Out‐of‐hospital versus in‐hospital status epilepticus: The role of etiology and comorbidities

Abstract Background and purpose Our objectives were to identify differences in clinical characteristics between patients with out‐of‐hospital and in‐hospital status epilepticus (SE) onset, and to evaluate the influence of SE onset setting on 30‐day mortality and SE cessation. Methods We included consecutive patients with SE admitted from 2013–2021 at Modena Academic Hospital. A propensity score was obtained with clinical variables unevenly distributed between the two groups. Results Seven hundred eleven patients were included; 55.8% (397/711) with out‐of‐hospital and 44.2% (314/711) with in‐hospital onset. Patients with in‐hospital SE onset were older and had a higher frequency of comorbidities, acute and/or potentially fatal etiologies, impaired consciousness before treatment, and nonconvulsive or myoclonic SE. No difference was found in SE cessation between the groups. Patients with in‐hospital SE had higher 30‐day mortality (127/314, 62.9% vs. 75/397, 37.1%; p < 0.001). In‐hospital onset was an independent risk factor for 30‐day mortality (adjusted odds ratio = 1.720; 95% confidence interval = 1.107–2.674; p = 0.016). In the propensity group (n = 244), no difference was found in 30‐day mortality and SE cessation between out‐of‐hospital and in‐hospital SE onset groups (36/122, 29.5% vs. 34/122, 27.9%; p = 0.888; and 47/122, 38.5% vs. 39/122; 32%; p = 0.347, respectively). Conclusions In‐hospital SE is associated with higher 30‐day mortality without difference in SE cessation. The two groups differ considerably for age, acute and possibly fatal etiologies, comorbidities, and SE semiology. The patient location at SE onset is an important prognostic predictor. However, the increased mortality is probably unrelated to the setting of SE onset and reflects intrinsic prognostic predictors.

recognized and adequately treated to minimize the risk of long-term consequences and mortality.
Several factors predicting short-term mortality have been identified, and some of them have been incorporated into outcome scoring systems like the Status Epilepticus Severity Score (STESS) [3] and the Epidemiology-Based Mortality Score in Status Epilepticus (EMSE) [4]. The main prognostic factors include age, level of consciousness impairment, SE semiology, history of previous seizures, etiology, comorbidities, and electroencephalographic (EEG) patterns. The knowledge of the factors that predict the outcome is important, because their control could guide treatment strategies and improve prognosis. For instance, identification of the underlying cause of SE through a rapid diagnostic workup is crucial for effective etiological treatment [5]. Therefore, considering the clinical context in which SE occurs may shorten diagnostic assessment and allow proper and tailored management, possibly reducing morbidity and mortality.
Studies comparing the clinical characteristics and prognosis of patients with out-of-hospital SE onset and those with SE developed during the hospital stay (de novo in-hospital onset) are scarce [6,7]. One comparative study showed that patients with in-hospital SE had more fatal etiologies and comorbidities, refractory SE, less return to functional baseline, and increased mortality compared to patients with out-of-hospital SE [7]. The difference in prognosis could be explained by the underlying etiology and comorbid medical conditions, as patients with in-hospital SE onset are more likely to have been admitted with acute or potentially fatal brain injuries, metabolic disturbances, or comorbidities that may influence the outcome.
Although patients already admitted to the hospital have more comorbidities and acute/possibly fatal etiologies, one could argue that in-hospital SE onset is recognized and treated more promptly compared to SE occurring out-of-hospital. Similar to cardiac arrest, which has a better prognosis if occurring in the hospital compared to out of the hospital [8], a prompt diagnosis and treatment could counterbalance the intrinsically negative prognostic features associated with in-hospital SE onset.
This study aimed to identify differences in clinical characteristics between patients with out-of-hospital and de novo in-hospital SE onset and to evaluate the influence of SE onset setting on 30-day mortality and SE cessation.

Study design, setting, and patients
We reviewed consecutive episodes of SE occurring in patients aged 14 years and older and prospectively registered at Baggiovara Civil Hospital (Modena, Italy) from September 1, 2013 to October 31, 2021. Before 2015, SE was considered to be a continuous seizure that lasts 5 min or longer or two or more discrete seizures without complete recovery of consciousness between them [9]. After 2015, the definition by the International League Against Epilepsy (ILAE) was systematically adopted and prospectively applied [1]. Accordingly, the operational time indicating when a seizure is likely to be prolonged leading to continuous seizure activity (i.e., SE), was set at 5 min for tonic-clonic SE, 10 min for focal SE with impaired consciousness, and 10-15 min for absence SE. All cases of SE that occurred before 2015 were reviewed by two of the authors (S.M. and G.G.) to ensure that all met the ILAE diagnostic criteria. The cases of nonconvulsive SE were diagnosed according to the Salzburg EEG criteria [10,11].
A specific data set was used to collect demographic and clinical information, including age, gender, setting of SE onset (outof-hospital or in-hospital), medical history and comorbid medical conditions, prior history of epilepsy, etiological ILAE classification [1] in which acute symptomatic causes were divided into hypoxic or nonhypoxic, impairment of consciousness before treatment, worst seizure type according to the STESS [3], EEG patterns according to the EMSE [4], scores of the EMSE and STESS [3,4], and modified Rankin Scale (mRS) at SE onset, at discharge, and after 30 days. The form was filled in by the first physician (neurologist or neurointensivist) taking care of the patient.
In our hospital, every patient with a suspicion or diagnosis of SE is referred to the consultant neurologist, both for diagnostic confirmation and treatment. The consultant neurologist and an EEG recording are available 24 h a day, 7 days per week, even in the intensive care units.

Outcome
SE cessation was defined according to the Sustained Effort Network for Treatment of Status Epilepticus (SENSE) study as follows: cessation of SE within the first hour after treatment initiation for generalized convulsive SE and cessation of SE within 12 h after treatment initiation for other SE types [15].
Data on follow-up of patients, SE cessation, and their 30-day mortality were obtained from the SE data set used to collect information and confirmed through the registry office.

Statistical analysis
The categorical variables were described as percentage and number of events out of the total, and the univariate comparisons were performed with the Fischer exact test or the χ 2 test. Continuous variables were reported as median and interquartile range (IQR) or as mean and SD, depending on the underlying distribution.
Comparisons were made with the Mann-Whitney or t tests. The possible independent association between in-hospital SE onset and 30-day mortality was studied through a multivariate model with logistic regression and adjusted stepwise method for variables that had been found significant in univariate analyses and included as possible multivariate confounders. Survival analysis between the two treatment groups (out-of-hospital and in-hospital SE onset) was conducted with the Kaplan-Meier method through the log-rank test.
Subsequently, considering the different baseline conditions that could have influenced the outcome and the fact that the study did not have preenrollment selection criteria, the baseline variables that were found to be unbalanced between the two groups were included in a propensity score matching.
After propensity score matching, the analyses were repeated to estimate the independent association between SE onset setting and 30-day mortality. All tests were two-sided, and a p value <0.050 was considered statistically significant. Statistical analyses were performed with Stata version 16.0 (StataCorp).

Data availability
Upon request from qualified investigators, we will share anonymized data.

RE SULTS
Seven hundred eleven patients with SE observed during the study period were included. There were 55.8% (397/711) who had an outof-hospital SE onset, whereas 44.2% (314/711) had an in-hospital onset. Patient baseline characteristics were strongly unbalanced between the two study groups ( Table 1). Patients with de novo inhospital SE onset were older (median = 76 years, IQR = 66-82 vs. 71 years, IQR = 57-81; p < 0.001) and had a higher frequency of ischemic heart disease, cerebrovascular disease, diabetes mellitus, heart failure, dementia, peripheral vascular disease, and chronic kidney disease. A previous history of epilepsy was found more frequently among patients with in-hospital onset (230, 73.2% vs. 231, 58.2%; p < 0.001). Acute symptomatic causes (hypoxic and nonhypoxic) were more frequent in patients with in-hospital SE, whereas remote and progressive symptomatic causes were found more frequently among patients with out-of-hospital onset. An impaired consciousness before treatment and nonconvulsive SE occurred more frequently among in-hospital onset patients, whereas generalized convulsive SE, focal motor, focal non-motor, absence, and myoclonic SE complicating idiopathic generalized epilepsy were more frequent in the out-of-hospital group. A higher proportion of EEG patterns with after status ictal discharges, generalized sharply and/or triphasic period potentials, lateralized periodic discharges, or spontaneous burst suppression was found among patients with in-hospital SE onset.
The severity of SE, evaluated by the STESS and EMSE, was higher in those with in-hospital onset. No difference was found in mRS at baseline, whereas it was higher at discharge and at 30 days among patients with in-hospital SE onset.
Patients with in-hospital onset had higher 30-day mortality log-rank test p < 0.001) (Figures 1 and 2).
Characteristics associated with 30-day mortality are reported in Table 2. Of note, nonconvulsive SE occurred more frequently in nonsurvivors, whereas generalized convulsive SE, focal motor, focal nonmotor, absence, and myoclonic SE complicating idiopathic generalized epilepsy were more frequent in survivors. Comorbidities associated with increased mortality were ischemic heart disease, cerebrovascular disease, dementia, ulcer, tumor, peripheral vascular disease, chronic obstructive pulmonary disease, chronic kidney disease, and liver failure. Hypoxic acute symptomatic causes were more frequent among nonsurvivors, whereas remote and progressive symptomatic causes were more frequent in survivors.
In the multivariate analysis performed with a logistic regression using the stepwise regression method, the in-hospital onset was an independent risk factor for mortality at 30 days, with an adjusted odds ratio of 1.720 (95% confidence interval = 1.107-2.674; A propensity score was performed with the clinical variables found to be differently distributed between the two groups. A oneto-one statistical matching was carried out, obtaining a restricted data set of 244 patients equally distributed into 122 pairs (i.e., 122 out-of-hospital and 122 in-hospital SE onset). The characteristics of the patient group obtained after propensity score matching are reported in Table 3.
In the propensity group (n = 244), no difference was found in 30-

DISCUSS ION
This study has identified clinical and prognostic differences between patients with out-of-hospital and de novo in-hospital SE onset.
SE cessation did not differ between groups, whereas 30-day mortality was higher among patients with in-hospital onset.
In this study, we pragmatically defined SE treatment response according to criteria proposed in the SENSE registry. We chose this definition for two reasons. The first is that the SENSE study evaluated the pharmacological treatment of SE in a real-life setting, as in our study. The second reason is that we could expect lower and later response rates in the out-of-hospital onset group compared to the in-hospital SE onset, and the definition of SE cessation provided by the SENSE registry appears particularly appropriate for the present study. With regard to this issue, in our study the percentage of patients who met the SE cessation criterion was very close to that reported in the SENSE study, and as previously commented by the authors of the SENSE study, lower than SE treatment response reported in clinical trials [16][17][18][19][20]. Second, unlike what could be expected, the intra-or extrahospital onset was not associated with a different response to pharmacological treatment.
The higher mortality in patients with in-hospital SE onset observed in our population (about 40%) is consistent with data available in the literature. One study found that in this patient group the mortality (60%) was higher than the overall mortality rate associated with refractory SE derived from prior studies including mixed (i.e., in-hospital and out-of-hospital SE) populations [6]. A subsequent 10-year comparative cohort study showed that in-hospital SE was associated with short-term mortality of about 30% and with more potentially fatal etiologies and comorbid conditions, longer duration and treatment refractoriness, and more infrequent return to functional baseline [7].
The poor outcome among patients with in-hospital SE onset can be explained by the fact that these patients are admitted for an intrinsically severe or critical condition, usually for acute or potentially fatal brain injuries, metabolic disturbances, or comorbidities. Each of these factors is associated with a high risk of death.
The combination of structural brain abnormalities and metabolic disturbances appears to be particularly predictive of short-term mortality [21].
In our study, patients with in-hospital SE onset were older, confirming the finding that age is a nonmodifiable predictor of mortality following SE, as consistently reported in the literature [22][23][24][25][26].
Differences in comorbidities were found between the two groups and were more frequent among patients with in-hospital onset. Of note, no comorbidity was found more often in the out-of-hospital group. Attention should be paid to specific comorbidities associated with poor outcomes; treating comorbidities could, hence, lead to a better prognosis reducing the risk of short-term mortality.
Furthermore, patients with out-of-hospital and in-hospital SE onset differed markedly with regard to underlying etiologies. In the latter group, acute and/or potentially fatal SE causes occurred more frequently [26][27][28][29].
The higher mortality rate in patients with in-hospital SE can be explained by the older age, higher morbidity, higher prevalence of periodic EEG patterns or spontaneous burst suppression, and a higher proportion of possibly fatal etiologies, but also by SE semiology.
Interestingly, generalized convulsive SE occurred more frequently in the out-of-hospital onset and survivors, whereas non-convulsive SE was more frequent in the in-hospital group and in patients who died within 30 days from SE. These findings are in line with the results of a large retrospective population-based study, which showed that prominent motor phenomena are associated with lower mortality; in this study, convulsive semiology at the end of SE, or as its only semiology had a case fatality of zero, whereas impaired awareness with somnolence, stupor, or coma was associated with higher mortality compared to awake patients [30]. The different prognosis can be explained by the fact that generalized convulsive SE with bilateral tonic-clonic activity is probably recognized much more quickly by bystanders, leading to prompter treatment than nonconvulsive SE.
Furthermore, unlike the latter, SE with generalized prominent motor phenomena could reflect a still preserved cerebral ability to seize without homeostatic metabolic or electric exhaustion [31].
Compared to out-of-hospital onset, the in-hospital onset was an independent risk factor for mortality at 30 days, but also for poorer functional outcome, as shown by higher mRS at discharge and at 30 days. Although patients already admitted to the hospital The term corresponds to the terminology reported in the current definition of SE [1]. The original term reported in the STESS was simple-partial [3]. b The term corresponds to the terminology reported in the current definition of SE [1]. The original term reported in the STESS was complex-partial [3]. c SE cessation was defined according to the Sustained Effort Network for Treatment of Status Epilepticus study as follows: cessation of SE within the first hour after treatment initiation for generalized convulsive SE and cessation of SE within 12 h after treatment initiation for other SE types [15].
have more comorbidities and acute/possibly fatal etiologies, one could argue that in-hospital SE onset is recognized and treated more promptly compared to SE occurring out-of-hospital. Similar to cardiac arrest, which has a better prognosis if occurring in the hospital compared to outside the hospital [8], prompter diagnosis and treatment could counterbalance the intrinsically more negative prognostic features associated with in-hospital SE onset. After balancing the clinical variables that were differently distributed between the out-of-hospital and the in-hospital SE onset groups by the propensity score matching, no difference was found in 30-day mortality or functional outcome assessed with mRS. This further confirms that the different prognosis is likely to be unrelated to the setting of SE onset but reflects intrinsic differences in relevant prognostic factors between the groups. This study has a few limitations. This study was conducted in one single tertiary care center, possibly limiting the generalizability of the findings. We investigated episodes of SE instead of patients with first SE, which might have biased the results. We had no data on total SE duration and on time of SE onset. This is particularly relevant for non convulsive status epilepticus (NCSE); however, in this condition it is very challenging and sometimes impossible to identify with precision the time of SE onset. In patients with out-ofhospital hypoxic SE, the diagnosis was made after the patient was admitted to the hospital; the etiological diagnosis of hypoxic SE due  The term corresponds to the terminology reported in the current definition of SE [1]. The original term reported in the STESS was simple-partial [3]. b The term corresponds to the terminology reported in the current definition of SE [1]. The original term reported in the STESS was complex-partial [3]. c SE cessation was defined according to the Sustained Effort Network for Treatment of Status Epilepticus study as follows: cessation of SE within the first hour after treatment initiation for generalized convulsive SE and cessation of SE within 12 h after treatment initiation for other SE types [15].
to cardiorespiratory arrest was hence made retrospectively. Finally, we did not have enough data to calculate the Charlson Comorbidity Index score [32], which is useful for assessing the burden of comorbidities, and had no information on single EEG patterns. Hence, further studies conducted in different cohorts and hospital settings are required to confirm our findings.
The major strengths of this study include the large sample size and the use of propensity score matching, a statistical sampling TA B L E 3 Characteristics of the data set obtained after propensity score matching The term corresponds to the terminology reported in the current definition of SE [1]. The original term reported in the STESS was simplepartial [3]. b The term corresponds to the terminology reported in the current definition of SE [1]. The original term reported in the STESS was complex-partial [3]. c SE cessation was defined according to the Sustained Effort Network for Treatment of Status Epilepticus study as follows: cessation of SE within the first hour after treatment initiation for generalized convulsive SE and cessation of SE within 12 h after treatment initiation for other SE types [15].

TA B L E 3 (Continued)
technique that can limit the selection bias of two study groups.
Through propensity score matching we obtained a subgroup of patients adequately matched 1 to 1 between the two study groups (out-of-hospital and in-hospital SE onset) to obtain a balance in basal characteristics.
In our hospital, patients from various departments in whom there is a suspicion of SE are referred to the consultant neurologist, available 24 h a day, 7 days per week, both for diagnostic confirmation and treatment following an internal standardized protocol; hence, there are almost no differences according to the department that initially evaluates the patients. The management of patients in the intensive care units also involves the consultant neurologist.
This greatly reduces the heterogeneity in the diagnostic process and treatment of patients with in-hospital SE onset. Conversely, it is possible that the initial management of out-of-hospital SE cases is affected by higher heterogeneity in treatment.
In conclusion, our study has shown that patients with in-hospital SE are at a higher risk of 30-day mortality compared to those with out-of-hospital SE. The two groups differ considerably in terms of age, acute and possibly fatal etiologies, comorbidities, and SE semiology.
The patient location at SE onset has important prognostic relevance.
However, the increased mortality is probably unrelated to the setting

ACKNOWLEDG EMENT
Open Access Funding provided by Universita degli Studi di Modena e Reggio Emilia within the CRUI-CARE Agreement.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Anonymized data that support the findings of this study are available from the corresponding author upon reasonable request.