Abstract
Terman SW, Guterman EL, Lin CC, Thompson MP, Burke JF. Epilepsia. 2024 May;65(5):1415-1427. Epub 2024 Feb 26. PMID: 38407370; PMCID: PMC11087197. doi: 10.1111/epi.17927 Objective: Understanding factors driving variation in status epilepticus outcomes would be critical to improve care. We evaluated the degree to which patient and hospital characteristics explained hospital-to-hospital variability in intubation and postacute outcomes. Methods: This was a retrospective cohort study of Medicare beneficiaries admitted with status epilepticus between 2009 and 2019. Outcomes included intubation, discharge to a facility, and 30- and 90-day readmissions and mortality. Multilevel models calculated percent variation in each outcome due to hospital-to-hospital differences. Results: We included 29 150 beneficiaries. The median age was 68 years (interquartile range [IQR] = 57-78), and 18 084 (62%) were eligible for Medicare due to disability. The median (IQR) percentages of each outcome across hospitals were: 30-day mortality 25% (0%-38%), any 30-day readmission 14% (0%-25%), 30-day status epilepticus readmission 0% (0%-3%), 30-day facility stay 40% (25%-53%), and intubation 46% (20%-61%). However, after accounting for many hospitals with small sample size, hospital-to-hospital differences accounted for 2% to 6% of variation in all unadjusted outcomes, and approximately 1% to 5% (maximally 8% for 30-day readmission for status epilepticus) after adjusting for patient, hospitalization, and/or hospital characteristics. Although many characteristics significantly predicted outcomes, the largest effect size was cardiac arrest predicting death (odds ratio = 10.1, 95% confidence interval = 8.8-11.7), whereas hospital characteristics (eg, staffing, accreditation, volume, setting, and services) all had lesser effects. Significance: Hospital-to-hospital variation explained little variation in studied outcomes. Rather, certain patient characteristics (eg, cardiac arrest) had greater effects. Interventions to improve outcomes after status epilepticus may be better focused on individual or prehospital factors, rather than at the inpatient systems level.
Commentary
Status epilepticus (SE) is the second most frequent nontraumatic neurological emergency, affecting 10 to 40 per 100 000 persons annually. It is associated with high morbidity and mortality rates. Most factors associated with outcome are nonmodifiable. They include patients’ characteristics (age and co-morbidities) and the etiology and semiology of SE. Several randomized controlled trials have progressively informed us on the first 2 steps of management, that is, administration of a parenteral benzodiazepine, eventually followed by intravenous administration of a nonsedating anti-seizure medication (ASM). 1 These first 2 steps typically occur out of the hospital or in the emergency room and, together, can achieve seizure control in 80% to 85% of cases. Although clearly defined, these initial steps are subject to frequent deviations from guideline in practice. As many as 20% patients do not receive a first-line benzodiazepine and most of those who receive one receive it too late, and at a dose lower than recommended.2,3
When the first 2 lines of treatment fail to control SE, it is deemed refractory, in which case the patient is usually admitted to an intensive care unit (ICU) for continuous administration of anesthetic drugs under continuous electroencephalography (EEG) monitoring. 1 Other in-hospital steps of care include the management of the underlying etiology and of the complications that might arise during the ICU stay, either from SE itself or from its treatment. While guidelines also cover these steps, they rely on much thinner evidence. Variations in practice between different centers is thus expected, possibly to an even larger degree than for initial management steps. Studying this variation, and how it influences outcomes, could inform us on how to improve SE care.
In this study, 4 the authors harness Medicare claims data to study the variability in mortality, readmission, and intubation rates in 29 150 patients who were admitted for SE between 2009 and 2019 at 2538 different hospitals. Slightly more than 60% patients had a history of epilepsy. Conversely, almost 40% had an acute etiology, including 8% with postanoxic SE. Fourteen percent were transferred from another institution, nearly 50% were intubated, and 27% were admitted to an ICU. Due to the source of data, the cohort comprises only elderly patients or patients with disability. This is a possible bias, although, as mentioned by the authors, the effect of practice variations might be emphasized in this more fragile group. Most hospitals were medium to large size academic centers, located in a metropolitan area and had both an ICU and a neurological department. The 30-day mortality rate was 25% and readmission rate for SE was 4%. In multivariate analyses, outcomes mostly depended on patient or hospitalization features: age, co-morbidities, lack of history of epilepsy, and acute brain injury (especially anoxic brain injury), being transferred from another institution, use of EEG, and intubation. Surprisingly, hospital features did not seem to matter much. A higher number of ICU beds was associated with a slightly higher rate of intubation and slightly lower mortality rate. A greater number of annual SE admissions was associated with a lower intubation rate and tended to be associated with lower mortality, but the latter did not reach significance. Overall, most of the variability of the mortality rate between hospitals was explained by differences in patient and hospitalization features rather than by differences in hospital features. In fact, hospital features contributed to only a few percent of the overall variability in mortality, when adjusting for patient and hospitalization features. Only the rate of readmission for SE seemed to be significantly affected by hospital features.
The study has some limitations. The effect of neurological ICU beds was not reported. This would have been of interest as these units are designed to provide specific care to patients with neurological emergencies, perhaps better so than in general ICU. 5 Given the strong contribution of postanoxic cases to mortality and intubation rates—it is likely that most of them were intubated and died—it would have been interesting to perform the same analyses after excluding this subgroup. As a side note, differentiating “true” postanoxic SE from other conditions such as status myoclonus is not trivial in clinical practice, and it might be even more difficult using claims data. More generally, outcome studies in SE are difficult because of the wide spectrum of semiology and severity, including degree of refractoriness. All these aspects dictate the need for in-hospital care and influence outcome. But they were not directly explored here. Further studies should focus on subtypes of SE that depend more on in-hospital management, such as refractory SE or nonconvulsive SE in comatose patients.
In the meantime, should we stop investigating in-hospital SE management and focus only on prehospital management? I would argue we still need to do both. The limited variability between hospitals suggests that all of them manage SE similarly. But that does not mean that they all do it right and that there is no room for substantial improvement. The evidence behind the guidelines for the in-hospital management steps of SE is still scarce. In the field of stroke medicine, hospital-to-hospital variation in outcome became more obvious when evidence for efficacious interventions became available, incorporated into time-constrained guidelines, and monitored with key performance measures.6,7 For in-hospital SE management, interventions of interest might include the choice of and time to third-line therapy, the use of and time to EEG or continuous EEG monitoring. “Seizure onset to treatment(s)” and “seizure onset to EEG” times should be routinely documented in medical records. There is a surprising dearth of quality improvement studies in SE. In fact, a Pubmed search on May 17, 2024 revealed only 8 studies, between 2016 and 2024 and almost all in the pediatric age group. The clinicaltrials.gov website only lists one ongoing study (NCT06194747). We ought to do more. In one such study, the implementation of a SE alert code led to a 30min reduction in the time to second-line treatment administration. 8 In another study, expediting EEG review and facilitating drug access for nonconvulsive SE reduced treatment delays and, more importantly, mortality. 9 Both were performed in large academic medical centers indicating that even in this setting, and likely every other setting, in-hospital SE care is not yet as efficient as it could be. There is thus hope that with some effort toward more efficiency, we might provide better in-hospital care and improve the outcome of SE.
