Your privacy, your choice

We use essential cookies to make sure the site can function. We also use optional cookies for advertising, personalisation of content, usage analysis, and social media.

By accepting optional cookies, you consent to the processing of your personal data - including transfers to third parties. Some third parties are outside of the European Economic Area, with varying standards of data protection.

See our privacy policy for more information on the use of your personal data.

for further information and to change your choices.
Skip to main content
Download PDF

Postoperative sepsis and its sequential impact on dementia

Critical Care volume 29, Article number: 66 (2025) Cite this article

Abstract

Background

Postoperative sepsis is a severe complication associated with increased mortality and potential long-term cognitive decline, including dementia. However, the relationship between postoperative sepsis and dementia remains poorly understood.

Methods

This retrospective cohort study used data from the National Database in Taiwan, covering the period from January 1, 2005, to December 31, 2022. The index period for surgeries was set between January 1, 2008, and December 31, 2013, allowing the identification of patients without prior dementia. A landmark period of 12 months following surgery was defined to capture the number of postoperative sepsis events, which were then analyzed for their impact on dementia risk. After 1:4 propensity score matching (PSM), dementia and mortality were evaluated using Cox proportional hazards and Fine-Gray competing risk models.

Results

Following PSM, 778 patients were in the postoperative sepsis group and 3,112 in the non-postoperative sepsis group. Dementia incidence was higher in the postoperative sepsis group (26%) compared to the non- postoperative sepsis group (13.6%), with a hazard ratio (HR) of 1.25 (95% CI, 1.03–1.52). A dose–response relationship was observed, with dementia rates of 24.5% for one postoperative sepsis event and 34.9% for two or more events, the latter showing an HR of 1.77 (95% CI, 1.17–2.66). Mortality was also elevated in the postoperative sepsis group (40.5% vs. 31.6%; HR 1.45, 95% CI, 1.28–1.65).

Conclusions

Postoperative sepsis is significantly associated with increased dementia risk in a dose-dependent manner. These findings highlight the importance of enhancing perioperative infection control to reduce both immediate and long-term cognitive complications.

Introduction

Dementia is a progressive neurodegenerative disease that poses significant public health and economic challenges. In 2019, approximately 57.4 million people were living with dementia globally—a number projected to reach 152.8 million by 2050 due to aging populations [1]. The financial burden is equally significant, with global annual costs estimated at $1.31 trillion, or $23,796 per person [2]. This burden is expected to rise and will place immense strain on healthcare systems, public infrastructure, and families who shoulder the emotional and financial responsibility of long-term care [3]. As dementia prevalence increases, there will be a critical need for enhanced healthcare resources and support systems [1, 2]. Effective preventive measures and early interventions are essential to mitigate these challenges.

Postoperative sepsis is a severe complication characterized by a dysregulated inflammatory response to infection, which can result in systemic organ dysfunction, increased morbidity, and a higher risk of mortality [4,5,6]. Preventive strategies include perioperative infection control through prophylactic antibiotics, sterile techniques, and timely recognition and management of infections [7, 8]. Optimizing glycemic control, maintaining proper nutrition, and effectively managing chronic diseases can further reduce the risk of developing sepsis [7, 8]. Beyond immediate complications, sepsis significantly raises the risk of long-term cognitive decline, including dementia [9,10,11,12]. Previous studies have linked sepsis to an increased dementia risk in general populations, but limited research has specifically examined this relationship in the context of postoperative settings [9,10,11,12]. Mechanisms involve systemic inflammation crossing the blood–brain barrier (BBB), inducing neuroinflammation and contributing to neurodegeneration [13,14,15]. Sepsis-induced BBB dysfunction leads to ischemic damage and microvascular injury, exacerbating neuronal loss [16,17,18,19,20]. Oxidative stress and mitochondrial dysfunction further impair neuronal survival, accelerating cognitive decline [16,17,18,19,20,21,22,23,24]. Understanding this association within postoperative populations is critical, as effective perioperative management of sepsis may reduce both immediate morbidity and the long-term burden of dementia.

We initiated this real-world database study to investigate the potential association between postoperative sepsis and dementia, as no clinical research to date has confirmed this relationship. Although prior observational studies, such as those by Janbek et al. and Lei et al., have explored the link between infections and dementia, they did not focus on the postoperative period—a critical window for intervention [11, 12]. Randomized controlled trials (RCTs) are limited in addressing long-term outcomes and complex multifactorial conditions like dementia [25]. By leveraging data from the national health insurance research database (NHIRD), we can analyze large, diverse patient populations with extended follow-up, allowing us to capture the long-term cognitive impacts of postoperative sepsis. If a link is established, this could lead to new preventive strategies that reduce both short-term complications and long-term dementia risk, offering significant clinical and public health benefits.

Patients and methods

We conducted a retrospective cohort study utilizing longitudinal data from the NHIRD covering the period from January 1, 2005, to December 31, 2022.

Data sources

We conducted a retrospective cohort study utilizing a statewide longitudinal dataset linked to the NHIRD. This population-based cohort study of patients with hospitalized postoperative sepsis utilized data from the NHIRD, a comprehensive repository that includes disease diagnoses, procedural details, medication prescriptions, demographics, and beneficiary enrollment profiles.[26,27,28,29] The database uses encrypted patient identifiers to ensure confidentiality and integrates with Taiwan’s Death Registry, providing precise determination of vital status and cause of death. This integration enhances the rigor and credibility of the study [26,27,28,29]. The study protocols were approved by the Institutional Review Board. For each event, one primary diagnosis and up to 50 secondary diagnoses were coded using the International Classification of Diseases, 9th or 10th Revision (ICD-9 or ICD-10). All patient data were linked to the NHIRD and the Registry of Births and Deaths, providing comprehensive data up to December 31, 2022, for analysis.

Study design and sample

The index period was defined as January 1, 2008, to December 31, 2013, allowing sufficient time to identify patients without prior dementia diagnoses and calculate comorbidities. This period also provided adequate follow-ups for all individuals. We first compared the incidence of outcomes (dementia or death) between postoperative sepsis and non-postoperative sepsis groups. Notably, during the index period, 666,740 patients underwent major operations. Several exclusions were applied to refine the cohort: 438,589 patients were excluded due to surgeries outside the index period, 52,220 due to missing anesthesia data, and 7,825 due to dementia before the index period. For the case group, patients who had dementia or sepsis prior to 2008 were excluded, and in the control group, those who developed sepsis before 2014 (942 patients) were excluded to ensure the integrity of the comparison. Additionally, 92 patients with two surgeries on the same day, 21,035 patients under the age of 18, and those with dementia or sepsis prior to the index period were excluded. Further exclusions included patients with incomplete data on income or urbanization, as well as those with less than one year of follow-up. After these exclusions, 1,087 patients with postoperative sepsis remained, along with 108,628 patients without postoperative sepsis. Propensity score matching (PSM) was then performed to balance the postoperative sepsis and non-postoperative sepsis groups.

Postoperative sepsis was identified using the ICD-9 code 998.59 for other postoperative infections, as well as 995.91 for sepsis and 995.92 for severe sepsis. Additional septicemia codes, such as 038.x, were also utilized to capture specific types of septicemia. In the ICD-10 system, postoperative sepsis was defined using codes T81.4XXA for infection following a procedure and A41.x for various forms of sepsis, including A41.9 for unspecified sepsis. Severe sepsis was categorized under R65.20 for cases without septic shock and R65.21 for those with septic shock. The specific code T81.44XA was used to identify sepsis following a procedure in the initial encounter.

Outcomes

The primary outcomes were newly diagnosed dementia and mortality. The onset of dementia was determined by the first recorded diagnosis in the database, using ICD-10 codes detailed in Supplemental Table 1. In Taiwan, dementia diagnoses are subject to stringent legal and clinical criteria due to their implications for major illness certification and associated benefits. These requirements enhance the reliability and validity of diagnostic data. Similarly, postoperative sepsis diagnoses are derived from ICU records and require documented clinical criteria, such as systemic inflammatory response, infection, and organ dysfunction, which are often corroborated by antibiotic use. These safeguards minimize the risk of misclassification. For mortality, the event time was recorded as the date of death, as documented in the linked dataset.

Statistical analysis

Baseline characteristics for the postoperative sepsis and non-postoperative sepsis groups were summarized using descriptive statistics. Linear relationships were assumed between continuous covariates and study outcomes. For mortality, patient follow-up began at the index event and continued until death. For dementia, follow-up was from the index event until dementia onset or death, whichever occurred first, with censoring at 62 months for event-free patients.

We first compared the incidence of death or dementia between the postoperative sepsis and non-postoperative sepsis groups. A landmarking approach was applied to assess a potential dose–response relationship between the number of postoperative sepsis and the outcomes. The landmark period was defined as the first 12 months of follow-up, during which the number of postoperative sepsis was recorded and analyzed for association with subsequent event rates. Postoperative sepsis events were treated as categorical variables (0, 1, or ≥ 2 events) in the full cohort, and as continuous variables in the postoperative sepsis group alone. Only patients event-free within the landmark period were included in the dose–response analysis, with the number of hospital events within the landmark period included as a covariate.

Different models were applied for each outcome. Cox proportional hazards models were used to assess mortality, presenting results as hazard ratios (HRs). For dementia, the Fine-Gray subdistribution hazard model was utilized to account for competing risks, particularly the risk of death, with results expressed as subdistribution hazard ratios (sHRs). Given the strong link between postoperative sepsis and mortality, the Fine-Gray model provided more accurate risk estimates for dementia. Cox regression models were constructed to account for various confounding factors. Model 1 was a univariable analysis. Model 2 was adjusted for age and sex. Model 3 included adjustments for age, sex, income levels, and urbanization. Model 4 further adjusted for type of surgery, surgical urgency, ASA physical status classification, types of anesthesia, first surgery, number of surgeries, and coexisting comorbidities, in addition to age, sex, income levels, and urbanization. Model 5 extended the adjustments to include patient habitus, concurrent medication use, and the Charlson Comorbidity Index, along with all variables from Model 4.

All analyses were conducted in the full matched cohort. Results were presented with 95% confidence intervals (CIs). The number of patients at risk, dementia events, person-years of follow-up, and incidence rates per 10,000 person-years were calculated for each group. Statistical analyses were performed using SAS software (version 9.4; SAS Institute, Cary, NC, USA).

Results

Baseline characteristics

Baseline characteristics of the study population are presented in Table 1. After 1:4 PSM, a total of 3,112 patients were classified as the non-postoperative sepsis group, and 778 patients were classified as the postoperative sepsis group. The mean age was similar between the two groups, with 55.99 ± 17.04 years in the non-postoperative sepsis group and 56.95 ± 17.73 years in the postoperative sepsis group (ASMD = 0.056). The median age in both groups was also comparable, with an interquartile range of 58.00 (43.00–70.00) in the non-postoperative sepsis group and 59.00 (44.00–71.00) in the postoperative sepsis group. The distribution of patients across different age groups, sex, and income levels showed minimal differences between the two groups, with all absolute standardized mean differences (ASMDs) below 0.1, indicating a well-matched cohort. The proportion of patients in different surgical categories, such as musculoskeletal, cardiovascular, and digestive surgeries, was similar across both groups, although patients in the postoperative sepsis group had a slightly higher incidence of multi-surgery events (13.0% vs. 12.0%, ASMD = 0.087). Other variables, including surgical urgency, ASA physical status classification, types of anesthesia, and the number of surgeries, were well-balanced between the two groups. Comorbidities such as diabetes, hypertension, coronary artery disease, and hyperlipidemia were similarly distributed between the groups. Notably, there was no significant difference in the use of concurrent medications, including statins and aspirin, between the non-postoperative and postoperative sepsis groups.

Table 1 Baseline characteristics of the study population after 1:4 propensity score matching for non-postoperative and postoperative sepsis
Full size table

Postoperative sepsis: association with dementia, mortality, and impact of event frequency

In Supplemental Table 2, the association between postoperative sepsis and the incidence of dementia and all-cause mortality was analyzed. A significantly higher incidence of dementia was observed in patients with postoperative sepsis compared to those without postoperative sepsis (26.0% vs. 13.6%, p < 0.0001). Similarly, the incidence of all-cause mortality was significantly higher in the postoperative sepsis group compared to the non-postoperative sepsis group (40.5% vs. 31.6%, p < 0.0001). Notably, the dementia risk in the postoperative sepsis group may be underestimated due to the high mortality rate in this group, meaning many patients may have died before dementia could be diagnosed. Supplemental Table 3 demonstrates the incidence of mortality and dementia based on the frequency of postoperative sepsis events within the first 12 months of follow-up. The incidence of dementia increased with the frequency of postoperative sepsis events: 13.6% in patients without sepsis, 24.5% in patients with one sepsis event, and 34.9% in those with two or more sepsis events (p < 0.0001). A similar pattern was observed for all-cause mortality, with rates of 31.6%, 37.8%, and 56.9% for patients with 0, 1, and ≥ 2 sepsis events, respectively (p < 0.0001).

Table 2 Cox regression models evaluating dementia and competing risk of mortality in patients with postoperative sepsis during the first 12 months of follow-up
Full size table
Table 3 Cox regression models evaluating dementia and competing risk of mortality based on frequency of postoperative sepsis events during the first 12 months of follow-up
Full size table

Cox regression models for dementia and mortality in postoperative sepsis patients

Table 2 presents the Cox regression models assessing dementia and risk of mortality in patients with postoperative sepsis during the first 12 months of follow-up. In the univariable analysis (Model 1), postoperative sepsis was significantly associated with an increased dementia risk (HR, 1.20; 95% CI, 1.09–1.45; p = 0.0148). After adjusting for age and sex (Model 2), the HR increased to 1.25 (95% CI, 1.03–1.52; p = 0.0216). Further adjustments for income levels and urbanization (Model 3) yielded a similar HR of 1.26 (95% CI, 1.04–1.52; p = 0.0203). Model 4, which adjusted for additional surgical and clinical factors, showed a slightly higher HR of 1.27 (95% CI, 1.05–1.54; p = 0.0157). In the fully adjusted model (Model 5), accounting for patient habitus, medication use, and comorbidities, the HR was 1.24 (95% CI, 1.02–1.51; p = 0.0286). The similarity in adjusted hazard ratios (aHRs) across Models 1 to 5 suggests that the PSM was effective in balancing confounders.

For mortality, postoperative sepsis was associated with a significantly increased risk across all models. In Model 1, the HR was 1.45 (95% CI, 1.28–1.65; p < 0.0001), which remained robust in the fully adjusted Model 5 (HR, 1.57; 95% CI, 1.38–1.79; p < 0.0001). To account for the competing risk of death, a Fine-Gray subdistribution hazard model was used for the dementia outcome, which provided more accurate estimates by adjusting for the high risk of mortality in postoperative sepsis patients. This competing risk analysis confirmed that postoperative sepsis significantly increased the risk of both mortality and dementia, with death occurring before dementia diagnosis in many cases.

Dementia and risk of mortality by frequency of postoperative sepsis events

Table 3 shows the Cox regression models evaluating dementia risk and the competing risk of mortality based on the frequency of postoperative sepsis events during the first 12 months of follow-up. In Model 1, patients with one sepsis event had a higher dementia risk (HR, 1.12; 95% CI, 1.02–1.38; p = 0.0240), which increased further for two or more events (HR, 1.77; 95% CI, 1.17–2.66; p = 0.0065). The competing risk analysis showed a similar pattern, with one event (HR, 1.20; 95% CI, 1.05–1.52; p = 0.0248) and two or more events (HR, 2.34; 95% CI, 1.54–3.57; p < 0.0001) associated with increased dementia risk. In the fully adjusted model, one sepsis event remained significantly associated with dementia (HR, 1.17; 95% CI, 1.04–1.44; p = 0.0241), and the risk remained higher for two or more events (HR, 1.76; 95% CI, 1.15–2.68; p = 0.0087). The competing risk analysis confirmed these results. For mortality, one sepsis event increased risk (HR, 1.45; 95% CI, 1.26–1.67; p < 0.0001), with a greater risk for two or more events (HR, 2.35; 95% CI, 1.80–3.07; p < 0.0001). The frequency of postoperative sepsis events was consistently associated with higher dementia and risk of mortality across all models and analyses.

Postoperative sepsis events and dementia incidence rates

Table 4 shows the incidence rates and IRRs for dementia and mortality by the frequency of postoperative sepsis events during the first 12 months of follow-up. Patients with postoperative sepsis had higher rates of both dementia and mortality compared to those without sepsis. For dementia, the IRR was 1.75 (95% CI, 1.14–2.62) for two or more events and 1.23 (95% CI, 1.02–1.50) for one event. The Kaplan–Meier analyses (Figs. 1 and 2) demonstrate that postoperative sepsis significantly increases dementia risk in surgical patients. Patients with postoperative sepsis exhibit a higher cumulative incidence of dementia compared to those without sepsis (Fig. 1), with the risk increasing further in patients experiencing multiple sepsis events (Fig. 2). For mortality, the IRR was 2.44 (95% CI, 1.89–3.15) for two or more events and 1.47 (95% CI, 1.26–1.72) for one event. A clear dose–response relationship is evident, as patients with two or more sepsis events have a markedly higher dementia risk and lower overall survival than those with one or no events. The survival curves (Supplementary Figs. 1 and 2) show early and sustained divergence, emphasizing the strong association between recurrent postoperative sepsis and worse survival outcomes. These findings underscore the cumulative impact of sepsis events on long-term cognitive decline and mortality.

Table 4 Postoperative sepsis events and dementia incidence rates and incidence rate ratios among surgical patients
Full size table
Fig. 1
figure 1

Kaplan–Meier analysis of dementia risk in surgical patients with and without postoperative sepsis

Full size image
Fig. 2
figure 2

Kaplan–Meier analysis of dementia risk in surgical patients by frequency of postoperative sepsis events in the first 12 months of follow-up

Full size image

Discussion

Our study offers new insights into the association between postoperative sepsis and long-term outcomes, specifically the increased risks of dementia and mortality in surgical patients. Postoperative sepsis is a significant complication that not only elevates the risk of death but also substantially increases the dementia risk in survivors. We found that patients who experienced postoperative sepsis had a dementia incidence of 26%, compared to 13.6% in patients without postoperative sepsis. Furthermore, the frequency of postoperative sepsis events was associated with a dose–response effect on dementia risk, with patients experiencing two or more sepsis events showing a markedly higher incidence of dementia (34.9%) compared to those with no sepsis events (13.6%). These findings underscore the clinical importance of preventing postoperative sepsis, as it contributes not only to short-term mortality but also to long-term cognitive decline. The results of this study align with previous research demonstrating the severe consequences of sepsis [9,10,11,12], but this is one of the first studies to show a clear association between postoperative sepsis and dementia in a large, population-based cohort. The use of PSM strengthens the validity of our results by minimizing confounding factors.

Our findings highlight the need for enhanced perioperative infection control strategies to mitigate the risk of postoperative sepsis and its associated long-term outcomes. These strategies include strict adherence to sterile surgical techniques, optimized prophylactic antibiotic use, and early postoperative infection detection and management [30]. Additionally, identifying patients at higher risk of sepsis preoperatively—through comprehensive assessments of comorbidities and immune function—could facilitate targeted interventions [31]. Monitoring sepsis survivors for cognitive decline during follow-up may also enable early dementia detection and intervention [32]. These approaches have the potential to significantly improve patient outcomes and reduce healthcare burdens.

In our study, the index period was critical to ensure that all patients were dementia-free at the start of follow-up, allowing us to observe the onset of dementia. By setting a six-year index period, we excluded patients with prior dementia and assessed comorbidities, providing a clearer analysis of postoperative sepsis and dementia risk. However, competing risks like death remained a concern, as patients could die before developing dementia. To address this, we introduced a landmark period of the first 12 months post-surgery to examine the relationship between postoperative sepsis frequency and dementia. This approach reduced the impact of early deaths and allowed for a more detailed analysis of recurrent postoperative sepsis. The Fine-Gray subdistribution hazard model accounted for death as a competing risk, providing a more accurate dementia risk estimation. These methods minimized confounding and offered important insights into the complex relationship between postoperative sepsis and dementia. The study design follows the approach of the BMJ study on delirium, which identified it as a strong risk factor for both mortality and incident dementia in older adult patients [33]. This design allows for a more robust analysis of the relationship between postoperative sepsis and dementia. The data presented in this study supports a causal interpretation of the association between postoperative sepsis and subsequent dementia, providing valuable insights into the long-term cognitive consequences of sepsis following surgery [33].

In comparing our study with previous literature, it is essential to note that most of the prior research on sepsis and dementia focused on general sepsis rather than specifically on postoperative sepsis [9, 10]. For example, the study by Janbek et al., which included approximately 1.5 million individuals, found an increased dementia risk after any infection, but it did not specifically address postoperative infections [11]. The increased incidence of dementia reported (IRR 1.49) was based on all types of infections, including respiratory, urinary tract, and other non-surgical infections [11]. This differs from our study, which uniquely focuses on the relationship between postoperative sepsis and dementia, a condition directly linked to surgical procedures and potentially preventable through enhanced perioperative care [4,5,6]. Furthermore, while previous meta-analyses (such as the one involving 891,562 individuals) have shown that survivors of severe sepsis face a heightened dementia risk, these studies often included a mix of infection sources, many of which were not related to surgical interventions [12]. Our study stands apart by specifically targeting postoperative sepsis, offering more actionable insights for surgical and perioperative management. The distinction between general sepsis and postoperative sepsis is critical because postoperative sepsis arises directly from surgical procedures and is more amenable to prevention strategies, such as better infection control during and after surgery [4,5,6]. This contrasts with general sepsis, where the infection source is often unclear and harder to manage preventively [34]. While prior studies align with our findings regarding the association between sepsis and increased dementia risk [9,10,11,12], our study is the first to specifically examine postoperative sepsis and demonstrate a dose-dependent relationship using a real-world national database, offering more targeted implications for clinical practice, particularly in surgical settings.

The relationship between postoperative sepsis and an increased dementia risk is multifaceted, involving both clinical and pathophysiological mechanisms [35, 36]. Although sepsis has been consistently associated with a higher dementia risk, the exact cause of this association remains debated. One prevailing theory, the ‘infectious hypothesis,’ suggests that infections act as a direct risk factor for dementia by triggering neuroinflammatory responses [35, 36]. Another view, the ‘infectious accelerant theory,’ posits that infections may exacerbate pre-existing dementia pathology, hastening its progression. The possibility of reverse causation—wherein preclinical dementia increases a patient’s susceptibility to sepsis—has also been proposed [35, 36]. A national cohort study by Richmond-Rakerd et al., which followed 1.7 million people over 30 years, found a consistent increase in dementia risk across various time intervals, ruling out reverse causality and suggesting a genuine relationship between infection and dementia [37]. Further prospectiare required to definitively distinguish between these hypotheses, though they may not be mutually exclusive. The underlying mechanisms linking sepsis and dementia remain under investigation, but evidence points toward systemic inflammation as a key driver [38, 39]. The inflammatory response in sepsis is not typically caused by direct brain infection but is instead a systemic immune response transmitted to the central nervous system through various pathways [40, 41]. Research by De Sousa et al. can induce long-term trained innate immune memory in the brain, making it more susceptible to neurotoxic insults like Aβ oligomers, which are linked to Alzheimer’s disease [42]. Other studies have suggested that sepsis interferes with critical cognitive processes, such as BDNF/TrkB signaling in the hippocampus, which is essential for memory formation [43]. Vascular mechanisms may also contribute to this relaystemic inflammation can impair blood–brain barrier (BBB) function, allowing neurotoxic substances and pathogens to enter the brain, resulting in neuroinflammation and neuronal damage [44]. Sepsis-related BBB dysfunction may also cause microbleeds and vascular injury [45], leading to ischemic damage, which has been observed in brain autopsies of patients who died from sepsis [46,47,48]. The occurrence of vascular brain injury, such as microvascular infarction, helps explain the cognitive impairment observed in sepsis survivors. Overall, the link between postoperative sepsis and dementia likely involves multiple mechanisms—including systemic inflammation, immune memory, neurovascular damage, and direct interference with cognitive processes—each contributing to the increased dementia risk in sepsis survivors. These mechanisms highlight the importance of further research to explore preventive strategies in the perioperative setting.

Our study has several strengths, particularly in its methodological design. We utilized a large population-based cohort and employed propensity score matching, which minimized confounding factors and enhanced the comparability between groups. The use of an index period allowed us to ensure patients were dementia-free at the study's start, and the landmark period (12 months) facilitated precise analysis of postoperative sepsis' dose–response relationship with long-term outcomes. Additionally, accounting for the competing risk of death through Fine-Gray models provided more accurate dementia risk estimates, especially in high-mortality groups. This comprehensive design improves the reliability of our findings on the serious long-term consequences of postoperative sepsis, including increased dementia and risk of mortality. Importantly, our study is the first to use a real-world national database to establish a clear link between postoperative sepsis and the increased dementia risk, demonstrating a dose-dependent effect. While previous research has recognized sepsis as a contributor to long-term cognitive decline, our study uniquely focuses on the postoperative context. By utilizing a large, population-based dataset, we were able to assess both the frequency of postoperative sepsis and its impact on dementia risk. Our findings provide compelling new evidence that repeated sepsis events significantly elevate the risk of developing dementia, emphasizing the critical need for preventive measures in perioperative care.

Our study has several limitations. First, as a retrospective cohort study, residual confounding is possible despite using propensity score matching. Unmeasured factors such as variations in surgical care may influence outcomes. Second, reliance on administrative data from the NHIRD introduces the possibility of coding inaccuracies, potentially leading to misclassification of sepsis and dementia diagnoses. However, stringent diagnostic criteria for dementia and the use of ICU-based clinical indicators for postoperative sepsis in Taiwan help minimize this risk. Third, the study population was limited to Taiwan, potentially affecting generalizability to other populations. Additionally, lifestyle factors such as physical activity was not accounted for, which could introduce bias. Lastly, while competing risk models were employed to account for mortality, the high mortality in the sepsis group may have led to an underestimation of dementia incidence, as some patients may have died before a diagnosis could be made. Further prospective studies with longer follow-up are necessary to confirm these findings.

Conclusions

Our study demonstrates a significant association between postoperative sepsis and an increased dementia risk, with a dose-dependent effect. Patients who experienced multiple episodes of sepsis had a markedly higher incidence of dementia compared to those without sepsis. These findings emphasize the importance of improving perioperative infection control and sepsis prevention strategies to reduce both immediate and long-term cognitive risks.

Availability of data and materials

The datasets supporting the findings of this study were obtained from the National Health Insurance Research Database (NHIRD) and the Taiwan Cancer Registry database. Due to the Personal Information Protection Act in Taiwan (enforced since 2012), these datasets cannot be publicly shared in the manuscript, supplementary materials, or any public repository to protect participant confidentiality. Researchers interested in accessing the data may submit a formal application to the relevant ethics review committees in Taiwan. Detailed information on data access procedures is available at: https://dep.mohw.gov.tw/DOS/cp-5119-59201-113.html. Dr. Szu-Yuan Wu, MD, PhD, had full access to all the data in this study and takes full responsibility for the integrity and accuracy of the data analysis.

Abbreviations

NHIRD:

National health insurance research database

PSM:

Propensity score matching

HR:

Hazard ratio

sHR:

Subdistribution hazard ratio

CI:

Confidence interval

ICD-9/10:

International classification of diseases, 9th/10th revision

BBB:

Blood–brain barrier

AβO:

Amyloid-β oligomers

ASA:

American society of anesthesiologists

ASMD:

Absolute standardized mean difference

RCT:

Randomized controlled trial

CCI:

Charlson comorbidity index

References

  1. Nichols E, Steinmetz JD, Vollset SE, Fukutaki K, Chalek J, Abd-Allah F, Abdoli A, Abualhasan A, Abu-Gharbieh E, Akram TT, et al. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health. 2022;7(2):E105–25.

    Article  Google Scholar 

  2. Wimo A, Seeher K, Cataldi R, Cyhlarova E, Dielemann JL, Frisell O, Guerchet M, Jonsson L, Malaha AK, Nichols E, et al. The worldwide costs of dementia in 2019. Alzheimers Dement. 2023;19(7):2865–73.

    Article  PubMed  Google Scholar 

  3. Wang L, Zhou Y, Fang X, Qu G. Care burden on family caregivers of patients with dementia and affecting factors in China: a systematic review. Front Psychiatr. 2022;13:1004552.

    Article  Google Scholar 

  4. Berrios-Torres SI, Umscheid CA, Bratzler DW, Leas B, Stone EC, Kelz RR, Reinke CE, Morgan S, Solomkin JS, Mazuski JE, et al. Centers for disease control and prevention guideline for the prevention of surgical site infection, 2017. JAMA Surg. 2017;152(8):784–91.

    Article  PubMed  Google Scholar 

  5. Cohen NS, Bock JM, May AK. Sepsis and postoperative surgical site infections. Surgery. 2023;174(2):403–5.

    Article  PubMed  Google Scholar 

  6. Plaeke P, De Man JG, Coenen S, Jorens PG, De Winter BY, Hubens G. Clinical- and surgery-specific risk factors for post-operative sepsis: a systematic review and meta-analysis of over 30 million patients. Surg Today. 2020;50(5):427–39.

    Article  PubMed  Google Scholar 

  7. Wang Q, Cao M, Tao H, Fei Z, Huang X, Liang P, Liu B, Liu J, Lu X, Ma P, et al. Evidence-based guideline for the prevention and management of perioperative infection. J Evid Based Med. 2023;16(1):50–67.

    Article  CAS  PubMed  Google Scholar 

  8. Dhole S, Mahakalkar C, Kshirsagar S, Bhargava A. Antibiotic prophylaxis in surgery: current insights and future directions for surgical site infection prevention. Cureus. 2023;15(10): e47858.

    PubMed  PubMed Central  Google Scholar 

  9. Muzambi R, Bhaskaran K, Smeeth L, Brayne C, Chaturvedi N, Warren-Gash C. Assessment of common infections and incident dementia using UK primary and secondary care data: a historical cohort study. Lancet Healthy Longev. 2021;2(7):e426–35.

    Article  PubMed  PubMed Central  Google Scholar 

  10. van Ton AMP, Meijer-van Leijsen EM, Bergkamp MI, Bronkhorst EM, Pickkers P, De Leeuw FE, Tuladhar AM, Abdo WF. Dementia risk and structural brain changes following nonneurological infections during 9-year follow-up. Crit Care Med. 2022;50(4):554–64.

    Article  Google Scholar 

  11. Janbek J, Laursen TM, Frimodt-Møller N, Magyari M, Haas JG, Lathe R, Waldemar G. Hospital-diagnosed infections, autoimmune diseases, and subsequent dementia incidence. JAMA Netw Open. 2023;6(9): e2332635.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Lei S, Li X, Zhao H, Feng Z, Chun L, Xie Y, Li J. Dementia risk or cognitive impairment in sepsis survivals: a systematic review and meta-analysis. Front Aging Neurosci. 2022;14: 839472.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Raz L, Knoefel J, Bhaskar K. The neuropathology and cerebrovascular mechanisms of dementia. J Cereb Blood Flow Metab. 2016;36(1):172–86.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Xie J, Van Hoecke L, Vandenbroucke RE. The impact of systemic inflammation on alzheimer’s disease pathology. Front Immunol. 2021;12: 796867.

    Article  CAS  PubMed  Google Scholar 

  15. Ahmad MA, Kareem O, Khushtar M, Akbar M, Haque MR, Iqubal A, Haider MF, Pottoo FH, Abdulla FS, Al-Haidar MB, et al. Neuroinflammation: a potential risk for dementia. Int J Mol Sci. 2022;23(2):616.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Iwashyna TJ, Ely EW, Smith DM, Langa KM. Long-term cognitive impairment and functional disability among survivors of severe sepsis. JAMA. 2010;304(16):1787–94.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Muzambi R, Bhaskaran K, Brayne C, Davidson JA, Smeeth L, Warren-Gash C. Common bacterial infections and dementia risk or cognitive decline: a systematic review. J Alzheimers Dis. 2020;76(4):1609–26.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Asby D, Boche D, Allan S, Love S, Miners JS. Systemic infection exacerbates cerebrovascular dysfunction in Alzheimer’s disease. Brain. 2021;144(6):1869–83.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Basak JM, Ferreiro A, Cohen LS, Sheehan PW, Nadarajah CJ, Kanan MF, Sukhum KV, Dantas G, Musiek ES. Bacterial sepsis increases hippocampal fibrillar amyloid plaque load and neuroinflammation in a mouse model of Alzheimer’s disease. Neurobiol Dis. 2021;152: 105292.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, Bellomo R, Bernard GR, Chiche JD, Coopersmith CM, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA. 2016;315(8):801–10.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Widmann CN, Heneka MT. Long-term cerebral consequences of sepsis. Lancet Neurol. 2014;13(6):630–6.

    Article  PubMed  Google Scholar 

  22. Semmler A, Widmann CN, Okulla T, Urbach H, Kaiser M, Widman G, Mormann F, Weide J, Fliessbach K, Hoeft A, et al. Persistent cognitive impairment, hippocampal atrophy and EEG changes in sepsis survivors. J Neurol Neurosurg Psychiatr. 2013;84(1):62–9.

    Article  Google Scholar 

  23. Fritze T, Doblhammer G, Widmann CN, Heneka MT. Time course of dementia following sepsis in German health claims data. Neurol Neuroimmunol Neuroinflamm. 2021. https://doi.org/10.1212/NXI.0000000000000911.

    Article  PubMed  Google Scholar 

  24. Kao LT, Sheu JJ, Lin HC, Tsai MC, Chung SD. Association between sepsis and dementia. J Clin Neurosci. 2015;22(9):1430–3.

    Article  PubMed  Google Scholar 

  25. Fernainy P, Cohen AA, Murray E, Losina E, Lamontagne F, Sourial N. Rethinking the pros and cons of randomized controlled trials and observational studies in the era of big data and advanced methods: a panel discussion. BMC Proc. 2024;18(Suppl 2):1.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Sun M, Chen WM, Wu SY, Zhang J. Dementia risk after major elective surgery based on the route of anaesthesia: a propensity score-matched population-based cohort study. EClinicalMedicine. 2023;55: 101727.

    Article  PubMed  Google Scholar 

  27. Sun M, Chen WM, Wu SY, Zhang J. Dementia risk amongst older adults with hip fracture receiving general anaesthesia or regional anaesthesia: a propensity-score-matched population-based cohort study. Br J Anaesth. 2023;130(3):305–13.

    Article  CAS  PubMed  Google Scholar 

  28. Sun M, Chen WM, Wu SY, Zhang J. Long-term opioid use and dementia risk in patients with chronic pain. J Am Med Dir Assoc. 2023;24(9):1420-1426.e1422.

    Article  PubMed  Google Scholar 

  29. Sun M, Chen WM, Wu SY, Zhang J. Metformin in elderly type 2 diabetes mellitus: dose-dependent dementia risk reduction. Brain. 2023;147:1474–82.

    Article  Google Scholar 

  30. Rosa R, Sposato K, Abbo LM. Preventing surgical site infections: implementing strategies throughout the perioperative continuum. AORN J. 2023;117(5):300–11.

    Article  PubMed  Google Scholar 

  31. Osagwu M, Okobi E, Ekor E. Postoperative infections: risk factors, prevention, and management strategies. Int J Med Sci Dent Res. 2024;7:52–76.

    Google Scholar 

  32. Li Y, Ji M, Yang J. Current understanding of long-term cognitive impairment after sepsis. Front Immunol. 2022;13: 855006.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Gordon EH, Ward DD, Xiong H, Berkovsky S, Hubbard RE. Delirium and incident dementia in hospital patients in New South Wales, Australia: retrospective cohort study. BMJ. 2024;384: e077634.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Ling L, Mui OOY, Laupland KB, Lefrant JY, Roberts JA, Gopalan PD, Lipman J, Joynt GM. CaBo OMi: scoping review on diagnostic criteria and investigative approach in sepsis of unknown origin in critically ill patients. J Intensive Care. 2022;10(1):44.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Itzhaki RF, Golde TE, Heneka MT, Readhead B. Do infections have a role in the pathogenesis of Alzheimer disease? Nat Rev Neurol. 2020;16(4):193–7.

    Article  PubMed  Google Scholar 

  36. Butler L, Walker KA. The role of chronic infection in Alzheimer’s disease: instigators, co-conspirators, or bystanders? Curr Clin Microbiol Rep. 2021;8(4):199–212.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Richmond-Rakerd LS, Iyer MT, D’Souza S, Khalifeh L, Caspi A, Moffitt TE, Milne BJ. Associations of hospital-treated infections with subsequent dementia: nationwide 30-year analysis. Nat Aging. 2024;4(6):783–90.

    Article  PubMed  Google Scholar 

  38. Sipilä PN, Heikkilä N, Lindbohm JV, Hakulinen C, Vahtera J, Elovainio M, Suominen S, Väänänen A, Koskinen A, Nyberg ST, et al. Hospital-treated infectious diseases and the dementia risk: a large, multicohort, observational study with a replication cohort. Lancet Infect Dis. 2021;21(11):1557–67.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Inoue Y, Shue F, Bu G, Kanekiyo T. Pathophysiology and probable etiology of cerebral small vessel disease in vascular dementia and Alzheimer’s disease. Molecular Neurodegener. 2023. https://doi.org/10.1186/s13024-023-00640-5.

    Article  Google Scholar 

  40. Manabe T, Heneka MT. Cerebral dysfunctions caused by sepsis during ageing. Nat Rev Immunol. 2022;22(7):444–58.

    Article  CAS  PubMed  Google Scholar 

  41. Wendeln A-C, Degenhardt K, Kaurani L, Gertig M, Ulas T, Jain G, Wagner J, Haesler LM, Wild K, Skodras A, et al. Innate immune memory in the brain shapes neurological disease hallmarks. Nature. 2018;556(7701):332–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. De Sousa VL, Araújo SB, Antonio LM, Silva-Queiroz M, Colodeti LC, Soares C, Barros-Aragão F, Mota-Araujo HP, Alves VS, Coutinho-Silva R, et al. Innate immune memory mediates increased susceptibility to Alzheimer’s disease-like pathology in sepsis surviving mice. Brain Behav Immun. 2021;95:287–98.

    Article  PubMed  Google Scholar 

  43. Hippensteel JA, Anderson BJ, Orfila JE, McMurtry SA, Dietz RM, Su G, Ford JA, Oshima K, Yang Y, Zhang F, et al. Circulating heparan sulfate fragments mediate septic cognitive dysfunction. J Clin Invest. 2019;129(4):1779–84.

    Article  PubMed  PubMed Central  Google Scholar 

  44. Ozben T, Ozben S. Neuro-inflammation and anti-inflammatory treatment options for Alzheimer’s disease. Clin Biochem. 2019;72:87–9.

    Article  CAS  PubMed  Google Scholar 

  45. Andjelkovic AV, Situ M, Citalan-Madrid AF, Stamatovic SM, Xiang J, Keep RF. Blood-brain barrier dysfunction in normal aging and neurodegeneration: mechanisms, impact, and treatments. Stroke. 2023;54(3):661–72.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Takeda S, Sato N, Morishita R. Systemic inflammation, blood-brain barrier vulnerability and cognitive/non-cognitive symptoms in Alzheimer disease: relevance to pathogenesis and therapy. Front Aging Neurosci. 2014;6:171.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Sweeney MD, Sagare AP, Zlokovic BV. Blood-brain barrier breakdown in Alzheimer disease and other neurodegenerative disorders. Nat Rev Neurol. 2018;14(3):133–50.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Ehlenbach WJ, Sonnen JA, Montine TJ, Larson EB. Association between sepsis and microvascular brain injury. Crit Care Med. 2019;47(11):1531–8.

    Article  PubMed  Google Scholar 

Download references

Funding

This work was supported by grants from the National Natural Science Foundation of China (NSFC: No. 82271209 awarded to Mingyang Sun, and Nos. 82471288 and 82271288 awarded to Jiaqiang Zhang). Additionally, Szu-Yuan Wu’s research was supported by the Lo-Hsu Medical Foundation, Lotung Poh-Ai Hospital (Grant Nos. 1303, 11304, 11403, and 11404).

Author information

Authors and Affiliations

  1. Department of Anesthesiology and Perioperative Medicine, People’s Hospital of Zhengzhou University, Henan Provincial People’s Hospital, 7 Weiwu Rd, Zhengzhou, Henan, China

    Mingyang Sun, Fangfang Li, Yangyang Wang, Mengrong Miao, Zhongyuan Lu & Jiaqiang Zhang

  2. Academy of Medical Sciences of Zhengzhou University, Zhengzhou, Henan, China

    Zhongyuan Lu

  3. Graduate Institute of Business Administration, College of Management, Fu Jen Catholic University, Taipei, Taiwan

    Wan-Ming Chen

  4. Artificial Intelligence Development Center, Fu Jen Catholic University, Taipei, Taiwan

    Wan-Ming Chen

  5. Big Data Center, Lo-Hsu Medical Foundation, Lotung Poh-Ai Hospital, No. 83, Nanchang St., Luodong Township, Yilan County, 265, Taiwan

    Szu-Yuan Wu

  6. Division of Radiation Oncology, Lo-Hsu Medical Foundation, Lotung Poh-Ai Hospital, Yilan, Taiwan

    Szu-Yuan Wu

  7. Department of Healthcare Administration, College of Medical and Health Science, Asia University, Taichung, Taiwan

    Szu-Yuan Wu

  8. Cancer Center, Lo-Hsu Medical Foundation, Lotung Poh-Ai Hospital, Yilan, Taiwan

    Szu-Yuan Wu

  9. Centers for Regional Anesthesia and Pain Medicine, Taipei Municipal Wan Fang Hospital, Taipei Medical University, Taipei, Taiwan

    Szu-Yuan Wu

  10. Department of Food Nutrition and Health Biotechnology, College of Medical and Health Science, Asia University, Taichung, Taiwan

    Szu-Yuan Wu

Authors
  1. Mingyang Sun
    View author publications

    Search author on:PubMed Google Scholar

  2. Fangfang Li
    View author publications

    Search author on:PubMed Google Scholar

  3. Yangyang Wang
    View author publications

    Search author on:PubMed Google Scholar

  4. Mengrong Miao
    View author publications

    Search author on:PubMed Google Scholar

  5. Zhongyuan Lu
    View author publications

    Search author on:PubMed Google Scholar

  6. Wan-Ming Chen
    View author publications

    Search author on:PubMed Google Scholar

  7. Szu-Yuan Wu
    View author publications

    Search author on:PubMed Google Scholar

  8. Jiaqiang Zhang
    View author publications

    Search author on:PubMed Google Scholar

Contributions

M.S. and F.L. conceptualized and designed the study. Y.W. and M.M. were responsible for data collection and statistical analysis. Z.L. and W.C. prepared figures and tables. J.Z. wrote the main manuscript text, with critical revisions from all other authors. All authors reviewed and approved the final manuscript.

Corresponding authors

Correspondence to Szu-Yuan Wu or Jiaqiang Zhang.

Ethics declarations

Ethics approval and consent to participate

This study was conducted using de-identified data from the National Health Insurance Research Database (NHIRD), a publicly available database in Taiwan. Since the data are fully anonymized and do not contain identifiable personal information, the study does not involve human participants directly. As such, ethical approval and informed consent were not required. This complies with the relevant guidelines and regulations for the use of anonymized secondary data in research.

Conflict of interest

The authors declare no competing interests related to this study.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Sun, M., Li, F., Wang, Y. et al. Postoperative sepsis and its sequential impact on dementia. Crit Care 29, 66 (2025). https://doi.org/10.1186/s13054-025-05276-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13054-025-05276-z

Keywords

Critical Care

ISSN: 1364-8535