Introduction
Status epilepticus (SE), characterized by prolonged seizures, Reference Trinka, Cock and Hesdorffer1 is a neurological emergency with high rates of morbidity and mortality. Reference Betjemann, Josephson, Lowenstein and Burke2 SE can occur in people with or without a diagnosis of epilepsy or prior seizures and can have an unknown or known etiology, such as intoxication or encephalitis. Reference Lee3 The treatment of SE is often complex, requiring lengthy hospitalizations. One study conducted in Alberta found that over half of patients admitted to a hospital with SE required treatment in an intensive care unit (ICU), with a mean ICU length of stay of 6.5 days, Reference Mahmoud, Marette, Lindqvist and Ahmed4 indicating that SE is a significant burden on the Canadian healthcare system.
Although there are many estimates of the frequency of SE, there are few estimates of the incidence of first-episode SE. Those that do exist vary considerably, ranging from 10.7 to 73.7 per 100,000 person-years. Reference Rodin, Krogstad and Aukland5,Reference Kantanen, Sairanen and Kälviäinen6 Additionally, no estimates are available from Canadian study populations. Understanding the incidence of first-episode SE is essential to ensure that medical professionals are adequately educated about the likelihood of first-episode SE. In addition, a baseline incidence estimate should be available so that changes may be monitored over time. Therefore, incidence estimates from Canadian study populations are needed to clarify the incidence of first-episode SE in Canada.
Epilepsy is the most common risk factor for SE, Reference Fountain7 with an estimated 16%–61% of patients with SE having a previous epilepsy diagnosis. Reference Ng and Pavlova8,Reference Bergin, Brockington and Jayabal9 Among those with epilepsy, patients who are non-adherent to their anti-seizure medications are at highest risk. Reference DeLorenzo, Pellock, Towne and Boggs10 Other known SE risk factors include stroke, traumatic brain injury, alcohol abuse, brain tumors and infections. Reference Fountain7 However, less is known about non-health-related risk factors for SE, such as income, educational attainment, ethnicity and other sociodemographic characteristics. Social and economic factors are known to influence health, and identifying these factors is necessary to reduce health inequities. 11 Most studies that have examined sociodemographic characteristics as potential SE risk factors were among patients with epilepsy and included only age and sex, as clinical risk factors (e.g. the etiology and type of epilepsy) were of primary interest. Therefore, it is necessary to further explore sociodemographic characteristics as potential risk factors for SE, as any inequities identified would indicate the sub-populations that would benefit most from prevention strategies.
Therefore, the objectives of this study were to estimate the incidence of first-episode SE in Ontario, Canada, and the associations between potential sociodemographic and health-related risk factors and first-episode SE.
Methods
The objectives of this study were addressed using a retrospective cohort design and linked census and health administrative data from Ontario, Canada. The following datasets were linked on an individual patient level: Canadian Institute for Health Information (CIHI) Discharge Abstract Database (DAD), CIHI National Ambulatory Care Reporting System (NACRS), Canada’s 2006 Census long-form questionnaire (CENSUS), 2006 Canadian Census Health and Environment Cohort (CanCHEC), Canadian Vital Statistics – Death Database (CVSD), Canadian Mortality Database (CMDB), Canadian Cancer Registry (CCR), Postal Code Conversion File Plus (PCCF+), Historical Postal Code file and the Canadian Marginalization Index (CAN-Marg).
The DAD and NACRS contain the health administrative data that were used. The DAD database contains administrative and clinical data from acute inpatient facilities in Ontario. 12 The NACRS database contains emergency department, day surgery and clinic data from all Ontario facilities. 13 Full descriptions of the datasets used in this study can be found in the supplementary material (eTable 1).
Population
The study population included respondents to the 2006 Census long-form questionnaire who consented to the use of their taxation data to determine their place of residence. Approximately 20% of the Canadian population receives the long-form questionnaire; the remainder receives a shorter version. The long-form questionnaire requests additional demographic, social and economic information about the individuals in the household. 14 We included only long-form respondents because we required information that was only collected from these individuals. We excluded nonresidents of Ontario, those with missing age or sex information, were younger than 18 or older than 105, or had a prior hospital admission or emergency department visit for SE (Fig. 1).
Figure 1: Flow diagram depicting the cohort build.
Potential Risk Factors
Most potential sociodemographic risk factors were obtained from the Census using variables that directly correspond to these concepts. These include sex, age, educational attainment, marital status, employment status, race, immigration status, second-generation Canadian, speaks one or more of Canada’s official languages, mother tongue, weekly number of hours spent on unpaid housework, childcare, senior care, household conditions (in need of repairs), individual-level household income and rurality.
Neighborhood household income and neighbourhood marginalization (dimensions: instability, deprivation, dependency and ethnic concentration) were assigned to participants using the PCCF + and CAN-Marg databases, respectively, using their postal code of residence at baseline, the census administration date of May 16, 2006. Participants’ baseline postal code of residence was available from the Historical Postal Code file.
We also investigated comorbid conditions that increase the risk of seizures, epilepsy, or SE as potential risk factors. Alcohol and drug abuse, chronic kidney disease (CKD) and diabetes increase the risk of provoked seizures. Reference Misra and Kalita15,Reference Verrotti, Scaparrotta, Olivieri and Chiarelli16 These conditions were identified using International Classification of Diseases 10th Revision (ICD-10) codes in the DAD and NACRS databases within the two years before baseline. Brain tumors, central nervous system (CNS) infections, dementia, depression or anxiety, epilepsy or seizures, stroke and traumatic brain injuries (TBIs) are known to increase the risk of epilepsy or SE. Reference Fountain7,Reference Caramelli and Castro17,Reference Tellez-Zenteno, Patten, Jetté, Williams and Wiebe18 These comorbid conditions were identified using ICD-9 and ICD-10 codes in all available DAD and NACRS data before baseline (since 2000 for the DAD and 2002 for the NACRS). As cancer also increases the risk of epilepsy, Reference Gonzalez Castro and Milligan19 diagnoses within the five years before baseline were identified using the CCR. The specific ICD codes used to define these comorbid conditions are available in the supplementary material (eTable 2).
Outcome
Participants were followed from baseline, May 16, 2006, until they were admitted to a hospital or visited an emergency room for SE, their death, or the end of the follow-up period, May 15, 2009. First-episode SE was defined using ICD-10 code G41: status epilepticus identified in the primary diagnostic position in the DAD and NACRS databases. The CVSD and CMDB were used to identify deaths.
Statistical Analysis
We estimated the cumulative incidence and incidence rate of first-episode SE in our study population. We also age-standardized the estimated incidence rate using the full 2006 Canadian population.
In the bivariate analysis, we assessed whether there were differences in each potential risk factor between those who did and did not develop SE using standardized differences, with differences greater than 0.1 considered statistically significant. In the multivariable analysis, we included selected risk factors in the Cox proportional hazards regression model to estimate the hazard ratios (HRs) for their associations with first-episode SE. These risk factors were selected based on the standardized differences from the bivariate analysis and how common the risk factors were among patients who developed SE. For example, we excluded the variables for time spent on child and senior care, as few people who developed SE spent more than five hours per week on these tasks. A p-value of 0.05 was used to determine statistical significance.
Missing values for categorical variables were categorized into a separate “missing” category for all analyses. However, missing data from the Census was minimal since Statistics Canada follows up with respondents of incomplete questionnaires. Data analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC) software and standardized differences were estimated using a macro from Yang and Dalton. Reference Yang and Dalton20 Data were accessed through the Research Data Centre at Western University. Statistics Canada requires at least five individuals per table cell, except for quantiles, which require at least twenty. To meet this requirement, we combined categories as needed. We also randomly rounded the cells in Table 1 to multiples of five, which Statistics Canada also requires.
Table 1: Risk factors of study sample by outcome
*Column of “Developed SE” does not always sum to 140 people, as cells were rounded to the nearest multiple of five. Standardized differences were calculated using unrounded values.
Results
We included 1,301,700 individuals in our study population. In the three years following the Census, 140 (0.011%) developed SE (3.6 per 100,000 person-years). The age-standardized incidence rate was 3.5 per 100,000 person-years. The total and mean follow-up time was 3,857,888 person-years and 3.0 (± 0.26) person-years, respectively. There were 31,140 (2.4%) deaths during follow-up.
Most differences in the distribution of potential risk factors between individuals who did and did not develop SE were statistically significant in the bivariate analysis (Table 1). In the multivariable analysis (Table 2), the sociodemographic factors that significantly increased the risk of first-episode SE included older age, male sex and being unemployed. Having a university degree, being an immigrant, spending more time on unpaid housework, and having a higher household income were protective. However, except for older age, all these hazard ratios were close to one and are likely of minimal clinical relevance. The comorbid conditions that statistically significantly increased the risk of first-episode SE include alcohol or drug abuse, brain tumor or cancer, chronic kidney disease, dementia, depression or anxiety, diabetes, epilepsy or seizures and stroke.
Table 2: Multivariable analysis of select risk factors and status epilepticus
Discussion
The estimated cumulative incidence of first-episode SE over the three-year study period was 0.011%, and the age-adjusted incidence rate was 3.5 per 100,000 person-years. Risk factors included older age, alcohol or drug abuse, brain tumors or cancer, chronic kidney disease, dementia, diabetes, epilepsy or seizures and stroke.
Our incidence rate estimate is lower than those previously reported in other high-income countries. Reference Rodin, Krogstad and Aukland5,Reference Knake, Rosenow and Vescovi21 Our lower estimate is likely due to the official International League Against Epilepsy (ILAE) definition of SE in use during the study period, the 1981 definition, which defined SE as a seizure with a minimum duration of 30 minutes. 22 Since the seizure duration defining SE was reduced to 5 minutes in 2015 by the ILAE, studies using data from after this date, are more likely to report a higher incidence rate. Reference Trinka, Cock and Hesdorffer1
Inconsistencies between incidence rates estimated across studies could also be due to the ICD codes used to identify SE. Some studies, including ours, used ICD-10 code G41: status epilepticus, while others have used G40.301 and G40.311: generalized idiopathic epilepsy and epilepsy syndromes, intractable or not intractable, with status epilepticus. It is currently unclear which ICD codes should be used to identify SE in health administrative data, as no studies estimating their accuracy could be identified. We could not use G40.301 and G40.311 as they are not used in the DAD or NACRS databases. Therefore, using only G41 is appropriate and likely highly accurate for our data. However, as with all studies using health administrative data to estimate first-episode SE incidence, it is possible that SE could be coded as a seizure (G40) or that a seizure could be incorrectly classified as SE.
We identified several sociodemographic factors statistically significantly associated with first-episode SE, but most of their hazard ratios were close to one. However, age was positively associated with the first episode of SE, with a clinically relevant hazard ratio of 1.35, a finding that is consistent with previous studies. Reference DeLorenzo, Pellock, Towne and Boggs10,Reference Knake, Rosenow and Vescovi21,Reference Tatro, Hamilton, Peters and Rowe23 We did not find meaningful associations with sex in the present study. Studies that have included sex as a risk factor have reported inconsistent findings, with most studies reporting a higher incidence in men, Reference Bergin, Brockington and Jayabal9,Reference Knake, Rosenow and Vescovi21,Reference Bhalla, Tchalla and Mignard24,Reference Tiamkao, Pranboon, Thepsuthammarat and Sawanyawisuth25 another reporting a higher incidence in women, Reference Vignatelli, Tonon and D’Alessandro26 and another reporting similar incidence rates in men and women. Reference Leitinger, Trinka and Giovannini27 Thus, it remains unclear whether sex is associated with SE.
In addition, we found no clinically relevant associations between first-episode SE and educational attainment, employment status, immigration status, individual-level household income, rurality, or race. We found a small but statistically significant reduced risk of first-episode SE among those who performed more than five hours of housework per week (5–14 hours: HR = 0.98, 95% CI = 0.98, 0.99; 15–29 hours: 0.98, 95% CI = 0.97, 0.98; 30+ hours: HR = 0.97, 95% CI = 0.97, 0.98). It is unclear why housework might be associated with a reduced risk of first-episode SE. However, the small risk difference may not be clinically meaningful.
Although few sociodemographic factors were associated with first-episode SE, we found multiple comorbid conditions associated with our outcome in the multivariable analysis. Comorbid conditions associated with SE include hypoxemia, systemic infection, dementia, TBI, cancer, drug poisoning, CNS infection, metabolic causes, alcohol-related causes and stroke. Reference Legriel and Brophy28 In addition to estimating significant associations with some of these conditions, we also estimated significant associations with CKD and diabetes. These findings are expected, as it is well-known that both CKD and diabetes can cause provoked seizures. Reference Misra and Kalita15,Reference Verrotti, Scaparrotta, Olivieri and Chiarelli16 Importantly, some comorbid conditions, such as alcohol abuse and diabetes, are more prevalent in the general population. Therefore, despite these conditions having smaller hazard ratios than others, such as CKD, they likely have a more significant effect on SE risk at the population level.
There are some limitations of this study. As we explored potential sociodemographic risk factors for first-episode SE, we required the linkage of census and health administrative data, which typically takes many years to become available. Therefore, although our estimate was measured using data from 2006, it was the most current data linked to the Census when we began the study. Additionally, gender is an essential sociodemographic variable we would have liked to include in this study. However, gender was not added to the Census until 2021 29 and was therefore unavailable.
We also identified relatively few people who had first-episode SE over follow-up (n = 140), which limited the power of the multivariable analysis. The limited number of patients with SE also required us to combine risk factor categories to meet the minimum cell size requirements of Statistics Canada when reporting the bivariate analysis results. Additionally, we did not account for the competing risk of death, which may have resulted in an overestimation of the hazard ratios.
In addition, we likely did not identify all patients with the comorbidities included. As these variables were measured between 2000 and 2006, we would have identified patients with the condition only if they had either an emergency department visit or hospitalization for the condition within this period. Therefore, patients with preexisting but well-controlled conditions would be more likely to have been missed. Additionally, we would not have identified patients with the conditions diagnosed between the index date and the end of the follow-up period. Another study limitation is the potential that some adjusted potential risk factors mediate the associations between other potential risk factors and SE, which would result in an underestimation of the associations between the exposure and SE. For example, alcohol abuse may partially mediate the association between socioeconomic status and SE, and adjusting for both may bias the association between socioeconomic status and SE toward the null hypothesis. Therefore, mediation effects between the potential risk factors included in this study should be explored in future research.
As high-quality healthcare databases were used for this population-based study, we believe our results are generalizable to adult populations in many other regions, particularly other parts of Canada and other high-income nations. The generalizability of our study results to lower- and middle-income countries is unclear, as they may have higher incidence rates due to a higher prevalence of risk factors. In addition, some variables in our study may poorly represent the sociodemographic characteristics of people in other countries, such as mother tongue and race, making the estimates for some risk factors in this study less relevant for those populations.
The incidence of SE has a bimodal distribution across the lifespan, with peaks in early childhood and older adulthood. Therefore, since we did not include children, our results are not generalizable to pediatric populations. Additionally, since we used census data to identify the study population, generalizability to individuals who do not reside in private households, such as those in collective dwellings or who are unhoused, may be limited. 30
Conclusion
The incidence of first-episode SE in our sample of Ontario adult residents between 2006 and 2009 was 3.5 per 100,000 person-years, somewhat lower than rates estimated in other high-income countries. Older age and several comorbid conditions, including alcohol or drug abuse, brain tumors or cancer, chronic kidney disease, dementia, diabetes, epilepsy or seizures and stroke, were associated with an increased risk of first-episode SE. This information can be used to educate medical professionals about the likelihood of first-episode SE and its risk factors and act as a baseline estimate to monitor changes over time. Future research should continue to explore sociodemographic and health-related risk factors for SE to confirm our findings.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/cjn.2024.11.
Acknowledgements
The analyses presented in this paper were conducted at the Western University RDC, part of the Canadian Research Data Centre Network.
Funding
This study was funded by the Schulich School of Medicine & Dentistry Collaborative Research Seed Grant and the Jack Cowin Endowed Chair in Epilepsy Research at Western University.
Competing interests
GC, TA, PW, and DD report no conflicts of interest. JB holds the Jack Cowin Endowed Chair in Epilepsy Research at Western University. TG holds research grants from the Canadian Institutes of Health Research, New Frontiers in Research Fund, and Academic Medical Organisation of Southwestern Ontario for unrelated research.
Statement of authorship
GC: literature search, project planning, study design, data analysis, data interpretation, manuscript writing; TA: project planning, study design, data interpretation, manuscript editing; PW: study design, manuscript editing; TG: manuscript editing; DD: manuscript editing; JB: project oversight, study design, data interpretation, manuscript editing.
