that increases in staffing levels across all professional groups were linked to reductions in SHMI in 62% of hospitals below SHMI control limits, further supporting the importance of adequate staffing in improving patient outcomes. However, the greatest benefits were observed in hospitals with lower SHMI, suggesting that these hospitals were not overstaffed but instead optimising workforce capacity. These findings are consistent with previous research showing that higher RN staffing levels correlate with better patient outcomes, particularly in terms of mortality reduction, less missed care, and prevention of adverse events.17,18 These findings underscore the risks of cost-driven workforce changes and reinforce the need for evidence-based staffing policies that prioritise patient safety and quality of care.
While these findings align with existing literature, other signals were observed on occasion, such as higher overall CHPPD correlating with increased mortality and lower staffing levels correlating with reduced mortality. These signals could be due to a number of factors including external confounders, but illustrate why caution is needed when only looking at overall staffing numbers and not breakdown in staff groups as this may mask effect. This data also covers the period of COVID-19 pandemic, which placed unprecedented strain on healthcare systems, is potentially, the main contributor to this effect to these positive cross-correlations as the majority of the data captures the period during and immediately after the COVID-19 pandemic. During peak periods, staff were redeployed across services, leading to skill mismatches and disruptions in workforce continuity. The influx of less experienced workers combined with system-wide overload aligned with the acuity of patients and limited treatments may have reduced the effectiveness of increased staffing levels.19
There are other factors which could contribute to these findings. Health inequalities and patient complexity were not fully accounted for in SHMI adjustments. Socioeconomic deprivation, multimorbidity, and delayed hospital presentations are known to influence mortality rates, yet these factors are not consistently adjusted for in SHMI calculations.4 Two further factors, more related with health systems operations can be due to workforce policy and hospital configurations, including reactive rather than proactive hiring, skill mix adjustments, replacements of staff, and changes in patient flow management such as delayed discharges and practices such as ‘boarding’ in which extra beds are added to clinical areas but extra staff are not provided. Staffing adjustments may have lagged behind patient needs, leading to misalignment between workforce capacity and actual demand, leaving care undone.20
Another potential explanation for the absence of improvements in some hospitals despite staffing increases is the quality of the work environment. Previous research demonstrated that in hospitals with poor work environments, increases in nurse staffing levels had no effect on reducing mortality.21 This suggests that staffing increases alone may be insufficient unless accompanied by organisational conditions that enable staff to work effectively, such as supportive leadership, adequate resources, and positive interprofessional dynamics.
In hospitals where SHMI exceeded control limits, the relationship between staffing levels and SHMI was often non-significant (42%–33% of cases, depending on the staff category). This suggests that, in these settings, CHPPD levels remain too low to generate measurable improvements. Previous studies have similarly found that staffing thresholds reach a saturation point before mortality reductions become evident.10,22 Conversely, in hospitals with lower SHMI, increases in staffing had a significant impact, reinforcing the importance of maintaining adequate workforce levels to achieve better patient outcomes.
Hospitals with high SHMI may require not only more staff but also enhanced workforce training, improved care coordination, and better alignment of workforce skills with patient needs. This aligns with research indicating that workforce planning must be paired with organisational and structural improvements to yield the greatest impact.23–25
Implications for policy and practice
These findings highlight the relationship of care provided by RNs in reducing excess mortality and maintaining care quality over time, while emphasising the limited impact of other members of staff. Workforce expansion alone may not improve outcomes, especially in hospitals with SHMI above control limits. The COVID-19 pandemic has also likely distorted workforce-related trends.
While SHMI is a useful performance indicator, it should not be the sole measure of care quality and patient safety. Mortality reflects the endpoint of a patient's journey but does not fully capture hospital performance. Other metrics are equally, if not more, informative in evaluating care effectiveness. Therefore, while findings highlight significant relationships between care hours provided by different workers and SHMI, they should be interpreted within the wider context of hospital performance as mortality at an aggregate is not a suitable surrogate measure of overall quality.26
Strengths and limitations
Prior studies examining the relationship between staffing and hospital mortality have typically used one of two approaches: either cross-sectional analyses across hospitals using average staffing levels, or longitudinal analyses within a limited number of trusts, often using shift-level or trust-level data to explore intra-trust variation. While both approaches have contributed with valuable insights – particularly into nurse-to-patient ratios and their association with mortality – each comes with limitations in terms of scalability and generalisability.
This study offers a novel contribution by applying a trust-level time series cross-correlation approach to explore the dynamic relationship between CHPPD and the SHMI across 122 NHS Trusts over 42 months. This design allows us to combine national-scale coverage with the ability to examine variation over time. Unlike shift-level studies that often focus on a small number of hospitals, this approach provides broader validity while still capturing meaningful temporal patterns in staffing.
One of the key strengths of this methodology is its capacity to explore lead–lag relationships between staffing and mortality, offering a more dynamic perspective than traditional cross-sectional studies. The use of standardised national datasets (CHPPD and SHMI) allows for consistent comparisons across Trusts and helps uncover trust-level patterns that may be masked in more localised studies. By stratifying CHPPD into two distinct workforce categories – RNs and other healthcare workers – and comparing results across SHMI performance groups (above, within, or below control limits), the analysis also captures how different staffing and workforce compositions influence mortality outcomes.
A key limitation of this study is the mismatch between the way SHMI is reported (as a 12-month rolling average) and the monthly frequency of CHPPD data. The smoothing inherent in SHMI reduces the visibility of short-term fluctuations and creates challenges for direct alignment with staffing measures. Due to these constraints, the data could not reasonably be transformed into a stationary format without obscuring the medium-term patterns of interest. To address this, lead–lag cross-correlation methods were applied, allowing the exploration of associations between monthly changes in staffing and medium-term mortality outcomes. This approach helps account for the temporal mismatch, though it restricts interpretation to medium-term patterns rather than immediate effects.
This approach focuses on negative lags to assess whether fluctuations in staffing precede changes in mortality, which is relevant for real-world planning scenarios where the effects of workforce changes may take months to emerge. As a result, the study complements both cross-trust comparisons and within-trust shift-level studies, helping to bridge the gap between granular variation and broader systemic trends.
Trusts were categorised into SHMI performance groups (above, within, or below expected) based on NHS England's October 2024 publication, which reflects data from June 2023 to May 2024. However, SHMI performance can vary over time, and this static classification does not account for longitudinal changes in Trust performance.
Another important consideration is the degree of variation in CHPPD within trusts over time. Cross-correlation relies on temporal variation to detect associations, and in trusts where CHPPD showed low variation over time, the ability to observe meaningful correlations with SHMI is reduced. In these settings, associations may appear weaker or more random, even if staffing still plays a role in outcomes. This factor should be noted as it may have contributed to heterogeneity in results.
In addition, SHMI has been risk adjusted to specific metrics; however it hasn’t been adjusted for Covid, not all health inequalities neither includes data on community hospitals where substitution of staff is even more apparent. It also does not account for unmeasured confounders such as patient acuity, staff experience, hospital resource allocation, variations in medical interventions and performance, or factors related to the work environment and organisational culture.
Similarly, CHPPD data are reported in absolute terms and are not adjusted for patient complexity or illness severity. This limits comparability across Trusts, particularly those with higher baseline mortality risks who may justifiably staff at higher levels.
Finally, while cross-correlation analysis enables the identification of temporal associations, it does not provide evidence of causality or quantify marginal effects. The approach is exploratory, aimed at uncovering potential temporal relationships rather than making causal inferences.
Future research
Incorporating patient-level data – such as comorbidities, socioeconomic factors, and acuity levels – could refine analyses and account for population-level differences. Additionally, exploring operational and Trust-level metrics would provide a more comprehensive view of system-wide variations affecting SHMI.
Incorporating SHMI monthly values, data by specialisation and from community hospitals could provide further information into mortality ratios and care hours per patient so future studies can better capture the short-term impact of CHPPD changes. This approach would help uncover temporal trends and mitigate variance dilution effects that may obscure important findings.
Researching practices in hospitals with SHMI below control limits could help identify best practices and inform targeted improvements, supported by interpretive methodologies utilising qualitative data to give further insight. Other techniques such as data mining of incident reporting have been used successfully in explanative studies to further understand effects.27 Lastly, research and policy efforts should expand to include broader patient-centred and process-oriented outcomes, ensuring that strategies address not only SHMI but also overall patient safety, experience, and long-term well-being.
Conclusion
This study provides new knowledge into the relationship between CHPPD and SHMI across NHS Trusts using a time series approach. These findings inform workforce composition and demonstrate that higher RN CHPPD is significantly associated with lower SHMI, particularly in Trusts with higher-than-expected SHMI. This reinforces the established cross-sectional international studies that assert the role of RNs in ensuring patient safety and reducing adverse outcomes or the absence of RNs associated with poor outcomes. This calls into question the strategy of replacing RNs with other workers. Additionally, other healthcare workers still have an effect in hospitals where mortality ratio is higher-than-expected, suggesting that workforce composition plays a crucial role in patient outcomes. Organisations should consider examining SHMI in relation to workforce data on a local basis for insight.
Acknowledgements
Thea Stein at Nuffield Trust.
Ethical considerations
This is a secondary analysis of routinely collected publicly available data hence ethical approval was not sought and the data is freely available.
Consent to participate
This is a secondary analysis of routinely collected publicly available data hence consent form individuals is not required-the data is freely available in the public domain.
Consent for publication
N/A
Declaration of conflicting interests
Just prior to submissoin, Alison Leary was appointed Deputy President of the Royal College of Nursing UK.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was partly funded by the Norfolk Initiative for Coastal and rural Health Equalities (NICHE) Anchor Institute at the University of East Anglia.
ORCID iDs
Geoffrey Punshon https://orcid.org/0000-0001-9247-9546
Alison Leary https://orcid.org/0000-0001-7846-5658
Cidália Eusébio https://orcid.org/0000-0002-3540-8386
Footnote
Guarantor The corresponding author takes responsibility for the integrity of the work and confirms that he controlled the decision to publish.
Data availability statement
This study uses publicly available data.
References
1. Liang CY, Kornas K, Bornbaum C, et al. Mortality-based indicators for measuring health system performance and population health in high-income countries: a systematic review. IJQHC Commun 2023; 3: lyad010.
2. Jarman B, Gault S, Alves B, et al. Explaining differences in English hospital death rates using routinely collected data. Br Med J 1999; 318: 1515–1520.
3. Bottle A, Faitna P, Brett S, et al. Factors associated with, and variations in, COVID-19 hospital death rates in England’s first two waves: observational study. BMJ Open 2022; 12: e060251.
4. Summary Hospital-level Mortality Indicator (SHMI) - Deaths associated with hospitalisation. NHS England Digital, https://digital.nhs.uk/data-and-information/publications/statistical/shmi (accessed 31 January 2025).
5. Investigation into Mid Staffordshire NHS Foundation Trust. Commission for Healthcare Audit and Inspection, https://cdn.ps.emap.com/wp-content/uploads/sites/3/2009/06/Investigation_into_Mid_Staffordshire_NHS_Foundation_Trust.pdf (March 2009).
6. Powell T. The Francis Report (Report of the Mid-Staffordshire NHS Foundation Trust public inquiry) and the Government’s response, https://commonslibrary.parliament.uk/research-briefings/sn06690/ (2025, accessed 31 January 2025).
7. Zaranko B, Sanford NJ, Kelly E, et al. Nurse staffing and inpatient mortality in the English national health service: a retrospective longitudinal study. BMJ Qual Saf 2023; 32: 254–263.
8. Griffiths P, Maruotti A, Saucedo AR, et al. Nurse staffing, nursing assistants and hospital mortality: retrospective longitudinal cohort study. BMJ Qual Saf 2019; 28: 609–617.
9. Rubbo B, Saville C, Dall’Ora C, et al. Staffing levels and hospital mortality in England: a national panel study using routinely collected data. BMJ Open 2023; 13: e066702.
10. Needleman J, Buerhaus P, Pankratz VS, et al. Nurse staffing and inpatient hospital mortality. N Engl J Med 2011; 364: 1037–1045.
11. Aiken LH, Sloane DM, Bruyneel L, et al. Nurse staffing and education and hospital mortality in nine European countries: a retrospective observational study. Lancet Lond Engl 2014; 383: 1824–1830.
12. Dall’Ora C, Rubbo B, Saville C, et al. The association between multi-disciplinary staffing levels and mortality in acute hospitals: a systematic review. Hum Resour Health 2023; 21: 30.
13. World Health Organisation. Strengthening Europe’s Nursing Workforce, https://iris.who.int/bitstream/handle/10665/379298/Policy-brief-66-1997-8073-eng.pdf?sequence=6 (2024).
14. England NHS. NHS England » Care hours per patient day (CHPPD): guidance for all inpatient trusts, https://www.england.nhs.uk/long-read/care-hours-per-patient-day-chppd-guidance-for-all-inpatient-trusts/ (accessed 3 February 2025).
15. Twigg DE, Myers H, Duffield C, et al. The impact of adding assistants in nursing to acute care hospital ward nurse staffing on adverse patient outcomes: an analysis of administrative health data. Int J Nurs Stud 2016; 63: 189–200.
16. Griffiths P, Saville C, Ball J, et al. Costs and cost-effectiveness of improved nurse staffing levels and skill mix in acute hospitals: a systematic review. Int J Nurs Stud 2023; 147: 104601.
17. Griffiths P, Ball J, Bloor K, et al. Nurse staffing levels, missed vital signs and mortality in hospitals: retrospective longitudinal observational study. Southampton (UK): NIHR Journals Library, http://www.ncbi.nlm.nih.gov/books/NBK534527/ (2018, accessed 5 February 2025).
18. Ball JE, Bruyneel L, Aiken LH, et al. Post-operative mortality, missed care and nurse staffing in nine countries: a cross-sectional study. Int J Nurs Stud 2018; 78: 10–15.
19. World Health Organization. State of the world’s nursing 2020: executive summary. Geneva: World Health Organization, https://iris.who.int/handle/10665/331673 (2020, accessed 5 February 2025).
20. Ball JE, Murrells T, Rafferty AM, et al. Care left undone’ during nursing shifts: associations with workload and perceived quality of care. BMJ Qual Saf 2014; 23: 116–125.
21. Aiken LH, Cimiotti JP, Sloane DM, et al. Effects of nurse staffing and nurse education on patient deaths in hospitals with different nurse work environments. Med Care 2011; 49: 1047.
22. Kane RL, Shamliyan TA, Mueller C, et al. The association of registered nurse staffing levels and patient outcomes: systematic review and meta-analysis. Med Care 2007; 45: 1195–1204.
23. Imison C, Castle-Clarke S, Watson R. Reshaping the workforce to deliver the care patients need.
24. Fagerström L, Kinnunen M, Saarela J. Nursing workload, patient safety incidents and mortality: an observational study from Finland. BMJ Open 2018; 8: e016367.
25. Taylor N, Clay-Williams R, Hogden E, et al. High performing hospitals: a qualitative systematic review of associated factors and practical strategies for improvement. BMC Health Serv Res 2015; 15: 244.
26. Freemantle N, Richardson M, Wood J, et al. Can we update the summary hospital mortality Index (SHMI) to make a useful measure of the quality of hospital care? An observational study. BMJ Open 2013; 3: e002018.
27. Leary A, Cook R, Jones S, et al. Using knowledge discovery through data mining to gain intelligence from routinely collected incident reporting in an acute English hospital. Int J Health Care Qual Assur 2020; 33: 221–234.
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
Cite
Cite
Cite
OR
Download to reference manager
If you have citation software installed, you can download citation data to the citation manager of your choice
Information, rights and permissions
Information
Published In
Article first published online: December 30, 2025
Keywords
Rights and permissions
© The Author(s) 2025.
This article is distributed under the terms of the Creative Commons Attribution 4.0 License (https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).
Data availability statement
Data is available for this article. View more information
Authors
Contributorship
JP, CE, SH, CSJ, SB, and AL contributed to study design, data collection, data analysis, and manuscript preparation. GP contributed to data collection and manuscript preparation.
Metrics and citations
Metrics
Journals metrics
This article was published in Journal of Patient Safety and Risk Management.
View All Journal MetricsPublication usage*
Total views and downloads: 1242
*Publication usage tracking started in December 2016
Publications citing this one
Receive email alerts when this publication is cited
Web of Science: 0
Crossref:
There are no citing articles to show.
Figures and tables
Figures & Media
Tables
View Options
View options
PDF/EPUB
View PDF/EPUBAccess options
If you have access to journal content via a personal subscription, university, library, employer or society, select from the options below:
I am signed in as:
View my profileSign out
I can access personal subscriptions, purchases, paired institutional access and free tools such as favourite journals, email alerts and saved searches.
loading institutional access options
Alternatively, view purchase options below:
Purchase 24 hour online access to view and download content.
Access journal content via a DeepDyve subscription or find out more about this option.
