The study revealed a marked increase in the incidence and mortality of sepsis over the 20-year study period. The rise in incidence is consistent with trends observed worldwide, where population aging, improved diagnostic capabilities, and increased clinical awareness led to more frequent identification of septic cases (WHO 2020, Rhee and Klompas 2020, Rudd et al. 2020, Kaukonen et al. 2015, Hall et al. 2011, Daviaud et al. 2015, Ibarz et al. 2024). Although no formal changes in local sepsis protocols or administrative coding systems were implemented at our institution during the study period, the introduction of the Sepsis-3 definitions may have indirectly contributed to increased case recognition by promoting greater attention to organ dysfunction and disease severity (Singer et al. 2016). These trends likely reflect a combination of epidemiological and healthcare-related factors rather than abrupt organizational changes. One significant reason is the increased survival of individuals with chronic illnesses, making patients vulnerable to severe infections and sepsis. For example, patients with underlying conditions such as malignancies, diabetes, and cardiovascular diseases are at a higher risk of developing infections that can lead to sepsis due to their defective immune systems and overall compromised health status (Daviaud et al. 2015). Additionally, the aging population contributes significantly to the rise in sepsis cases. Older adults, particularly those over 65, are more susceptible to infections due to age-related physiological changes (Guarino et al. 2023, Ibarz et al. 2024). Infections, especially in hospital and long-term care settings, are often resistant to antibiotics, making them more challenging to treat and more likely into evolve to sepsis (Ibarz et al. 2024). Furthermore, the increase in sepsis incidence may reflect changes in coding practices and definitions, as the criteria for diagnosing sepsis have been revised over the years. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3, 2016) simplified the diagnostic criteria by focusing on life-threatening organ dysfunction caused by a dysregulated host response to infection (Singer et al. 2016). These revised definitions may have contributed to an increased case identification and documentation, thus explaining, at least in part, the observed rise in incidence.

Sepsis- and septic shock–related mortality observed in our cohort falls within the broad range of mortality rates previously reported in the literature, which are known to vary substantially according to patient characteristics, disease severity, and healthcare settings (WHO 2020). However, when mortality trends are examined over time, our findings diverge from part of the recent literature. Several large analyses documented a stabilization or even a gradual decline in sepsis-related mortality over the last two decades (Luhr et al. 2019, Bauer et al. 2020). A meta-analysis of randomized sepsis trials conducted between 2002 and 2016 reported a modest but significant annual reduction in mortality of approximately 0.42% (Luhr et al. 2019), while a subsequent analysis focusing on Europe, North America, and Australia between 2009 and 2019 described a general stabilization or slight decrease in mortality rates (Bauer et al. 2020).

The divergence between these findings and our results may be explained by several factors. First, differences in patient populations are likely relevant, as older age and a higher burden of comorbidities are consistently associated with worse outcomes. In our cohort, the mean age was 73.9 years, compared with 63 years in the study by Luhr et al. (Luhr et al. 2019) and 64 years in the analysis by Bauer et al. (Bauer et al. 2020). Second, variability in administrative coding accuracy across studies and healthcare systems may substantially affect mortality estimates derived from ICD-based datasets. Previous investigations have reported low sensitivity, but high specificity of sepsis-related ICD codes, with wide variability in predictive values potentially leading to under- or misclassification of cases (Liu et al. 2022). In line with this, a case review conducted in North America demonstrated that ICD codes on death certificates were modified in up to 93% of cases, with changes in the underlying cause of death occurring in 60%, highlighting the intrinsic instability of administrative mortality attribution (McGivern et al. 2017). In addition, temporal and contextual factors not fully captured in earlier analyses, including local resource availability and the impact of the SARS-CoV-2 pandemic, may have contributed to higher mortality rates observed in more recent years. Finally, the progressive global increase in antimicrobial resistance represents an additional concern, with potential implications for sepsis-related mortality in the coming decades (GBD 2021 Antimicrobial Resistance Collaborators). In this context, it is also plausible that the transition from Sepsis-1 to Sepsis-3 definitions contributed to the observed increase in sepsis and septic shock incidence and mortality, likely reflecting enhanced sensitivity to organ dysfunction and improved recognition of more severe cases (Singer et al. 2016).

This study found significant differences in IHM based on the underlying etiology of sepsis, in line with previous reports highlighting a strong association between infection source and clinical outcomes (Guarino et al. 2022, Leligdowicz and Matthay 2019, Rhee et al. 2019, Khandelwal et al. 2016). Research has consistently shown that different infectious sources of sepsis lead to varying mortality risks. Respiratory infections, particularly pneumonia, have been identified as one of the leading causes of sepsis and septic shock, and they were associated with higher mortality rates compared to other etiologies (Leligdowicz and Matthay 2019). The prevalence of respiratory etiology is notable, likely reflecting the impact of respiratory pathogens such as influenza and, more recently, COVID-19. Notably, the indeterminate etiology of sepsis, where no clear source of infection can be identified, was also strongly associated with IHM. This could be due to delayed or inappropriate treatments, as clinicians may have less information to tailor specific antimicrobial therapies (Rhee et al. 2019, Khandelwal et al. 2016). Alternatively, an indeterminate etiology may also reflect an ultra-acute and severe clinical presentation, in which rapid deterioration or early mortality precludes a complete diagnostic work-up, rather than a true diagnostic failure (Guarino et al. 2023, Rhee et al. 2019). Other studies also emphasized that etiologies, such as urinary tract infections or abdominal sepsis, were generally less fatal than respiratory or indeterminate sources, likely due to more effective diagnosis and available treatment options for these infections (Leligdowicz and Matthay 2019, Rhee et al. 2019, Khandelwal et al. 2016, Kramarow 2021). In patients under 65 years of age, respiratory etiology was the most significant contributor to fatal outcomes, while in older patients (those between 65 and 84 years, and those ≥ 85 years), both respiratory and indeterminate etiologies were associated with increased IHM. This age-related variation in the impact of sepsis etiology is consistent with the idea that older adults may be more susceptible to atypical presentations of sepsis, making timely diagnosis and treatment challenging (Ibarz et al. 2024, Luhr et al. 2019, Bauer et al. 2020, Liu et al. 2022, McGivern et al. 2017, GBD 2021 Antimicrobial Resistance Collaborators, Guarino et al. 2022, Leligdowicz and Matthay 2019, Rhee et al. 2019, Khandelwal et al. 2016, Kramarow 2021, Fleischmann-Struzek et al. 2020).

The increase in LOS suggested a growing burden of sepsis on healthcare system, as prolonged hospitalizations are inherently resource-intensive and costly. Patients with sepsis require longer hospitalizations due to the complexity of their care, which often involves intensive monitoring, aggressive antimicrobial therapy, and management of organ dysfunction (Bauer et al. 2020, Fleischmann-Struzek et al. 2020). The correlation between prolonged LOS and increased mortality further underscores the importance of early identification and rapid intervention in sepsis management. Efforts to reduce LOS and improve patient outcomes may depend on the development of more effective sepsis protocols, early detection technologies, and post-acute care strategies (Fleischmann-Struzek et al. 2020).

The COVID-19 pandemic has profoundly affected healthcare systems worldwide, and its impact on sepsis outcomes was no exception (Shappell et al. 2023, Shappell et al. 2023). The findings of this study showing that both IHM and 21-day mortality increased during the SARS-CoV-2 pandemic is in line with the emerging literature on the detrimental effects of COVID-19 on patients with sepsis (Shappell et al. 2023, Shappell et al. 2023, Heubner et al. 2022). This excess mortality likely reflects both direct and indirect effects of the pandemic. Direct mechanisms include the development of severe viral sepsis and secondary infections in patients with SARS-CoV-2, whereas indirect effects can be related to healthcare system overload, delayed hospital presentation, reduced ICU availability, and reallocation of medical resources (Heubner et al. 2022, Li et al. 2020). COVID-19 may also contribute to secondary bacterial or fungal infections in critically ill patients, further increasing the risk of sepsis and death (Grasselli et al. 2021, Ripa et al. 2021, Bardi et al. 2021, Zuniga-Moya et al. 2024). The pandemic highlighted disparities in healthcare access and outcomes, as vulnerable populations, including the elderly and those with preexisting conditions, have been disproportionately affected by both COVID-19 and sepsis (Grasselli et al. 2021).

This study has several limitations. A major limitation is the reliance on administrative ICD-9CM codes to identify sepsis and septic shock cases. The administrative nature of the dataset also precluded adjustment for important clinical confounders, including comorbidities, disease severity scores, and organ support therapies. ICU admission was not included as a covariate, as it may represent an intermediate outcome rather than a true confounder in the relationship between sepsis severity and mortality.

In addition, the retrospective design of our study may introduce biases in data collection and interpretation. Given that data were collected over a 20-year period, changes in clinical practices, diagnostic criteria, and treatment protocols may have influenced the observed outcomes. In particular, the evolution of sepsis definitions from Sepsis-1 to Sepsis-3 may have affected case identification and reporting over time. Although administrative ICD-9-CM coding was consistently used throughout the study period, changes in clinical awareness, diagnostic thresholds, and coding practices may have also contributed to an apparent increase in sepsis incidence and severity. This potential classification bias is inherent to long-term retrospective studies and should be considered when interpreting temporal trends.

Furthermore, this was a single-center study, which may limit the generalizability of the findings to other regions or healthcare systems with different patient populations, organizational models, and treatment strategies. Finally, the retrospective nature of the study does not allow disentangling the relative contribution of COVID-19-related viral sepsis from indirect pandemic-related factors, such as delayed access to care or resource constraints, on mortality and length of stay.

Despite these limitations, the study also has several strengths. Indeed, the single-center, long-term design allowed for a continuous and homogeneous evaluation of sepsis incidence and outcomes over two decades within the same healthcare system. This approach minimizes inter-institutional variability related to coding practices, organizational models, and referral pathways, which often limits the interpretability of large multicenter or administrative datasets. Moreover, the extended observation period enabled the assessment of major temporal transitions, including changes in sepsis definitions, advances in critical care, and the impact of the SARS-CoV-2 pandemic. The large sample size involving over 15,000 sepsis patients provides robust statistical power, allowing for a comprehensive analysis of trends in incidence, mortality, and LOS. The inclusion of a detailed analysis of sepsis etiologies and their correlation with IHM across different age groups further added significance to our study, offering insights into how age and infection type influence patients’ outcomes. These findings may help inform future strategies aimed at reducing sepsis-related mortality across different patient subgroups.