Hello, I'm Jonathan Parker, a fifth-year neurosurgery resident at Stanford, and I'm going to be sharing with you today some work we've done on the incidence and outcome of perioperative status epilepticus after intracranial neurosurgery. I have no disclosures. I'd like to start this talk with a case example of a patient that I saw as a junior resident. Briefly, this is a 65-year-old male with a right-sided sphenoid wing meningioma who presented to us with altered mental status and progressive visual deficit. He underwent a right tyrional cardiotomy. I was called to his bedside on the evening of postoperative day four for acute onset of altered mental status, and after a neurologic exam and metabolic workup and CT imaging were largely unrevealing for the extent of his altered mental status, a continuous EEG was obtained, which demonstrated the patient was in nonconvulsive status epilepticus. Unfortunately, this patient required uptitration of multiple antiepileptic drugs, including intubation and transfer to the ICU, and experienced a long ICU course and suffered multiple complications in the ICU, which eventually resulted in the patient's demise approximately two months after surgery. This case example inspired me to understand better what risk factors potentially can be modified in terms of risk of post-neurosurgical status epilepticus. Status epilepticus is a rare complication after intracranial neurosurgery, and no doubt those of you listening may have experienced some isolated cases of your own in your own practice. However, this has been reported in the literature several times before, both actually in cases of meningioma resection. We believe this is a significant event to study, although rare because of its association with morbidity, mortality, increased healthcare costs, and even long-term quality of life. The risk factors for which patients will go on to develop post-neurosurgical status epilepticus are poorly understood. We hypothesized that both pathology-specific and procedure-specific risk factors may be associated with post-neurosurgical status epilepticus risk. We thus chose to asset this in a nationwide cross-sectional database to try to capture variability in neurosurgical procedures across different geographies and practice settings. We chose to do this in the IBM Watson Health MarketScan database, which captures over 250 million individual records of patients, is largely a private insurance-based database, and also includes dependents and has some Medicare patients as well. This has really been a proving ground for neurosurgical hypotheses, and the outcomes group at Stanford has used this database extensively, which really provides an opportunity to generate a large neurosurgical cohort, which is otherwise difficult to do in a single institution fashion. Inclusion criteria for our study were as follows. All patients had to be above 18 years of age and have an eligible intracranial neurosurgical procedure between 2007 and 2015. In the groupings that are seen on the right, these were assigned via CPT codes. We also excluded patients that underwent craniotomies for epilepsy-related indications. We defined post-neurosurgical status epilepticus in two ways, one both immediate and delayed. Immediate post-neurosurgical status epilepticus had to occur during the index admission, and delayed status epilepticus were patients that were readmitted to the hospital after initial index procedure with a primary diagnosis of status epilepticus. Status epilepticus was coded on the first day that it appeared in the administrative database, and we used both ICD-9 and ICD-10 codes to control for variable application across hospitals and time in terms of coding standards. Importantly, there's a manual validation study that demonstrated a positive predictive value of greater than 75% for grand mal status coding between what appears in the chart and what appears in a large administrative database, which gives some support overall for this method. In approximately 59,000 admissions, the patient had a primary procedure of tumor resection. Among these admissions, approximately 61% of the time, we could identify a specific underlying tumor pathology. We chose to break this down into meningioma, metastasis, and primary brain tumor. These are some exemplary ICD-9 and 10 codes that we used to define these subgroupings. In terms of our cohort, we captured approximately 197,000 admissions, again demonstrating the power of these large administrative databases, covering approximately 218,000 procedures with 15% of patients having a prior history of seizure or status epilepticus based off of ICD-9 and 10 codes. As you can see here, overall, the rate of immediate status epilepticus was low at 0.3%, and those patients developing delayed status epilepticus after their initial discharge from the hospital was approximately 0.6%. On the right, you can see the breakdown in percentages of the various craniotomy codes and groupings. This largely follows the expected distribution of the incidence of neurosurgical procedures with tumor resections being the most common at 29.9%, closely followed thereafter by trauma, hematoma, and intracranial pressure-related craniotomies. We first looked at the incidence of immediate status epilepticus over time in our cohort. As you can see in the graph on the left, there was a significant increase over time in the incidence of status epilepticus. When we looked at inpatient mortality of those admissions that had immediate status epilepticus versus all neurosurgical admissions, you could see that there was a relative increase in mortality in those patients, although this did not change significantly over time. A potential explanation of the increased recognition of status epilepticus over time in this cohort can be variances in the neurosurgical pathology that is captured, or this may represent an underlying difference in the coding or capture of status events. We next looked at specific risk factors which may modify the incidence of immediate status epilepticus. We looked at this in two fashions, one on the left in a Kaplan-Meier analysis where an event was considered the first day that a status epilepticus occurred. This was on average four and a half days after the index procedure. As you can see in both the red and the dotted blue lines, this represented infection and trauma, hematoma, and elevated intracranial pressure cases respectively. In our Cox proportional hazards analysis, we found that trauma, hematoma had the highest elevated increased risk factor for this complication, although this did not reach significance for infection, likely because of the smaller case numbers. A similar analysis was then performed for delayed status epilepticus cases. As you can see here again on the left with the Kaplan-Meier curves, both infection and now CSF diversion procedures were associated with the highest risk of delayed status epilepticus. And again, supratentorial tumor location was associated with increased risk, although no operative characteristics as you can see listed here passed significance. We then chose to look specifically at a subpopulation of patients that underwent tumor resection. As you can see here, the risk of immediate post neurosurgical status epilepticus in this group was highest for those undergoing meningioma resection, similar to the patient we discussed in the case example at the beginning of the talk. At delayed status epilepticus, there was a different risk profile, and the patients with primary brain tumor resections had a highest risk of delayed status epilepticus, which happened on average approximately a year and a half after discharge from their index procedure. Lastly, we went on to look at metrics of how this affected their inpatient stay. So we looked at length of stay in terms of ICU and overall hospital length of stay. As one would imagine, those with immediate post neurosurgical status epilepticus had longer ICU stays and hospital stays. And even when you are readmitted for a status epilepticus, a higher percentage of these patients had to go to the ICU is approximately 75% versus 39% of those that were readmitted for other reasons. And when they did go to the ICU, they had a longer stay. Additionally, we looked at the use of continuous EEG, which in the study was defined as an EEG that was at least done overnight, versus spot EEG, which was an EEG of one hour or less. You can see that over the study interval in both cases of immediate and delayed status epilepticus, the use of continuous EEG significantly increases with mild changes in the use of spot EEG. Further, we looked at whether or not the type of EEG used was associated with outcome, and we did find in those cases with immediate post neurosurgical status epilepticus, use of spot EEG but not continuous EEG was associated with increased risk of death or discharge to hospice. Importantly, there are several limitations to our study. First and foremost is identification of status epilepticus events is really dependent solely on the accuracy of the ICD-9 codes in their capturing this administrative database. Unfortunately, we're unable to differentiate important clinical entities such as convulsive or non-convulsive status epilepticus in this study. And lastly, there's a challenge in that in a retrospective study, we're unable to better dissect out what may be pathology-specific versus procedural-specific risk profiles. Lastly, to summarize the study findings, we found that craniotomy for trauma, hematoma, and elevated ICP were associated with the highest risk of immediate status epilepticus, whereas procedures for CSF diversion and craniotomy for infection were associated with highest risk of delayed status. Among those patients who underwent tumor resection, meningioma patients had the highest risk of immediate post neurosurgical status epilepticus. Lastly, we observed increasing use of continuous EEG over the study interval in patients who developed post neurosurgical status, and in those patients with immediate post neurosurgical status, use of spot EEG but not continuous EEG was associated with death or discharge to hospice and non-home discharge. Lastly, I'd like to acknowledge Michael Jin, an excellent medical student at Stanford University who helped with this work in putting together the data, and various mentors, including Dr. Skirball at Stanford. Thank you very much for your time.

perioperative status epilepticus

intracranial neurosurgery

complications

morbidity

mortality