Hospital‐acquired infections as a risk factor for post‐traumatic epilepsy: A registry‐based cohort study

Abstract Objective Hospital‐acquired infections are a common complication for patients with moderate or severe traumatic brain injury (TBI), contributing to morbidity and mortality. As infection‐mediated immune responses can predispose towards epilepsy, we hypothesized that post‐injury hospital‐acquired infections increase the risk of post‐traumatic epilepsy (PTE). Methods A retrospective cohort study of adults with moderate to severe TBI was conducted using data from the Victorian State Trauma Registry in Australia. Infections were identified from the International Statistical Classification of Diseases and Related Health Problems 10th Revision–Australian Modification (ICD‐10‐AM) codes, and diagnosis of PTE was determined by the Glasgow Outcome Scale – Extended questionnaire regarding epileptic fits at 24 months follow‐up. Results Of all TBI patients (n = 15 152), 24% had evidence of having had any type of infection, with the most common being pneumonia, urinary tract, and respiratory infections. Of those who responded to the PTE question at 24 months (n = 1361), 11% had developed PTE. Univariable analysis found that the incidence of PTE was higher in patients who had any type of infection compared to patients without an infection (p < 0.001). After adjustment for covariates associated with both development of PTE and risk of infection, multivariable analysis found a solid association between infection and PTE (adjusted RR = 1.59; 95% CI: 1.11–2.28; p = 0.011). Having any type of complicating infection acquired during admission was also associated with poor GOSE outcomes at subsequent follow‐ups (adjusted OR = 0.20; 95% CI: 0.11–0.35, p < 0.001). Significance These findings suggest that hospital‐acquired infections contribute to PTE development after TBI. Future investigation into infections as a modifiable target to reduce poor outcomes after TBI is warranted. Plain Language Summary Hospital‐acquired infections are common in patients with traumatic brain injuries. A database study of adults with moderate or severe brain injuries in Australia examined whether these infections are associated with the development of epilepsy after a brain injury. 24% of patients had infections, with pneumonia and urinary tract infections being the most common. Of those surveyed 2 years after the injury, 11% developed post‐traumatic epilepsy. Patients with infections had a significantly higher risk of epilepsy, even when accounting for other known risk factors, and infections were also linked to poor outcomes more broadly. The study suggests that preventing hospital‐acquired infections could be a crucial target for improving outcomes after traumatic brain injuries.


| INTRODUCTION
4][5] Prevention or suppression of PTE requires an understanding of the risk factors that contribute to its development, a process known as epileptogenesis.Risk factors identified to date include injury severity (low presenting Glasgow Coma Scale [GCS] score), [6][7][8] presence of a skull fracture, 7 penetrating or additional craniofacial injuries, 9 acute post-traumatic seizures, 9,10 and genetic variation. 11,12Another potential risk factor unexplored in this context is post-injury infection.
Infections are common complications for patients with TBI during hospitalization. 135][16] The high incidence is multifactorial, primarily attributable to the need for mechanical ventilation, urinary catheters, and other invasive medical/surgical procedures that may introduce an infection source. 13Furthermore, reported immunosuppression involving acute lymophopenia and a dysregulated T cellmediated response after TBI may contribute to increased vulnerability to infections. 14,15][19][20][21] This evidence raises the question of whether infections might also promote the development of PTE.3][24] Independent of TBI, infections are well-known contributors to seizures and epilepsy, such as febrile seizures during infancy. 23,25In adults, a large study of almost 2 million Danish individuals found that hospital contact with infection (primary diagnosis of infection) was associated with a 78% increase in the risk of a subsequent epilepsy diagnosis compared to people without exposure to infection. 22fter TBI, we hypothesize that the additional insult of sustaining an infection may drive the development of PTE in an injured brain that is already at a higher risk of epilepsy as a result of the injury.However, very few studies have considered a potential link between infectious complications and PTE, and most relate to the specific context of penetrating head injuries in military settings. 26,27imited preclinical studies incorporating inoculation of live bacteria with experimental TBI models typically support the hypothesis that infections worsen outcomes [as recently reviewed in Ref. 28], although seizures and epilepsy have not been explored in this context to date.Clinically, a 1169386, 1176426 and 2009998; Veski the development of epilepsy after a brain injury.24% of patients had infections, with pneumonia and urinary tract infections being the most common.Of those surveyed 2 years after the injury, 11% developed post-traumatic epilepsy.Patients with infections had a significantly higher risk of epilepsy, even when accounting for other known risk factors, and infections were also linked to poor outcomes more broadly.The study suggests that preventing hospital-acquired infections could be a crucial target for improving outcomes after traumatic brain injuries.

K E Y W O R D S
bacterial, epilepsy, meningitis, nosocomial, seizure, sepsis

Key points
• Post-traumatic epilepsy (PTE) is a common consequence of traumatic brain injuries.
• We explored whether the risk of PTE was altered by a prior hospital-acquired infection.
• Univariable analysis found a higher incidence of PTE in those who had sustained an infection.
• The association remained after adjustment for covariates such as injury severity.
• Future studies to explore this potential relationship are warranted.
clear association between infections (while hospitalized) and PTE in TBI patients has not been reported, nor has this potential association been examined in a large civilian population of individuals with moderate to severe TBI.
Here, we therefore used an Australian population-based trauma registry to test the hypothesis that sustaining an infection during the acute hospitalization period following a moderate to severe TBI would increase the risk of PTE at 2-year follow-up.Moderate to severe TBI was defined as an Abbreviated Injury Scale (AIS) head severity score of 3 (serious, nonlife threatening) to 6 (unsurvivable), indicating moderate to severe intracranial injury or skull fracture. 30The AIS is an anatomical injury severity scoring system with established correlation with TBI outcomes, 30 which reportedly correlates poorly with the clinical examination-based GCS. 31 However, some studies have reported that head AIS injury categories are better able to cluster TBI patients into homogenous groups, reflecting its superior ability to indicate injury severity. 32We also focused on the AIS head score as it reflects a confirmed intracranial injury or skull fracture, while GCS is a measure of consciousness which can be impaired even in the absence of intracranial injury.Patients with a pre-existing diagnosis of epilepsy were excluded from the study, as identified using the International Statistical Classification of Diseases and Related Health Problems 10th Revision-Australian Modification (ICD-10-AM) codes from the admission.Reporting of patient inclusion/exclusion is in alignment with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Figure 1).

registrations, and patient consents
The VSTR dataset was collected with ethical approval from the Victorian Department of Health and Human Services Human Research Ethics Committee (Project ID #78938) and the Monash University Human Research Ethics Committee (CF13/3040-2001000165).An anonymized dataset was extracted from the VSTR of eligible patients following project-specific ethics approval from Monash University's Human Ethics Committee (MUHREC Project ID 18104).

| Data collection
ICD-10-AM codes were completed at discharge for the period of admission.Extracted data included demographics (e.g., age, sex), injury-related information (e.g., polytrauma, cause of injury, type of injury), and discharge information including in-hospital mortality and comorbidities (e.g., history of drug or alcohol abuse).In addition, ICD-10-AM diagnosis related to infections of various specified and unspecified causes were extracted (e.g., sepsis, meningitis, pneumonia, wound infection) where coded as complicating diagnoses (prefix C), indicating that the condition(s) were acquired following admission/primary diagnosis.Included ICD-10-AM codes are presented in Table 1.As previous literature has suggested that pneumonia is the most common complicating infection in this high-risk patient group, 13,34,35 we considered this as a distinct category as well as grouping all "respiratory infections" together for analysis.
Isolated TBI was defined as the absence of an injury in any other AIS body region with a severity score > 1.The Injury Severity Score (ISS) is derived from the AIS to provide an overall severity score. 30The initial Glasgow Coma Scale (GCS) score was used as a further measure of TBI severity.The Charlson Comorbidity Index (CCI) and indicators for pre-existing mental health, drug, and alcohol conditions were also mapped from the ICD-10-AM codes. 36,37The Accessibility and Remoteness Index of Australia (ARIA), a measure of geographical remoteness, was determined by residential postcode and was used as an indicator of accessibility to services including healthcare.Early seizures were defined based on the ICD-10-AM codes R56.8 (unspecified convulsions; 91.6%), G41 (status epilepticus) and G40 (acute symptomatic seizures), to identify new onset seizures during the acute admission stage but exclude pre-existing epilepsy, as previously reported. 10BI cases on the VSTR were routinely followed up by standardized, structured telephone interviews at 6, 12, and 24 months post-injury, for collection of the Glasgow Outcome Scale-Extended (GOS-E). 38For the purposes of this study, data from all three follow-up time points were accumulated, and PTE status was analyzed at the 24month time point.This decision was based upon previous studies indicating a higher incidence of PTE as more time passes after a TBI, up to at least 2 years. 39,40Diagnosis of PTE was determined by patient responses to the GOS-E question: "Since the injury, has the patient had any epileptic fits?". 38The question was routinely asked at all follow-up interviews from 12 March 2018.We thus considered a PTE diagnosis according to the most recent ILAE definition, where epilepsy can be diagnosed after "One unprovoked (or reflex) seizure and a probability of further seizures similar to the general recurrence risk (at least 60%) after two unprovoked seizures, occurring over the next 10 years." 1 Previous studies have found that up to 86% of patients who have had a first late seizure following a moderate-severe TBI will have recurrent seizures within two years. 2,41

| Statistical analysis
Descriptive statistics for hospital-acquired infections included all patients who met the inclusion criteria.Frequencies and percentages for categorical variables, and median and interquartile ranges (IQR) for nonnormally distributed continuous variables, were used to summarize the data.Analysis of the incidence of PTE at 24 months follow-up was limited in patients who were due to have 24 months follow-up on 12 March 2018 or afterward.Univariable associations between any complicating infection or infection subtypes and development of PTE were assessed using chi-square test.Any complicating infection and all the infection subtypes with p-value < 0.25 in the univariable analysis were  33 then put into multivariable Poisson regression models with robust variance, separately. 42From a list of demographic and clinical factors that were described in our previous study, 10 we used purposeful variable selection method 42 to select potential confounders that might be associated with both development of PTE (the outcome of interest) and the risk of any complicating infection acquired during admission (the exposure of interest).Factors that had univariable p-value < 0.25 with both development of PTE and the risk of any complicating infection were selected as covariates and included in the initial model.The Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC) were calculated for all possible combinations of covariates, where lower BIC and AIC values indicate preferable combinations.To minimize overfitting while maintaining a relatively good model fit, a two-step model selection approach was employed.Firstly, combinations with the smallest BIC values and which were not significantly different (defined as having an absolute difference of less than 2) were shortlisted.Subsequently, the combination with the lowest AIC was chosen for inclusion in the multivariable analyses.Adjusted risk ratios (RR) and the corresponding 95% confidence intervals (CIs) were reported.Mixed-effect multilevel ordinal logistic regression with random intercept at patient level was used to assess the effect of any complicating infection on recovery outcomes GOSE, with adjustments of relevant covariates including follow-up time point, early posttraumatic seizures, age, sex, history of mental health conditions, and Glasgow Coma Scale category.The statistical significance level was set at p < 0.05.All statistical analyses were performed using Stata version 16.1 (StataCorp).
T A B L E 1 ICD-10-AM codes for the corresponding conditions included in the analysis.

| Frequency of hospital-acquired infections after TBI
Over the 15-year period of captured data for this study, a total of 15 152 patients with moderate and severe TBI met the inclusion criteria (Figure 1).Of these, 3576 (24%) experienced any type of infection based on relevant ICD-10-AM codes (Table 2).Eighty percent of codes with an ICD-10-AM code for infections were classified with a C-prefix as a complicating condition acquired during hospitalization (79.6%).Most of the remaining cases were diagnosed with P-prefix as primary condition presented at the time of admission (19.9%), and rarely with A-prefix as an associated condition at the time of admission (0.2%) or U-prefix as an unknown condition onset flag (0.3%).The proportion of infections classified as a complication was particularly high for meningitis, wound infections, pneumonia, intestinal infections, and sepsis (>85%), and was relatively lower for respiratory infections (63%), and urinary tract infections (72%).The details of types of infections reported are presented in Table 2.

| Incidence of PTE at 24 months follow-up
After the PTE question was routinely asked at all followups from 12 March 2018, 1361 (73%) of the 1871 patients survived the first 24 months post-injury were followed up at 24 months (Figure 1).Among them, 152 (11%) had developed PTE by 24 months post-injury.Of these 152 patients, 45 (29.5%) were first recorded to have developed PTE at the 6-month follow-up time point.19 patients (12.5%) reported PTE developing between 6-12 months and 34 (22.4%) between 12-24 months.A further 23 patients (15.1%) developed PTE any time within 12 months (as they missed the 6-month follow-up), and 21 (20.4%)developed PTE any time within 24 months (as they missed both the 6-and 12-month follow-ups).
Univariable analysis showed the proportion of PTE was higher among patients who had any type of complicating infection acquired during admission (20%) compared to patients that did not sustain an infection (9.2%; RR = 2.13; 95% CI: 1.56-2.90;p < 0.001) (Table 3).
The "infection burden," categorized as 1 (n = 179) vs.With adjustment for selected covariates (Table 4) in multivariable analysis, patients who experienced any type of complicating infection had a higher risk of developing PTE compared to patients who did not sustain an infection (adjusted RR = 1.59; 95% CI: 1.11-2.28;p = 0.011).Further multivariable analyses in infection subtypes with a p < 0.25 on univariable analysis did not find any infection subtypes associated with PTE (Table 5).
Patients who had any type of complicating infection acquired during initial admission for TBI were also associated with poor GOSE outcomes at subsequent followups (adjusted odds ratio [OR] = 0.20; 95% CI: 0.11-0.35,p < 0.001), after adjusted for follow-up time point, early posttraumatic seizures, age, sex, history of mental health conditions, and Glasgow Coma Scale category.

| DISCUSSION
This study aimed to determine whether infections sustained during acute hospitalization following severe TBI were associated with the risk of PTE up to 2 years postinjury.Secondarily, we sought to evaluate whether infections were associated with outcomes.We found through univariate analysis that the incidence of PTE was higher in patients who had sustained a hospital-acquired infection (20%) compared to those that did not sustain an infection (9.2%), particularly for pneumonia, respiratory tract, and urinary tract infections.This association was a bit weaker yet remained significant by multivariate analysis (adjusted RR = 1.59; 95% CI: 1.11-2.28).Several past studies have indicated that injury severity measures such as ISS head AIS are risk factors for the occurrence of ventilator-associated pneumonia in patients with severe TBI, 43,44 suggesting that high injury severity has a considerable influence over one's susceptibility to infection.
Overall, the statistical significance of our findings indicates a solid association between infection and PTE that can be detected with the current sample size.scroft (1941) examined the reports of World War I veterans who sustained gunshot wounds to the head.Of 317 cases, 34% reported one or more seizure "fits," and these were twice as frequent when wound healing was delayed more than 60 days (indicating prolonged wound sepsis), 26 compared to patients with wounds that healed in less than 60 days, or less than 15 days (the lowest incidence).This finding was independent of dural penetration, which in itself was associated with an increased risk of seizures. 26ore recently, a retrospective study of 489 veterans with penetrating head injuries from the Iran-Iraq War reported that 32% of the population developed epilepsy across a median follow-up time of 39 months, and central nervous system (CNS) infections (intracranial sepsis) were independently associated with the risk of epilepsy by univariate analysis. 27Finally, a recent study from the University of Pittsburgh also found that intracranial infections were independently associated with the onset of the first late post-traumatic seizure. 4510][45][46][47][48] It is well established that infection risk is high in the intensive care unit, particularly for patients with TBI who often present with multiple comorbidities. 13ndwelling catheters and arterial lines, neurosurgical procedures, ventilator use, and injury-induced immunosuppression, all likely increase susceptibility to infection in this population. 34,49,50We found that 24% of patients with a severe TBI over a 15-year reporting period also sustained an infection, with the vast majority of these (80%) recorded as being acquired during hospitalization.Consistent with previous literature, 13,34,35 the most common complicating condition (i.e., reported as onset after hospital admission) was pneumonia, followed by urinary tract infections and other respiratory tract infections.2][53] Our new findings provide some evidence to suggest that these infections may also drive secondary neuropathological mechanisms, such as inflammation, to promote epileptogenesis after TBI.Independent of TBI, it is increasingly understood that immune challenges such as bacterial and viral infections can promote neuronal network hyperexcitability and lead to seizures or epilepsy. 54or example, infections often trigger the release of pro-inflammatory cytokines such as interleukin-1 beta (IL-1β).Preclinical evidence has shown that signaling via IL-1β can enhance long-term excitability following immune activation, 55 including in the context of experimental TBI 56 ; while biomarker studies have reported that IL-1β is associated with the development of PTE. 11ur current findings support future studies to delineate T A B L E 3 Univariable analysis of proportion of patients developed post-traumatic epilepsy among the underlying complicating conditions.the cellular molecular mechanisms underlying this interaction in the context of infections, immune responses, and epilepsy following brain injuries.There were some limitations to note regarding our approach.Firstly, we relied on recorded ICD codes to assess infection exposure, rather than positive microbiology cultures (not available), which may contribute to underreporting of infections or reporting bias.Secondly, epilepsy diagnosis at 2 years follow-up was based on response to a self-reporting questionnaire, as the only available information, and to our knowledge, the sensitivity and specificity of this questionnaire have not been established.Further, evidence that a single reported post-TBI seizure developed into recurrent seizure (i.e., PTE) in our cohort was not available.A clinical diagnosis according to the most widely accepted classifications of epilepsy according to the International League Against Epilepsy 57 would have been preferred if available in future studies.Further, the sample size was unavoidably restricted to a subset of available patients in the Registry who had routinely been asked this question at follow-up since 2018, rather than using the entire dataset, thus limiting the statistical power.We also did not have detailed information on pharmacological drug use in this population which may have influenced our findings, as it has been reported that as many as 90% of TBI patients with hospital-acquired pneumonia receive antibiotic treatments, 58 and prophylactic antiseizure medications may also be routinely administered. 59he study design did not allow us to determine whether an infection might alter the time course of PTE development -for example, it is feasible that an infection could accelerate the process of epileptogenesis after injury, resulting in a shorter latency to the onset of the first late post-traumatic seizure. 60Finally, infections are a leading cause of potentially avoidable re-admission episodes to acute care for TBI patients, 61 and it remains unclear how or whether such post-injury infection challenges during the chronic post-injury period might influence the development of PTE.These unanswered questions should be addressed in future studies.

| CONCLUSION
Our findings indicate that hospital-acquired infections are associated with the development of PTE after severe TBI.This raises the question of whether infections represent a modifiable risk factor and therapeutic target to reduce or prevent epilepsy in this population.For example, systemic infections can be minimized through the implementation of non-pharmacological infection control strategies, 62 while acute post-injury prophylactic broad-spectrum antibiotics to treat infections have been associated with improved survival. 63Early screening for pneumonia in intubated trauma patients via assessment of bronchoalveolar lavage fluid has also been shown to reduce the incidence of ventilator-associated pneumonia. 64longside previous reports of a high incidence of infections detected in patients with moderate and severe TBI, our findings suggest that minimizing the incidence of hospital-acquired infections in patients after severe TBI as a means to potentially prevent poor outcomes including worse GOSE and the development of PTE in a subset of vulnerable individuals.Future studies in which a clinical diagnosis of epilepsy can be examined following a prior hospital-acquired infection are warranted to confirm this relationship.

AUTHOR CONTRIBUTIONS
Project conceptualization: Bridgette D. Semple, Terence J. O'Brien, Joshua Laing, Jian Li.Data access and analysis: Zhibin Chen, Joshua Laing, Belinda J. Gabbe.Manuscript draft: Zhibin Chen, Bridgette D. Semple, Belinda J. Gabbe.Manuscript edits: Zhibin Chen, Joshua Laing, Jian Li, Terence J. O'Brien, Belinda J. Gabbe, Bridgette D. Semple.(Li), and TOB.BDS was supported by a Veski Near-Miss Grant.JL (Laing) was supported by an NHMRC PhD scholarship (GNT1169386) and the Brain Foundation.BJG was supported by an NHMRC Investigator Grant (APP2009998).ZC was supported by an Early Career Fellowship from the NHMRC of Australia (GNT1156444).TOB is supported by an NHMRC Investigator Grant (APP1176426).Open access publishing facilitated by Monash University, as part of the Wiley -Monash University agreement via the Council of Australian University Librarians.

CONFLICT OF INTEREST STATEMENT
None of the authors has any conflict of interest to disclose.The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.We confirm that we have read the Journal's position on issues involved in ethical T A B L E 5 Multivariate analyses of associations between hospital-acquired infection and post-traumatic epilepsy.

F
I G U R E 1 STROBE flow chart, illustrating included and excluded patients with moderate to severe traumatic brain injury (TBI) from the Victorian State Trauma Registry (VSTR), to perform an analysis on potential associations between hospital-acquired infections and post-traumatic epilepsy (PTE).STROBE = Strengthening the Reporting of Observational Studies in Epidemiology, as per von Elm et al.

PTE among patients with the condition PTE among patients without the condition RR (95% CI) p-Value n % n %
Other infections include meningitis, post-traumatic, wound, and other mycobacteria infections that each had 5 or less patients.The p-value was from chi- a

Univariable p-value for association with infection Univariable p-value for association with development of PTE Included in the initial model Selected in the final model
Factors screened for their associations with infection and development of PTE using the purposeful variable selection method.
T A B L E 4 We thank the Victorian State Trauma Outcome Registry and Monitoring (VSTORM) group for collaborating and providing data for this study.The Victorian State Trauma Registry is funded by the Department of Health, State Government of Victoria, and the Transport Accident Commission.This project was supported by an Epilepsy Research Program Idea Development Award (#W81XWH-19-ERP-IDA) from the US Department of Defense, awarded to BDS, JL infection 1.59(1.11-2.28)0.011Allanalyses were adjusted for early posttraumatic seizures, age, cause of injury -low-fall, history of mental health conditions, Glasgow Coma Scale category, and ARIA category.Abbreviations: ARIA, accessibility and remoteness index of Australia; CI, confidence interval; PTE, post-traumatic epilepsy; RR, risk ratio. Note: