Abstract
Objective
Psychological stress is commonly reported as a seizure precipitant in epilepsy, yet evidence from major crises remains limited and inconsistent. We examined whether the October 2023 Israel–Hamas conflict affected acute seizure‐related healthcare utilization among adults with epilepsy.
Methods
We conducted a retrospective cohort study of 34 941 adults with established epilepsy (International Classification of Diseases, Tenth Revision [ICD‐10] G40 diagnosis ≥ 2 years prior) using the Clalit Health Services database in Israel. We compared seizure‐related emergency department (ED) visits in two 1‐year windows: pre‐war (October 1, 2022–October 6, 2023) vs during‐war (October 7, 2023–October 1, 2024). Primary outcome was ED visit rate per person‐year. Given non‐normal distributions, paired Wilcoxon tests assessed rate changes. Multivariable logistic regression identified predictors of worsening (during > pre), and Cox proportional hazards modeling evaluated time‐to first ED visit, adjusting for demographics, socioeconomic status (SES), and comorbidities including dementia and post‐traumatic stress disorder (PTSD).
Results
Overall 3056 patients (8.8% of cohort) experienced a during‐war ED visit. Cohort‐level visit rates were similar across periods (224.13 vs 217.34 per 1000 person‐years; rate ratio .97, 95% confidence interval [CI] .94–1.00). Among 1039 patients with ED visits in both periods, median within‐patient change was negligible (−0.003 visits per person‐year), but the distribution shifted toward lower wartime rates (Wilcoxon Z = −3.11, p = .002; paired t test p = .730). Dementia and older age were associated with the shift was toward lower wartime rates. Odds of worsening were negligibly higher with age (odds ratio [OR] = 1.007). Older age (hazard ratio [HR] = 1.03) and dementia (HR = 1.20) predicted earlier ED visits, whereas PTSD (HR = .73) and higher SES (HR = .82) reduced hazard.
Significance
War‐related stress correlated with stable seizure‐related acute care utilization, with minor redistribution among recurrent ED users. Older age and dementia identify vulnerable subgroups requiring enhanced monitoring during national emergencies. Targeted monitoring and clinical support for high‐risk patients with epilepsy may help reduce acute seizure‐related ED utilization during high‐stress situations.
Keywords: epilepsy, seizures, stress, wartime
Key points.
Seizure‐related emergency department (ED) utilization remained stable during the October 2023 Israel–Hamas conflict despite prolonged national stress.
Older age and dementia were the most consistent markers of vulnerability to seizure‐related ED utilization during wartime.
Patients with PTSD and higher socioeconomic status showed distinct patterns of ED presentation without increased utilization.
Targeted monitoring of older adults and patients with dementia may improve epilepsy care during national crises.
1. INTRODUCTION
Epilepsy is a chronic neurological disorder characterized by recurrent unprovoked seizures and affects about 56.7 million people worldwide. 1 Although its pathophysiology is primary neurobiological, seizure occurrence is strongly influenced by environmental and psychological factors. Psychological stress and anxiety have long been considered as one of the major triggers of seizures and contributors to disease exacerbation. 2 , 3 , 4 , 5 , 6
The biological pathways that links stress to seizure exacerbation are complex, and several mechanisms were proposed including activation of the hypothalamic–pituitary–adrenal axis, autonomic dysregulation, sleep disorders with sleep deprivation, neuroinflammatory pathways, and impaired adherence to anti‐seizure medications. 2 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15
Despite the presumed association between stress and seizures, establishing a stress threshold for seizure precipitation remains challenging, as the stress perception is highly individualized and shaped by personal, social, and environmental factors. As a result, most of today's evidence is based on self‐reported seizure frequency and small cohort observational studies and have heterogeneous findings. 16 , 17 , 18 , 19 , 20 , 21 , 22
National scale traumatic events, such as natural disasters or military conflicts, provide an opportunity to assess the impact of stress at a population level. Previous studies conducted during armed conflicts or mass displacement presented mixed results. 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 Some demonstrated increased seizure frequency, 30 , 33 whereas others attributed this to indirect effect mediated by factors such as sleep deprivation or medication non‐adherence. 23 These inconsistencies highlight the need to move beyond viewing the population as a homogenous group, and more specifically identify the vulnerable individuals who are more sensitive to stress exposure.
On October 7, 2023, Israel experienced an unprecedented violent attack followed by a prolonged military conflict, resulting in mass casualties, widespread population displacement, repeated day and night missile alerts, and prolonged disruption of civilian life. 34 , 35 Subsequent population‐based studies documented substantial increases in post‐traumatic stress disorder (PTSD), depression, and stress‐related behavioral changes across the civilian population. 36 , 37 , 38
The following study aims to assess the impact of stress related to the October 2023 Israel–Hamas conflict on seizure control among Israeli epilepsy patients. We compared the rate of seizure‐related emergency department (ED) visits among patients with diagnosed epilepsy during the 1 year preceding the war and 1 year after the war outbreak. We also attempted to identify demographic and clinical factors that may contribute to individual stress vulnerability for seizure exacerbation.
2. METHODS
2.1. Study design and data source
A retrospective cohort study was conducted using anonymized electronic medical records derived from seizure‐related ED visits in hospitals affiliated with Clalit Health Services (CHS). Data were extracted from CHS using the Clalit Research data sharing platform powered by MDClone (https://www.mdclone.com). Only hospital‐based ED encounters were included, as ED utilization was used as a proxy measure for seizure exacerbation.
Adult patients (≥ 18 years) with a documented diagnosis of epilepsy (International Classification of Diseases, Tenth Revision [ICD‐10] G40) at least 2 years before the study period were included. Continuous medical documentation during the study period was required for inclusion. Patients without a prior documented diagnosis of epilepsy or seizure disorder (i.e., de novo seizures) were not included in the analytic cohort, as the study focused exclusively on seizure exacerbation among patients with established epilepsy. Patients with a documented diagnosis of psychogenic non‐epileptic seizures (PNES) were included in the cohort only if a concurrent diagnosis of epilepsy (ICD‐10 G40) was present.
Two predefined periods were analyzed: pre‐war period (October 1, 2022–October 6, 2023) and during‐war period (October 7, 2023–October 1, 2024). Follow‐up time within each period was defined as the interval from the start of the period (or from the beginning of Clalit membership coverage, whichever was later) until the end of the period (or until death or termination of Clalit membership, whichever came first). Accordingly, a patient contributed incomplete follow‐up only if their Clalit coverage did not span the entire calendar window of the period, or if they died within it. No administrative censoring other than death and loss of Clalit coverage were applied.
The primary outcome was the change in seizure‐related ED visit frequency between the pre‐war and during‐war periods. The secondary outcomes included time to first seizure‐related ED visit following the war onset and seizure exacerbation during wartime, as well as their association with demographic and clinical variables.
Collected variables included age, sex, socioeconomic status (SES), psychiatric diagnoses (including PTSD), vascular risk factors, and other relevant medical comorbidities.
2.2. Statistical analysis
All statistical analyses were performed using Statistical Package for the Social Sciences (SPSS) version 27.0 (IBM Corp., Armonk, NY). Descriptive statistics were computed for all study variables, including means, standard deviations (SDs), medians, and interquartile ranges (IQRs) for continuous variables, and frequencies and percentages for categorical variables. The normality of continuous variables was assessed using the Kolmogorov–Smirnov and Shapiro–Wilk tests.
For the primary analysis comparing ED visit rates before and after the war onset, the Wilcoxon signed‐rank test was employed due to the non‐normal distribution of the data. This non‐parametric paired test was used to compare person‐year normalized ED visit rates between the pre‐war period (October 1, 2022–October 6, 2023) and the wartime period (October 7, 2023–October 1, 2024).
Subgroup analyses examining differences in the magnitude of change in ED visit rates by comorbid conditions were conducted using the Mann–Whitney U test, a non‐parametric alternative to the independent samples t‐test. Correlations between continuous demographic variables (age) and change in ED visit rates were assessed using Spearman's rank correlation coefficient.
Multivariable logistic regression analysis was performed to identify independent predictors of exacerbation, defined dichotomously as having more ED visits in the wartime period compared to the pre‐war period. The model included demographic variables (age, gender, SES) and comorbid conditions (dementia, PTSD) as covariates. Results are reported as odds ratios (OR) with 95% confidence intervals (CI). Because patients could have more than one ED visit, the logistic regression was conducted at the patient level and not the visit level in order to avoid overrepresentation of patients with frequent ED utilization.
Time to ED visit analysis was conducted using Kaplan–Meier survival curves to estimate the probability of remaining event free over the follow‐up period. Cox proportional hazards regression was employed to identify predictors of time to first seizure‐related ED visit following the war onset. To distinguish war‐related effects from baseline subgroup tendencies, the Cox proportional hazards model was also fitted for the pre‐war period using an identical set of predictors (age, dementia, PTSD, gender, SES). The pre‐war survival time was defined as the number of days from the start of each patient's observed pre‐war window to the first seizure‐related ED visit within that window, with administrative censoring at the end of the observed pre‐war window (October 6, 2023, or at the patient‐level end of Clalit membership coverage, whichever came first).
Results are reported as hazard ratios (HRs) with 95% CIs. The proportional hazards assumption was evaluated. Model fit was assessed using the −2 Log Likelihood statistic and omnibus tests of model coefficients. All statistical tests were two‐tailed, and p‐values < .05 were considered statistically significant.
3. RESULTS
3.1. Sample characteristics
The study cohort comprised 34 941 adult patients with established epilepsy diagnoses (ICD‐10 G40), identified from the CHS database. The majority of index ED visits were inpatient diagnoses (83.4%), with 16.6% identified through ED visits. The mean age at the reference ED visit was 40.05 years (SD = 21.78, median = 38.01). The sample was relatively balanced by sex, with 52.1% male and 47.9% female patients. Regarding SES, the majority of patients (60.8%) were classified as medium SES, whereas 23.6% were classified as low and 15.6% as high.
Observed follow‐up was near‐complete in both study periods. During the pre‐war period, mean follow‐up was 344.67 days (SD = 82.65, median = 365), and 32 865 patients (94.1%) contributed the full 365‐day window. During the wartime period, mean follow‐up was 339.76 days (SD = 89.26, median = 366), and 31 740 patients (90.8%) contributed the full window (Table S1).
Incomplete follow‐up in the pre‐war period (n = 2077, 5.9%) was attributable to late Clalit enrollment within the pre‐war window (n = 510, 1.5%) or to termination of follow‐up before the end of the pre‐war window (n = 1567, 4.5%), the latter including 114 patients (.3%) who died within the pre‐war period and 80 patients (.2%) who disenrolled from Clalit within the period (Table S2).
Incomplete wartime follow‐up (n = 3201, 9.2%) was similarly partitioned into late enrollment (n = 327, .9%) and early termination (n = 2890, 8.3%), with 1101 patients (3.2%) dying during the wartime period and 1323 (3.8%) disenrolling from Clalit. Sixteen patients joined Clalit after October 7, 2023 (late enrollment) and died or disenrolled before October 1, 2024 (early termination); therefore they appear in both groups (Table S2). The ~10‐ to 16‐fold higher wartime mortality and disenrollment relative to the pre‐war period account for the 3.3 percentage‐point reduction in complete follow‐up during the wartime period.
The demographic and clinical characteristics of the study population listed at Table 1, and their comorbidities at Table 2. Notable comorbidities included depression (27.3%), hypertension (35.6%), stroke (21.4%), dementia (14.1%), and intellectual disability (10.8%). PTSD was present in 3.2% of patients, whereas autism spectrum disorder was diagnosed in 1.3% of the cohort.
TABLE 1.
Demographic and clinical characteristics of the study population (N = 34 941).
| Characteristic | n/Mean | %/SD |
|---|---|---|
| Gender | ||
| Male | 18 199 | 52.1% |
| Female | 16 742 | 47.9% |
| Age at ED visit, years | 40.05 | 21.78 |
| SES | ||
| Low | 7676 | 23.6% |
| Medium | 19 759 | 60.8% |
| High | 5069 | 15.6% |
| Follow‐up duration, days | 339.70 | 76.31 |
TABLE 2.
Prevalence of comorbid conditions (N = 34 941).
| Comorbid condition | n | % |
|---|---|---|
| Dementia | 4910 | 14.1% |
| Depression | 9547 | 27.3% |
| PTSD | 1122 | 3.2% |
| Hypertension | 12 451 | 35.6% |
| Stroke | 7462 | 21.4% |
| Myocardial infarction | 5291 | 15.1% |
| Intellectual disability | 3774 | 10.8% |
| Autism spectrum disorder | 459 | 1.3% |
| PNES | 67 | .2% |
3.2. Primary outcome–emergency department visits for epileptic seizures
During the pre‐war period, 7390 seizure‐related ED visits were recorded across 32972.66 person‐years of follow‐up (224.13 visits per 1000 person‐years), contributed by 3152 patients (9.0% of the cohort). During the wartime period, 7064 seizure‐related ED visits were recorded across 32502.77 person‐years (217.34 visits per 1000 person‐years), contributed by 3056 patients (8.8% of the cohort) (Table 3, Figure S1).
TABLE 3.
Seizure‐related ED visits and person‐years by period (N = 34 941).
| Metric | Pre‐war | Wartime |
|---|---|---|
| Total seizure‐related ED visits | 7390 | 7064 |
| Total person‐years of follow‐up | 32972.66 | 32502.77 |
| Patients with ≥1 ED visit, n (%) | 3152 (9.02%) | 3056 (8.75%) |
| ED visit rate per person‐year | .2241 | .2173 |
| ED visit rate per 1000 person‐years | 224.13 | 217.34 |
| Patient‐level event rate per person‐year | .0956 (95.6 per 1000 PY) | .0940 (94.0 per 1000 PY) |
The cohort‐level visit‐rate ratio was .97 (95% CI .94–1.00). However, the distribution of changes within patients was not uniform. Of the 34 941 patients, 2105 (6.0%) had ED visits exclusively during the pre‐war period, 2009 (5.7%) had ED visits exclusively during the wartime period, and 1047 (3.0%) had ED visits in both periods. The paired Wilcoxon signed‐rank test was restricted to the 1039 patients in the last group with non‐missing person‐year–normalized rates in both periods.
Within this subset, 382 patients (36.8%) had a higher wartime than pre‐war rate (positive ranks, mean rank 628.54) and 657 (63.2%) had a lower wartime than pre‐war rate (negative ranks, mean rank 456.89); no patients had identical rates. The Wilcoxon signed‐rank test detected an asymmetric shift in the distribution of signed ranks (Z = −3.11, p = .002) (Table 4), but the absolute magnitude of the within‐patient change was small: median Δ rate = −.003 visits per person‐year, interquartile range −1.01 to +.99, paired t‐test mean difference −.057, 95% CI −.38 to +.27, p = .730.
TABLE 4.
Wilcoxon signed‐rank comparison of person‐year‐normalized seizure‐related ED visit rates among the 1039 patients with ED visits in both study periods.
| Category | n | Mean rank | Sum of ranks |
|---|---|---|---|
| Positive ranks: during‐war rate > pre‐war rate (wartime INCREASED) | 382 | 628.54 | 240 102 |
| Negative ranks: during‐war rate < pre‐war rate (wartime DECREASED) | 657 | 456.89 | 300 178 |
| Ties: pre‐war rate = wartime rate | 0 | — | — |
| Total | 1039 | — | — |
Note: Analysis restricted to patients with seizure‐related ED visits in both study periods (n = 1039). Positive ranks indicate higher wartime than pre‐war person‐year‐normalized ED visit rates; negative ranks indicate lower wartime than pre‐war rates. Mean rank and sum of ranks are Wilcoxon signed‐rank test statistics and should not be interpreted as direct clinical effect sizes. The test uses the asymmetry between positive and negative rank sums to evaluate whether the distribution of within‐patient differences is symmetric about zero. Z = −3.11, p = .002 (two‐tailed) Median within‐patient Δ rate (wartime − pre‐war) = −.003 visits per person‐year; IQR −1.01 to +.99.
3.3. Subgroup analyses by comorbid conditions
Mann–Whitney U tests compared the distribution of patient‐level change scores in person‐year–normalized ED visit rates (wartime minus pre‐war) between patients with and without each comorbid condition (Table 5). Pre‐war and wartime rates are additionally displayed for each subgroup so that the direction and magnitude of the change can be read directly from the table.
TABLE 5.
Person‐year–normalized seizure‐related ED visit rates before and during the war period, by comorbidity status, in the paired‐analysis subset (N = 1039).
| Comorbidity (ICD‐10) | Group | n | Pre‐war rate, mean (median) | Wartime rate, mean (median) | Within‐patient Δ rate, median | Mann–Whitney Z | p |
|---|---|---|---|---|---|---|---|
| Dementia | Diagnosis absent | 817 | 3.29 (2.00) | 3.21 (2.00) | −.003 | −3.89 | < .001 |
| Diagnosis present | 222 | 2.73 (2.00) | 3.30 (2.00) | −.003 | |||
| PTSD | Diagnosis absent | 991 | 3.11 (2.00) | 3.20 (2.00) | −.003 | −.99 | .322 |
| Diagnosis present | 48 | 4.52 (2.50) | 3.77 (2.00) | −.506 | |||
| Depression | Diagnosis absent | 633 | 2.96 (2.00) | 3.21 (2.00) | −.003 | −1.01 | .311 |
| Diagnosis present | 406 | 3.51 (2.00) | 3.26 (2.00) | −.003 | |||
| Hypertension | Diagnosis absent | 628 | 3.31 (2.00) | 3.29 (2.00) | −.003 | −1.71 | .087 |
| Diagnosis present | 411 | 2.97 (2.00) | 3.14 (2.00) | −.003 | |||
| Stroke | Diagnosis absent | 736 | 3.21 (2.00) | 3.20 (2.00) | −.003 | −1.43 | .153 |
| Diagnosis present | 303 | 3.09 (2.00) | 3.31 (2.00) | −.003 | |||
| Myocardial infarction | Diagnosis absent | 843 | 3.29 (2.00) | 3.27 (2.00) | −.003 | −1.51 | .131 |
| Diagnosis present | 196 | 2.66 (2.00) | 3.06 (2.00) | −.003 | |||
|
Intellectual disability |
Diagnosis absent | 850 | 3.04 (2.00) | 3.12 (2.00) | −.003 | −.32 | .751 |
| Diagnosis present | 189 | 3.77 (2.00) | 3.75 (2.00) | −.003 | |||
| Autism spectrum disorder | Diagnosis absent | 1017 | 3.15 (2.00) | 3.22 (2.00) | −.003 | −1.81 | .070 |
| Diagnosis present | 22 | 4.05 (3.00) | 3.60 (2.49) | −.010 | |||
| PNES | Diagnosis absent | 1030 | 3.12 (2.00) | 3.20 (2.00) | −.003 | −.80 | .421 |
| Diagnosis present | 9 | 9.56 (5.00) | 7.10 (3.99) | −.003 |
Note: Analysis restricted to the 1039 patients with ≥ 1 seizure‐related ED visit in both the pre‐war and wartime periods. Rates are expressed as seizure‐related ED visits per person‐year of follow‐up. “Diagnosis present” denotes documented ICD‐10 coding of the condition at any time prior to or during the study period. “Diagnosis absent” denotes no such coding. Within‐patient Δ rate is the per‐patient (wartime − pre‐war) difference in person‐year–normalized rates; the median is reported because the distribution is highly non‐normal. Mann–Whitney Z and p‐values compare the distribution of Δ rate between the two comorbidity groups.
Abbreviations: ED, emergency department; PNES, psychogenic non‐epileptic seizure; PTSD, post‐traumatic stress disorder.
Dementia was the only comorbidity with a statistically significant association with the within‐patient change. Patients with dementia (n = 222) moved from a mean pre‐war rate of 2.73 visits per person‐year to a mean wartime rate of 3.30 visits per person‐year, whereas patients without dementia (n = 817) showed essentially no net change (3.29 → 3.21). The Mann–Whitney test confirmed a significantly greater upward shift in the dementia subgroup (U = 75308.5, Z = −3.89, p < .001).
Other comorbidities did not show statistically significant differences in the univariate subgroup analyses: PTSD (Z = −.99, p = .322), depression (Z = −1.01, p = .311), hypertension (Z = −1.71, p = .087), stroke (Z = −1.43, p = .153), myocardial infarction (Z = −1.51, p = .131), intellectual disability (Z = −.32, p = .751), and autism (Z = −1.81, p = .070).
3.4. Demographic correlates
Spearman correlation analysis revealed a weak but statistically significant positive correlation between age and the magnitude of change in ED visit rates (ρ = .112, p < .001), indicating that older patients tended to experience greater increases in seizure‐related ED visits following the war onset. SES showed no significant correlation with change in ED visit rates (ρ = .003, p = .915). Gender differences were examined using the Mann–Whitney U test, which revealed no significant difference between males (n = 558) and females (n = 481) in the magnitude of change (p > .05).
3.5. Logistic regression analysis: Risk factors for exacerbation
A multivariable logistic regression model was constructed to identify independent predictors of exacerbation, defined as having more ED visits in the wartime period than in the pre‐war period. This regression was performed at the patient level. Analysis was restricted to the 3056 patients with at least one seizure‐related ED visit during the wartime period (October 7, 2023–October 1, 2024); 229 excluded for missing covariates, leaving 2827 patients in the regression. The overall model was statistically significant and explained a small but meaningful proportion of variance in exacerbation risk.
Age emerged as statistically significant predictor of exacerbation; however, the effect size was not clinically significant (OR = 1.007, 95% CI 1.003–1.012, p = .003) (Table 6, Figure 1). Dementia, PTSD, and male gender showed a trend toward seizure exacerbation, although the associations did not reach statistical significance. High SES showed a nonsignificant tendency toward a protective effect (OR = 1.04, 95% CI .77–1.41, p = .798).
TABLE 6.
Wartime‐users logistic regression: Predictors of exacerbation among patients with ≥ 1 wartime ED visit (N = 2827 patients).
| Variable | B | SE | OR | 95% CI | p |
|---|---|---|---|---|---|
| Age | .007 | .002 | 1.007 | 1.003–1.012 | .003 |
| Dementia | −.060 | .126 | .94 | .74–1.20 | .632 |
| PTSD | −.189 | .220 | .83 | .54–1.27 | .391 |
| SES, high vs low | .039 | .154 | 1.04 | .77–1.41 | .798 |
| SES, medium vs low | .131 | .106 | 1.14 | .93–1.40 | .217 |
| Gender, male | −.071 | .090 | .93 | .78–1.11 | .431 |
Note: Analysis restricted to the 3056 patients with at least one seizure‐related ED visit during the wartime period (October 7, 2023 – October 1, 2024); 229 excluded for missing covariates, leaving N = 2827 in the regression. Outcome: exacerbation = 1 if wartime ED visits > pre‐war ED visits, else 0 (2187 patients, 77.4%, met this criterion). Model χ 2 = 14.098, df = 6, p = .029; Nagelkerke R 2 = .008; Hosmer–Lemeshow χ 2 = 3.697, df = 8, p = .883 (good fit).
Abbreviations: B, unstandardized regression coefficient; CI, confidence interval; ED, emergency department; OR, odds ratio; PTSD, post‐traumatic stress disorder; SE, standard error; SES, socioeconomic status.
FIGURE 1.
Patient level logistic regression—predictors of exacerbation in seizure among patients with ≥ 1 wartime emergency department (ED) visit (N = 2827 patients). Wartime‐users logistic regression. Age was a modest predictor among wartime ED users with limited clinical significance (odds ratio [OR] 1.007 per year, p = .003). Although the confidence interval is narrow, the effect is statistically significant. PTSD, post‐traumatic stress disorder; SES, socioeconomic status.
3.6. Survival analysis: Time to first wartime ED visit
Kaplan–Meier survival analysis was conducted to estimate the time to first seizure‐related ED visit following the conflict onset. Of the 34 941 patients in the cohort, 3094 (8.9%) experienced an ED visit during the follow‐up period, whereas 31 847 (91.1%) were censored. This number differs from the 3056 in the rate‐normalized analyses by 38 patients with zero or missing wartime person‐time, who are excluded from rate‐based calculations but retained in the time‐to‐event analysis. The mean survival time was 346.25 days (standard error [SE] = .36, 95% CI 345.55–346.96). The median survival time was not reached, indicating that more than 50% of patients did not experience seizure‐related ED visits during the follow‐up period.
A Cox proportional hazards regression model was performed on complete cases with available covariate data to identify predictors of time to first seizure‐related ED visit; therefore the analysis included 2903 observed events (Table 7, Figure 2). The model included both PTSD and dementia alongside demographic variables and demonstrated excellent fit (−2 Log Likelihood = 58784.52; χ 2 = 1167.00, df = 6, p < .001).
TABLE 7.
Cox proportional hazards regression: Predictors of time to first wartime seizure‐related ED visit (cohort N = 34 941; N 32504 = in analysis; 2903 ED visits, 29 601 censored).
| Variable | B | SE | HR | 95% CI | p |
|---|---|---|---|---|---|
| Age | .027 | .001 | 1.028 | 1.026–1.030 | < .001 |
| Dementia | .179 | .048 | 1.196 | 1.089–1.313 | < .001 |
| PTSD | −.320 | .134 | .726 | .558–.944 | .017 |
| Gender, male | .073 | .037 | 1.076 | 1.000–1.157 | .050 |
| SES, medium vs low | −.108 | .047 | .898 | .819–.984 | .021 |
| SES, high vs low | −.197 | .062 | .821 | .727–.927 | .001 |
Note: Model − 2LL = 58784.52; χ 2 = 1167.00, df = 6, p < .001.
Abbreviations: B, unstandardized regression coefficient; CI, confidence interval; ED, emergency department; HR, hazard ratio; PTSD, post‐traumatic stress disorder; SE, standard error; SES, socioeconomic status.
FIGURE 2.
Cox proportional hazards regression—predictors of time to first wartime seizure‐related emergency department (ED) Visit (cohort N = 34 941; N 32504 = in analysis; 2903 ED visits, 29 601 censored).High and medium socioeconomic status (SES) and post‐traumatic stress syndrome (PTSD) were associated with delayed ED presentation. Male gender showed a borderline significant association with shorter time to seizure‐related ED visit, whereas older age was significantly associated with earlier ED presentation. Each additional year of age was associated with 2.8% increase in the instantaneous risk of ED presentation at any given time point. Dementia was a significant risk factor (hazard ratio [HR] = 1.196, 95% confidence interval [CI]: 1.089–1.313, p < .001) with ~20% earlier ED presentation. PTSD, post‐traumatic stress disorder; SES, socioeconomic status.
PTSD demonstrated a protective effect on time to first ED visit (HR = .726, 95% CI .558–.944, p = .017). In contrast, dementia was a significant risk factor (HR = 1.196, 95% CI 1.089–1.313, p < .001), with affected patients presenting ~20% faster to the ED. Older age was significantly associated with shorter time to first ED visit (HR = 1.028, 95% CI 1.026–1.030, p < .001), with each additional year of age associated with a 2.8% increase in the instantaneous risk of ED presentation at any given time point. Male gender showed a borderline significant association with shorter time to ED visit (HR = 1.076, 95% CI 1.000–1.157, p = .050). Protective factors included medium SES (HR = .898, 95% CI .819–.984, p = .021) and high SES (HR = .821, 95% CI .727–.927, p = .001), both relative to low SES.
3.7. Comparison with the pre‐war period
To assess whether the survival‐analysis predictors reflect war‐specific effects or general subgroup tendencies, we fitted a parallel Cox model for the pre‐war period. The pre‐war Cox model (N = 30 821; 2843 events) identified dementia (HR 1.68, 95% CI 1.52–1.87, p < .001), PTSD (HR 1.31, 95% CI 1.09–1.57, p = .004), and SES (High‐vs.‐Low HR .64, p < .001; medium‐vs‐low HR .89, p = .006) as significant baseline predictors of time to first seizure‐related ED visit, whereas age (HR 1.001, p = .468) and female gender (HR 1.03, p = .473) were not significant. The side‐by‐side comparison with the wartime Cox model is presented in Table S3.
4. DISCUSSION
Psychological stress is frequently cited by patients as a seizure precipitant. Yet objective evidence linking stress exposure to seizure worsening has been inconsistent. Retrospective and narrative studies have shown that stress remains one of the most commonly self‐reported triggers for seizures. However, prospective evaluation is challenging due to confounding factors such as sleep disruption, medication adherence, and psychiatric symptoms. 4 , 21
As mentioned previously, studies examining seizure outcomes during wars, terrorist attacks, and other major national scale crises have produced mixed findings. During the 1991 Gulf War, only a subset of patients reported increased seizure frequency despite widespread stress exposure. 23 A study that followed patients residing near the Pentagon before and after 9/11 found that a stressful event is associated with seizure exacerbation only in a small proportion (12%) of exposed patients with epilepsy, but in a higher proportion (29%) of patients stressed by the event and in 50% of patients directly affected by the event. 24 The October 2023 Israel–Hamas conflict provided a unique setting for examining the association between wartime stress and seizure‐related ED utilization.
In this nationwide cohort of adults with established epilepsy, exposure to this conflict was not associated with a population‐wide increase in seizure‐related ED utilization. Cohort‐level visit rates were essentially unchanged between periods. A paired Wilcoxon signed‐rank analysis restricted to the 1039 patients with ED visits in both periods detected a statistically significant but clinically modest asymmetry in the direction of lower wartime rates within this chronic‐user subgroup. When classified under clinically reasonable stability thresholds, ~27%–40% of paired patients had essentially unchanged rates across the two periods. These findings do not support strong association between psychological stress and seizure exacerbation, and suggest that the aggregate impact of a prolonged national crisis on acute epilepsy‐related healthcare use is limited.
One possible explanation is that seizure‐related ED utilization may not fully reflect seizure burden during periods of prolonged conflict. During periods of recurrent missile attacks and security alerts, patients may be less willing to travel or spend time in open public areas and may prefer to remain near protected spaces and shelters. Such behavioral adaptations could lead to delayed ED presentation or preference for seeking advice through community healthcare services and direct communication with their treating neurologist. Such changes in healthcare‐seeking behavior could partially attenuate observed increases in seizure‐related ED utilization despite substantial psychological stress.
Although no overall increase in seizure‐related ED utilization was observed during wartime at the population level, patients with dementia demonstrated a distinct pattern within the paired subgroup analysis. Among patients with ED visits during both study periods, the dementia subgroup showed an increase in seizure‐related ED visit rates during wartime, suggesting increased acute care utilization in this cognitively vulnerable subgroup. However, this finding was not consistently supported across all multivariable models.
Predictors identified in the time‐to ED visit analysis do not necessarily mirror those associated with cumulative exacerbation, as these models reflect different aspects of seizure worsening. The logistic model reflects cumulative exacerbation across the study period, whereas the Cox model reflects the timing of the first seizure‐related ED visit.
In the patient‐level logistic regression restricted to patients with wartime ED visits, dementia was not independently associated with exacerbation. In contrast, survival analysis demonstrated that dementia was associated with a shorter time to first seizure‐related ED presentation during wartime. Notably, a similar association was also observed during the pre‐war period, with an even higher HR, suggesting that dementia may primarily reflect a baseline tendency toward a higher hazard of a first seizure‐related ED visit within the observation window rather than a war‐specific effect. Collectively, these findings suggest that dementia may be associated with a baseline tendency toward earlier seizure‐related ED utilization, whereas wartime stress may further accentuate this tendency in a subset of cognitively vulnerable patients. Cognitive impairment may amplify vulnerability through reduced stress coping, sleep disturbance, impaired anti‐seizure medication adherence, and reduced caregiver ability to cope with seizures.
Pediatric data from conflict settings have suggested increased seizure occurrence compared with non‐affected regions, highlighting age‐specific vulnerability. 25 In line with these pediatric findings, our study suggests that age‐specific vulnerability may also extend to older adults. Spearman correlation analysis showed that older patients tended to experience greater increases in seizure‐related ED visits following the war onset. The logistic regression model revealed age as statistically significant predictor of exacerbation, although the effect size was very small and of uncertain clinical significance. The survival analysis showed that older age was slightly but statistically significant, associated with shorter time to first ED visit during wartime. A similar trend was observed during a pre‐war period; however, the HR was slightly attenuated and the association was not statistically significant. Taken together, these findings alongside the pediatric data, suggest that individuals at the extremes of the age spectrum may be particularly sensitive to stress during wartime.
In contrast to broader literature showing that lower SES is associated with higher emergency utilization and poorer epilepsy outcomes, 39 , 40 , 41 we found no significant association between SES and seizure‐related ED visits during wartime. Logistic regression analysis showed a slight tendency for high SES toward a protective effect, but it was nonsignificant. However, survival analysis demonstrated that both high and medium SES were associated with longer time to ED presentation in both study periods. These findings suggest that individuals with greater SES resources may experience a slower progression toward clinically significant seizure exacerbations requiring emergency care, potentially reflecting differences in access to outpatient care, medication adherence, social support, or coping resources.
Similar differences between the logistic and time‐to‐event models were also observed for PTSD. In the patient‐level logistic regression restricted to patients with wartime seizure‐related ED visits, PTSD was not independently associated with exacerbation. However, survival analyses demonstrated opposing temporal patterns across study periods. During the pre‐war period, PTSD was associated with a shorter time to first seizure‐related ED presentation. In contrast, during wartime, PTSD was associated with a lower hazard of early ED presentation, indicating delayed time to first seizure‐related ED visit. These findings further support the distinction between cumulative exacerbation and timing of presentation, as reflected by the different analytic models. One possible explanation is that during wartime, patients with PTSD may have demonstrated altered healthcare‐seeking behavior, avoidance of hospital settings, reduced mobility, or prioritization of psychological coping over acute medical evaluation, resulting in delayed ED presentation despite ongoing stress exposure.
Of note, these findings should not be interpreted as being fully discordant with prior literature, which has consistently shown that trauma‐related disorders are relatively prevalent among people with epilepsy and associated with poorer outcomes. 42 , 43 Rather, our results likely reflect a difference in the outcome measure assessed. The current study is based on ED presentations for seizure‐related events, which represent only a subset of clinical exacerbations and are influenced by healthcare‐seeking behavior and access to care. ED utilization does not capture seizure frequency or milder events that did not prompt medical evaluation. Therefore, the observed temporal associations with PTSD may reflect changes in presentation patterns rather than true differences in underlying seizure burden.
Despite these subgroup effects, the absolute increase in seizure‐related ED utilization was small, suggesting that most individuals with epilepsy maintained a relative seizure stability during a prolonged conflict. Possible explanations include adaptation to chronic stress in a population exposed to recurrent security threats, continuity of care within an integrated healthcare system, and effective baseline seizure control.
Several limitations warrant consideration. A seizure‐related ED visit represents a surrogate marker for seizure worsening and does not capture seizures managed in outpatient or home settings. Therefore, the representation of seizure exacerbation by ED visits may underestimate the overall seizure burden, as ED presentation generally reflects the more severe and complicated events rather than the full spectrum of seizures manifestations. Further studies assessing seizure frequency using patient diaries or other outpatient data sources could provide additional and more comprehensive information. Moreover, because our study included only seizure‐related ED visits recorded in hospitals affiliated with CHS, visits to non‐affiliated hospitals may have been missed. Diagnostic misclassification may occur despite standardized coding. Key mediators including sleep disruption, medication adherence, direct exposure to violence, and displacement could not be directly measured. Finally, as an observational study, causal inference is not possible.
5. CONCLUSION
In summary, exposure to war‐related stress during the October 2023 Israel–Hamas conflict was associated with a cohort‐level stability in seizure‐related acute care utilization. Most patients demonstrated stable ED utilization patterns, highlighting that in a chronic stress environment with recurrent security threats, overall stability in seizure‐related ED utilization can be achieved.
Although overall utilization remained stable, different patterns of ED presentation were observed across patient subgroups, particularly among older adults, patients with dementia, patients with PTSD, and those with higher SES, suggesting that certain vulnerable subgroups may respond different to wartime stress. Among these subgroups, older age and dementia demonstrated the most consistent evidence of increased vulnerability during wartime. These findings support targeted monitoring for vulnerable subgroups during a national scale crisis and contribute to nuanced understanding of stress‐epilepsy relationship.
AUTHOR CONTRIBUTIONS
Anda Eilam: Conceptualization, methodology, investigation, data interpretation, and original draft writing. Mahdi Arar: Data curation and investigation. Merav Greenstein: Statistical analysis, data interpretation, and writing – review and editing. Michael Zouari: Statistical analysis, data interpretation, and writing – review and editing. Volodymyr Samogalskyi: Data interpretation and writing – review and editing. Yacov Balash: Data interpretation and writing – review and editing. Nadejda Khmelev: Conceptualization, methodology, supervision, data interpretation, and original draft writing.
FUNDING INFORMATION
This study received no external funding.
CONFLICT OF INTEREST STATEMENT
Author Merav Greenstein has served as a paid consultant for statistical analysis and has received support from Kaplan Medical Center–Clalit Health Services. The remaining authors declare no Conflict of interest.
ETHICS STATEMENT
We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
Supporting information
Table S1. Observed follow‐up duration per period (N = 34 941). Pre‐war follow‐up was derived from the patient‐level variable pre‐days (actual observed days). Wartime follow‐up from post‐days. The survival variable days_km_final (survival time to first seizure‐related emergency department [ED] visit) reported in the article (mean 339.70, standard deviation [SD] 76.31) is censored at first seizure‐related ED visit and therefore differs slightly from the raw wartime follow‐up shown here.
Table S2. Reasons for incomplete follow‐up per period (N = 34 941). “Other membership gaps” represents residual cases whose follow‐up window was shortened by the insurance‐coverage–adjusted calculation but could not be attributed to a clear death or disenrollment event in the same period (e.g., membership interruptions flanking the study period). In a “wartime group” 16 patients had a late enrollment and early termination; therefore the total number of incomplete follow‐up patients is 3201 rather than 3217.
Table S3. Pre‐war Cox N = 30 821 (events = 2843). Wartime Cox N = 32 504 (events = 2903). Both models were adjusted for the same set of predictors. Gender hazard ratio (HR) expressed for female relative to male (reference = male in the Statistical Package for the Social Sciences [SPSS] indicator coding used). HRs > 1 indicate a shorter time to first seizure‐related emergency department (ED) visit.
Figure S1. Flow diagram of study cohort selection, seizure‐related emergency department (ED) visit distribution and paired analysis inclusion.
ACKNOWLEDGMENTS
AI–Artificial Intelligence tools were used for language editing only. The authors are fully responsible for the article content.
DATA AVAILABILITY STATEMENT
Data are available upon reasonable request.
REFERENCES
- 1. GBD Epilepsy Collaborators . Global, regional, and national burden of epilepsy, 1990‐2021: A systematic analysis for the global burden of disease study 2021. Lancet Public Health. 2025;10(3):e203–e227. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Salpekar JA, Basu T, Thangaraj S, Maguire J. The intersections of stress, anxiety and epilepsy. Int Rev Neurobiol. 2020;13(152):195–219. [DOI] [PubMed] [Google Scholar]
- 3. Temkin NR, Davis GR. Stress as a risk factor for seizures among adults with epilepsy. Epilepsia. 1984;25(4):450–456. [DOI] [PubMed] [Google Scholar]
- 4. HR MK, Privitera MD. Stress as a seizure precipitant: Identification, associated factors, and treatment options. Seizure. 2017;44:21–26. [DOI] [PubMed] [Google Scholar]
- 5. Haut SR, Hall CB, Masur J, Lipton RB. Seizure occurrence: Precipitants and prediction. Neurology. 2007;69(20):1905–1910. [DOI] [PubMed] [Google Scholar]
- 6. Lai C‐W, Trimble MR. Stress and epilepsy. J Epilepsy. 1997;10(4):177–186. [Google Scholar]
- 7. Hitiris N, Mohanraj R, Norrie J, Sills GJ, Brodie MJ. Predictors of pharmacoresistant epilepsy. Epilepsy Res. 2007;75(2–3):192–196. [DOI] [PubMed] [Google Scholar]
- 8. Kanner AM. Depression and epilepsy: A bidirectional relation? Epilepsia. 2011;52(Suppl 1):21–27. [DOI] [PubMed] [Google Scholar]
- 9. Joëls M. Stress, the hippocampus, and epilepsy. Epilepsia. 2009;50(4):586–597. [DOI] [PubMed] [Google Scholar]
- 10. Fitzgerald PJ. Is elevated norepinephrine an etiological factor in some cases of epilepsy? Seizure. 2010;19(6):311–318. [DOI] [PubMed] [Google Scholar]
- 11. Catalán‐Aguilar J, González‐Bono E, Cano‐López I. Perceived stress in adults with epilepsy: A systematic review. Neurosci Biobehav Rev. 2025;170:106065. [DOI] [PubMed] [Google Scholar]
- 12. Espinosa‐Garcia C, Zeleke H, Rojas A. Impact of stress on epilepsy: Focus on neuroinflammation‐a mini review. Int J Mol Sci. 2021;22(8):4061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Catalán‐Aguilar J, González‐Bono E, Villanueva V, Hampel KG, Lozano‐García A, Hidalgo V, et al. Buffered cortisol response to stress in patients with epilepsy and its association with memory and quality of life. Neurology. 2025;105(7):e214103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Shekhar A, Truitt W, Rainnie D, Sajdyk T. Role of stress, corticotrophin releasing factor (CRF) and amygdala plasticity in chronic anxiety. Stress. 2005;8(4):209–219. [DOI] [PubMed] [Google Scholar]
- 15. Jhaveri DJ, McGonigal A, Becker C, Benoliel J‐J, Nandam LS, Soncin L, et al. Stress and epilepsy: towards understanding of neurobiological mechanisms for better management. eNeuro. 2023;10(11):ENEURO0200232023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Gilboa T. Emotional stress‐induced seizures: another reflex epilepsy? Epilepsia. 2012;53(2):e29–e32. [DOI] [PubMed] [Google Scholar]
- 17. Haut SR, Vouyiouklis M, Shinnar S. Stress and epilepsy: A patient perception survey. Epilepsy Behav. 2003;4(5):511–514. [DOI] [PubMed] [Google Scholar]
- 18. Neugebauer R, Paik M, Hauser WA, Nadel E, Leppik I, Susser M. Stressful life events and seizure frequency in patients with epilepsy. Epilepsia. 1994;35(2):336–343. [DOI] [PubMed] [Google Scholar]
- 19. Christensen J, Li J, Vestergaard M, Olsen J. Stress and epilepsy: A population‐based cohort study of epilepsy in parents who lost a child. Epilepsy Behav. 2007;11(3):324–328. [DOI] [PubMed] [Google Scholar]
- 20. Novakova B, Harris PR, Ponnusamy A, Reuber M. The role of stress as a trigger for epileptic seizures: A narrative review of evidence from human and animal studies. Epilepsia. 2013;54(11):1866–1876. [DOI] [PubMed] [Google Scholar]
- 21. Lang JD, Taylor DC, Kasper BS. Stress, seizures, and epilepsy: Patient narratives. Epilepsy Behav. 2018;80:163–172. [DOI] [PubMed] [Google Scholar]
- 22. Gélisse P, Genton P, Coubes P, Tang NPL, Crespel A. Can emotional stress trigger the onset of epilepsy? Epilepsy Behav. 2015;48:15–20. [DOI] [PubMed] [Google Scholar]
- 23. Neufeld MY, Sadeh M, Cohn DF, Korczyn AD. Stress and epilepsy: The Gulf war experience. Seizure. 1994;3(2):135–139. [DOI] [PubMed] [Google Scholar]
- 24. Klein P, van Passel L. Effect of stress related to the 9/11/2001 terror attack on seizures in patients with epilepsy. Neurology. 2005;64(10):1815–1816. [DOI] [PubMed] [Google Scholar]
- 25. Bosnjak J, Vukovic‐Bobic M, Mejaski‐Bosnjak V. Effect of war on the occurrence of epileptic seizures in children. Epilepsy Behav. 2002;3(6):502–509. [DOI] [PubMed] [Google Scholar]
- 26. Doron A, Fahoum F, Goldstein L. Stress and seizures: The October 7th experience. Epilepsy Behav. 2025;173:110560. [DOI] [PubMed] [Google Scholar]
- 27. Amsalem D, Haim‐Nachum S, Lazarov A, Levi‐Belz Y, Markowitz JC, Bergman M, et al. The effects of war‐related experiences on mental health symptoms of individuals living in conflict zones: A longitudinal study. Sci Rep. 2025;15(1):889. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. De Cauwer H, Somville FJMP. Neurological disease in the aftermath of terrorism: A review. Acta Neurol Belg. 2018;118(2):193–199. [DOI] [PubMed] [Google Scholar]
- 29. Appel S, Gelfand A, Kotlyar G, Milo R. A retrospective analysis of the development of seizure frequency in patients with seizures during a period of military conflict. Seizure. 2018;61:119–121. [DOI] [PubMed] [Google Scholar]
- 30. Swinkels WA, Engelsman M, Kasteleijn‐Nolst Trenité DG, Baal MG, de Haan GJ, Oosting J. Influence of an evacuation in February 1995 in The Netherlands on the seizure frequency in patients with epilepsy: A controlled study. Epilepsia. 1998;39(11):1203–1207. [DOI] [PubMed] [Google Scholar]
- 31. Shibahara I, Osawa S‐I, Kon H, Morita T, Nakasato N, Tominaga T, et al. Increase in the number of patients with seizures following the great East‐Japan earthquake. Epilepsia. 2013;54(4):e49–e52. [DOI] [PubMed] [Google Scholar]
- 32. Inatomi Y, Nakajima M, Yonehara T, Ando Y. Clinical characteristics of patients seizure following the 2016 Kumamoto earthquake. J Clin Neurosci. 2017;40:123–129. [DOI] [PubMed] [Google Scholar]
- 33. Moshe S, Shilo M, Chodick G, Yagev Y, Blatt I, Korczyn AD, et al. Occurrence of seizures in association with work‐related stress in young male army recruits. Epilepsia. 2008;49(8):1451–1456. [DOI] [PubMed] [Google Scholar]
- 34. Detailed findings on attacks carried out on and after 7 October 2023 in Israel. https://digitallibrary.un.org/record/4051246?v=pdf
- 35. Codish S, Frenkel A, Klein M, Geftler A, Dreiher J, Schwarzfuchs D. October 7th 2023 attacks in Israel: Frontline experience of a single tertiary center. Intensive Care Med. 2024;50(2):308–310. [DOI] [PubMed] [Google Scholar]
- 36. Shamoa‐Nir DL. Psychological trauma of evacuated Israeli residents following the October 7th Hamas terror attack. Psychiatry Res. 2024;342:116210. [DOI] [PubMed] [Google Scholar]
- 37. Feingold D, Neria Y, Bitan DT. PTSD, distress and substance use in the aftermath of October 7th, 2023, terror attacks in southern Israel. J Psychiatr Res. 2024;174:153–158. [DOI] [PubMed] [Google Scholar]
- 38. Levi‐Belz Y, Shoval‐Zuckerman Y, Blank C, Groweiss Y, Neria Y. The moderating role of belongingness in the contribution of depression to suicide ideation following the October 7, 2023, terrorist attack in Israel: A nationwide prospective study. J Affect Disord. 2024;356:292–299. [DOI] [PubMed] [Google Scholar]
- 39. Spooner J, Bandyopadhyay S, Baraka M, Sander JW. The impact of socioeconomic status on the management of epilepsy: A qualitative study. Epilepsy Behav. 2025;172:110685. [DOI] [PubMed] [Google Scholar]
- 40. Elliott JO, Lu B, Shneker BF, Moore JL, McAuley JW. The impact of “social determinants of health” on epilepsy prevalence and reported medication use. Epilepsy Res. 2009;84(2–3):135–145. [DOI] [PubMed] [Google Scholar]
- 41. Li X, Sundquist J, Sundquist K. Socioeconomic and occupational risk factors for epilepsy: A nationwide epidemiological study in Sweden. Seizure. 2008;17(3):254–260. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Pepi C, Mercier M, Salimbene L, Galati C, Specchio N, de Palma L. Post‐traumatic stress‐disorder in epilepsy: Meta‐analysis of current evidence. Epilepsy Behav. 2024;157:109833. [DOI] [PubMed] [Google Scholar]
- 43. Soncin L‐D, McGonigal A, Kotwas I, Belquaid S, Giusiano B, Faure S, et al. Post‐traumatic stress disorder (PTSD) in patients with epilepsy. Epilepsy Behav. 2021;121(A):108083. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. Observed follow‐up duration per period (N = 34 941). Pre‐war follow‐up was derived from the patient‐level variable pre‐days (actual observed days). Wartime follow‐up from post‐days. The survival variable days_km_final (survival time to first seizure‐related emergency department [ED] visit) reported in the article (mean 339.70, standard deviation [SD] 76.31) is censored at first seizure‐related ED visit and therefore differs slightly from the raw wartime follow‐up shown here.
Table S2. Reasons for incomplete follow‐up per period (N = 34 941). “Other membership gaps” represents residual cases whose follow‐up window was shortened by the insurance‐coverage–adjusted calculation but could not be attributed to a clear death or disenrollment event in the same period (e.g., membership interruptions flanking the study period). In a “wartime group” 16 patients had a late enrollment and early termination; therefore the total number of incomplete follow‐up patients is 3201 rather than 3217.
Table S3. Pre‐war Cox N = 30 821 (events = 2843). Wartime Cox N = 32 504 (events = 2903). Both models were adjusted for the same set of predictors. Gender hazard ratio (HR) expressed for female relative to male (reference = male in the Statistical Package for the Social Sciences [SPSS] indicator coding used). HRs > 1 indicate a shorter time to first seizure‐related emergency department (ED) visit.
Figure S1. Flow diagram of study cohort selection, seizure‐related emergency department (ED) visit distribution and paired analysis inclusion.
Data Availability Statement
Data are available upon reasonable request.