In this study we demonstrated that sepsis-coded hospital admissions increased 7.5-fold from 27.9 admissions per 100,000 population in 1998 to 210.4 admissions per 100,000 population in 2023. Sepsis admissions increased as a proportion of all-cause hospital admissions over the same time-period. This observation corroborates previously published data from English ICUs that demonstrated a rising incidence of sepsis between 1996 and 2004 [13] and 2011 to 2015 [5]. Our results align with the findings of studies from Europe, Australia and New Zealand and the United States which have demonstrated rising rates of sepsis [6,7,8,9, 22]. We found that sepsis-coded admissions did not increase at a constant rate but displayed a complex pattern with periods of varying acceleration and decline. The most pronounced increases occurred between 2010 and 2014 and in 2017.

Changes to coding methodology may have had a major impact on the incidence of sepsis-coded admissions described in this study. Specifically, an apparent spike in hospital admissions due to sepsis was observed in 2017 for all pathogens except Listeria spp. and meningococcal sepsis. This coincided with NHS Digital-mandated changes to sepsis diagnostic coding guidance in April 2017 that aimed to increase the identification of sepsis as the primary admission diagnosis [23]. Further guidance was issued in 2018 allowing local hospital coding departments a greater level of discretion and acknowledging that the previous guidance had led to a probable over-reporting of sepsis [24]. Following these adjustments, our study found that hospital admissions declined in 2018 for most pathogens, except group D streptococci and enterococci. Previous research using HES data also found that hospital admissions in England with a primary diagnosis code for sepsis increased significantly after the 2017 coding changes, followed by a decline in 2018, although admission rates remained significantly higher than in pre-2017 [25].

Changes to the definition of sepsis over the study time-period may also have affected recorded incidence of sepsis-coded admissions. There have been three major updates to the definition of sepsis, in 1991 [26], 2001 [27] and 2016 [1]. The gradual increase in sepsis-coded hospital admissions observed from the early 2000s, as well as the sharper rise seen in 2017, may have been impacted by the introduction of the new definitions in 2001 and 2016.

The introduction of sepsis prevention programmes, for example the Surviving Sepsis Campaign, may have impacted the recording of hospital admissions due to sepsis. The Surviving Sepsis Campaign guidelines were first published in 2004 [28], were updated in 2008 [29] and 2012 [30], and again most recently in 2021 [12]. Such initiatives may have facilitated improved recognition of sepsis within emergency departments leading to the increases in diagnostic coding observed in this study. The guideline updates in 2008 and 2012 may also have contributed to the pronounced rise in sepsis-coded hospital admissions between 2010 and 2014 demonstrated in this study.

One factor that may be contributing to the rising incidence of sepsis-coded hospital admissions in England, aside from changes to coding practices and definitions, is an ageing population. This study demonstrated a change in the age-distribution of sepsis-coded admissions over time, with a rising incidence amongst patients aged over 75 years. Of note, the number of people aged over 65 years in England and Wales increased by approximately 20% between 2011 and 2021 demonstrating a demographic shift over the study time period [31]. A further contributing factor may be rising usage of immunosuppressive therapies for autoimmune diseases and cancers, conditions which have increased in prevalence over the study period [32, 33].

Another possible contributor to rising rates of sepsis-coded admissions in England is antimicrobial resistance (AMR). Recent data from UKHSA show that antimicrobial resistant infections are increasing in England, with the majority caused by E. coli [34]. Rising rates of AMR are likely to lead to an increase in sepsis as infections become resistant to first-line antibiotics and more difficult to treat. One study from the United States found that state-specific prevalence of resistance to several antibiotics in different bacteria is positively correlated with rates of hospitalization with septicaemia and sepsis-mortality in multiple age groups of adults [34]. A study from Taiwan found that multidrug resistance was associated with the development of severe sepsis or septic shock among patients with bacteraemic urinary tract infections [35].

The second main finding from our study was that sepsis-coded admissions caused by most pre-specified pathogens of interest rose across the study time-period. The UKHSA collects surveillance data on bacteraemias in England due to various pathogens of interest [36]. Bacteraemia is a diagnosis that is closely related to sepsis. Although not all patients admitted with sepsis will have a proven bacteraemia, it is helpful to compare trends in surveillance data with data on hospital admissions due to sepsis. An important caveat to our results is that new microbiological diagnostic techniques have been introduced over the time-period of this study, for example matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF) [37, 38]. The implementation of MALDI-TOF and other new technologies may have impacted pathogen identification which could have influenced the results of this study.

Our study found a significant increase in sepsis-coded admissions caused by enterococcal spp over time. Bacteraemia surveillance data from UKHSA demonstrate a 63.5% rise in bacteraemia due to enterococcal spp across the decade from 2013 to 2023 with the highest rates reported in those aged over 75 years [39]. UKHSA surveillance data also corroborate the rise in sepsis-coded admissions due to S. aureus shown in this study with the incidence of S. aureus bacteraemia (SAB) increasing 41.6% between 2011 and 2024 [36].

Surveillance data demonstrate a rise in bacteraemias due to Gram-negative bacteria including E. coli and Klebsiella spp across the study time-period in keeping with our findings of rising hospital admissions due to Gram-negative sepsis [36]. Sepsis due to “other Gram-negative organisms” was the commonest pre-specified ICD-10 code of interest associated with a hospital admission in this study. Whilst this was in keeping with studies of patients with sepsis admitted to intensive care units internationally [40,41,42], it is important to note that this code includes multiple different species of bacteria compared to the other codes which are single species only.

The systematic implementation of vaccination programmes over the study period may have affected the incidence of sepsis-coded admissions. Admissions due to meningococcal sepsis were unique among the pre-specified pathogens of interest as they declined over the study period. This was likely driven by the introduction of vaccines against meningococcus in England. A vaccine against group C was introduced in 1999, an infant MenB vaccine was introduced in 2015, along with a teenage programme against meningococcal A, C, W and Y, also in 2015 [43]. This finding reinforces the positive impact of vaccination campaigns despite increasing vaccine-hesitancy.

The third main finding from this study is that sepsis-coded hospital admissions have returned to the pre-pandemic trend despite a decrease in the period from April 2020 to March 2021, coinciding with the first and second waves of the COVID-19 pandemic. The largest decreases were seen for sepsis due to pathogens spread via droplets or close contact including group A streptococcus, meningococcus, Haemophilus spp, and pneumococcus. This is likely due to the impact of social distancing measures. Sepsis-coded admissions due to foodborne pathogens also decreased. This is in keeping with previous research published by the UKHSA that demonstrated a decrease in gastrointestinal illness activity during the COVID-19 pandemic [44] This may have been due to changes in behaviour, for example social distancing measures leading more people to prepare food at home. While discussing these trends, it is important to note that recorded hospital admissions due to sepsis had been declining since 2018, possibly driven by the coding change described above. An important caveat is that any data collected during the COVID-19 pandemic must be interpreted with caution due to the extraordinary pressures on the NHS and other governmental institutions at that time. The return to the pre-pandemic trend in sepsis admissions highlights the continued importance of sepsis prevention measures.

The fourth key finding in our study was that sepsis-related mortality has returned to pre-pandemic levels after decreasing sharply during 2020. The decrease in sepsis-related mortality may have been artefactual related to the competing risks of COVID-19, due to a reduction in the incidence of sepsis or increasing deaths due to COVID-19. Our study also corroborates the decline in sepsis-related mortality in England in recent decades observed in other studies using national data from hospital admissions [13] and ICU admissions [45, 46]. This decline in sepsis-related mortality occurred on a background of rising sepsis-coded hospital admissions. It is not possible to conclusively explain the underlying reasons behind the discrepancy between recorded admission-incidence and mortality. One possible explanation may be that the rise in coded hospital admissions reflects coding changes and improved recognition of sepsis rather than a rise in clinically severe sepsis. Another possible factor may be that the introduction of initiatives including the Surviving Sepsis Campaign [12], may have led to improvements in the clinical management of sepsis in hospitals that have reduced mortality.

The final important finding in our study was that sepsis contributed > 800,000 hospital bed days in 2023, compared to < 100,000 in 1998. Many cases of sepsis will require an admission to an ICU which is significantly more expensive than a standard ward. Recent data from the British government suggest that an intensive care bed costs the NHS £1881 per night compared to £901 for a standard non-elective bed [47]. Rising bed-days due to sepsis admissions and the high costs involved are likely to place a significant financial burden on the NHS. The mean median length of stay was 6.1 days across the study time-period. This is shorter than in other previously published British studies from intensive care units [45]. This likely reflects the broader range of disease severity represented in our study which captured data from all hospital admissions, including those to general hospital wards. British data suggest that only 16% of sepsis cases are referred to critical care [48]. Patients with sepsis admitted to critical care in England are likely to be those with the most severe disease who will have more complex and longer recovery trajectories than many patients admitted elsewhere.

This study has several strengths. We used population-level, aggregate datasets from NHS Hospitals, accounting for 98–99% of hospital activity in England [16]. This gives a very large sample size allowing us to draw inferences about rarer causes of sepsis such as Listeria spp. and non-typhoidal Salmonella. It also provides insights into trends across the whole of England over a 25-year period. The HES data analysed in this study is derived from diagnostic coding performed by hospital trusts. This process is audited and has been demonstrated to be sufficiently robust for use in research [17].

Our study also has limitations. Two major updates to the definition of sepsis occurred during the study period [1, 27]. These changes are likely to have impacted clinical decision-making around whether individual cases met the diagnostic criteria for sepsis. Likewise, rising awareness and recognition of sepsis amongst healthcare professionals due to awareness campaigns [12] may also have led to an increase in diagnoses. These practice changes may have affected coding outcomes as trust coding departments rely on medical notes for correct attribution of ICD-10 codes. Changes to NHS-mandated coding guidance in 2017 and 2018 will also have impacted recorded rates of sepsis-coded admissions. These factors reflect the limitations, and possible imprecision, of relying on administrative data for epidemiological analysis. Without patient-level data it is not possible to objectively confirm whether the label of “sepsis” has been correctly attributed in all cases. Despite these limitations, the concurrent rise in many bacteraemias described by UKHSA over the study period [36], supports the conclusion that there may have been a genuine rise in sepsis cases over the study period. Data from other healthcare settings indicate that ICD-10 codes in administrative data have relatively low sensitivity but high specificity for identifying pathogen-specific infections and sepsis [49, 50]. This suggests that while ICD-10 codes may identify pathogen-specific infections correctly, cases may be missed, limiting the validity of using coding to capture the full burden of pathogen-specific sepsis cases. This suggests that our study may have underestimated the burden of sepsis in England. There are no validation studies assessing the validity of pathogen-specific sepsis codes in HES-APC, highlighting an important evidence gap for future research.

A further limitation is that it was not possible to explore sepsis severity. Changes to the use and relative positions of codes for septic shock (R57.2) and organ dysfunction (R65.2) within the NHS over time meant that it was not possible to perform a meaningful analysis of temporal trends for these codes. Due to the aggregate nature of the data, it was not possible to explore the impact of individual, patient-level factors including co-morbidities on hospital admissions due to sepsis. A further limitation is that we do not have individual-level data on outcomes of patients admitted with sepsis although we have been able to analyse population-level data on sepsis mortality in England and Wales provided by the ONS across a similar time-period. Both hospital admissions due to sepsis in England and overall sepsis mortality in England and Wales declined in 2020 but due to the aggregate nature of the HES and ONS death data it is not possible to comment on the case-fatality rate of those admitted with sepsis in this period. This was noted to be higher in a large Danish study [14]. We also did not have data on how the diagnosis of sepsis due to a particular pathogen of interest was defined. Microbiological confirmation for admissions coded as due to pathogen-specific sepsis were not available in this administrative dataset. We were therefore unable to confirm whether these diagnostic codes were correctly attributed. This may have decreased the reliability of our findings.

Our study shows that the recorded incidence of sepsis-coded hospital admissions has increased in England over recent years. This is true for most of the bacterial pathogens studied. Hospital admissions due to sepsis caused by meningococcus declined, likely due to the success of vaccination campaigns. Although coding changes and improved recognition of sepsis may have contributed to rising sepsis diagnoses, the true burden of sepsis in England may be rising. Sepsis-coded admissions and sepsis-related mortality decreased during the COVID-19 pandemic but have since returned to the pre-pandemic trend. Further research is needed to monitor the trends over the coming years. An aging population, more invasive procedures and increasing levels of immunosuppression are likely to lead to a greater at-risk population. Further research into sepsis epidemiology using individual patient-level data and research into predicting and preventing sepsis is urgently needed.