Factors Associated With Prolonged Length of Stay in Patients Hospitalized With Generalized Convulsive Status Epilepticus in the United States

Abstract

Objective:

Generalized convulsive status epilepticus (GCSE) is a severe complication of epilepsy, which typically requires extended hospitalization, resulting in substantial resource utilization, hospital expenditures, and patient costs. In this nationwide analysis, we examined hospital length of stay (LOS) patterns for GCSE, and the factors that influence prolonged LOS.

Methods:

We extracted data for adult patients (age 18 years and above) with a primary discharge diagnosis of GCSE from the National Inpatient Sample (NIS) from 2006-2014, the largest all-payer inpatient care database in the United States. We computed LOS (≤1, 2-6, and ≥7 days), overall, and across pre-specified patient-related, hospital-related, and healthcare system-related variables available in the NIS. We identified factors independently associated with prolonged hospitalization (2 or more days), using a multivariable logistic regression model.

Results:

Of 57,832 discharged with a primary diagnosis of GCSE, 6,133 (10.7%) had a LOS ≤1 day, 27,327 (7.3%) stayed for 2-6 days, and 24,372 (42.1%) stayed for ≥7 days. After adjusting for confounders, patients who were older, female, Black, and Hispanic, who underwent continuous EEG video monitoring, were Medicare beneficiaries, had medical comorbidities, or were admitted to large/urban hospitals, were all significantly more likely to have prolonged LOS.

Conclusion:

Over 40% of patients hospitalized for GCSE in the United States spend at least a week in the hospital. Efforts to shorten hospitalization for GCSE may need to primarily focus on patient groups with select sociodemographic and clinical characteristics.

Introduction

Generalized convulsive status epilepticus (GCSE) is arguably the most dreaded of all the complications of epilepsy. Mortality from GCSE varies between 3 and 39%.^1-7 As such, GCSE often requires hospital admission and results in major health resource utilization and hospital costs.⁸ Characterizing hospitalizations after GCSE is an essential step in curbing fatality rates and healthcare expenditures related to GCSE. A simple surrogate metric of the quality of initial and subsequent in-hospital care is the length of hospitalization. Median duration of hospital stay for patients with GCSE is roughly 3 days in the United States, with prolonged stays correlating with higher costs.³ However, factors that contribute to prolonged hospital length of stay (LOS) among patients with GCSE are largely unknown. In this analysis, we explored the distribution and factors associated with prolonged hospitalization, as defined by more 2 or more days, in patients with a primary discharge diagnosis of GCSE using the largest sample of hospital admissions in the United States, covering 9 years.

Methods

We used the National Inpatient Sample (NIS) from 2006-2014 to identify adult patients (ages 18 and above) with a discharged primary discharge diagnosis of GCSE. The NIS is one of the databases developed for the Healthcare Cost and Utilization Project (HCUP). The database is designed to produce national and regional estimates of inpatient utilization, access, charges, quality, and outcomes. The NIS “approximates a 20-percent stratified sample of all discharges from U.S. community hospitals, excluding rehabilitation and long-term acute care hospitals” and contains more than 7 million unweighted (35 million weighted) hospital discharges.⁹ Generalized convulsive status epilepticus was identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 345.3.¹⁰

For this analysis, the primary independent variable of interest was the LOS in days, divided into 3 categories: 0-1 day, 2-6 days, and 7 or more days. A previous study using the same database has used the same categories.¹¹ Length of stay was considered prolonged if ≥ 2 days. We chose this cut-off to account for the observation that reversible cellular damages that cause complications typically occur within the first few minutes to hours of generalized convulsive status epilepticus.¹² We identified the following covariates from the database: age (18-44 years, 45-64 years, 65-84 years, and ≥85 years), gender (male, female), race/Ethnicity (Non-Hispanic White, Black, Hispanic, and other), death status at the end of hospitalization period (alive vs. death), primary insurance payer (Medicare, Medicaid, private, and other), hospital bed size—defined as the number of beds in the hospital (small, medium, and large), urban-teaching status of the hospital (rural, urban non-teaching, and urban teaching), the day of admission (weekday or weekend), median household income for ZIP code (divided into quartiles), hospital census region (Northeast, Midwest, South, and West), and year group (2006-2008, 2009-2011, and 2012-2014).

To provide more specific information for some of these classifications, the National Inpatient Sample defines an urban vs. rural hospital based on the Core Based Statistical Area (CBSA) designation of the county that the hospital is located in.⁹ Hospitals in counties designated as metropolitan (having at least one population center with >50,000 people) were defined as urban, while hospitals in counties designed as micropolitan (having at least one population center 10,000-49,999 people) or non-core (counties that do not meet the criteria for metropolitan or micropolitan) were defined as rural.¹³ The National Inpatient Sample based hospital bed size classifications were based on the census region of the hospital as well as the urban-teaching status.⁹ In the Northeast, hospitals were considered small, medium, and large, respectively, if they had sizes of 1-49, 50-99, and 100+ in rural hospitals; 1-124, 125-199, and 200+ in urban non-teaching hospitals; and 1-249, 250-424, and 425+ for urban teaching hospitals. In the Midwest, hospitals were considered small, medium, and large, respectively, if they had sizes of 1-29, 30-49, and 50+ in rural hospitals; 1-74, 75-174, and 175+ in urban non-teaching hospitals; and 1-249, 250-374, and 375+ for urban teaching hospitals. In the South, hospitals were considered small, medium, and large, respectively, if they had sizes of 1-39, 40-74, and 75+ in rural hospitals; 1-99, 100-199, and 200+ in urban non-teaching hospitals; and 1-249, 250-449, and 450+ for urban teaching hospitals. In the West, hospitals were considered small, medium, and large, respectively, if they had sizes of 1-24, 25-44, and 45+ in rural hospitals; 1-99, 100-174, and 175+ in urban non-teaching hospitals; and 1-199, 200-324, and 325+ for urban teaching hospitals.

In addition, medical comorbidities were evaluated using the Charleson comorbidities index (CCI), a composite score of 17 medical comorbidities.¹⁴ Also, a variable was included based on whether patients underwent continuous video electroencephalographic (EEG) monitoring, which was identified using the Procedure code 89.19.¹⁵ Data used for this analysis are publicly available de-identified data; therefore, this study did not require any institutional board review approval.

Data Availability

The National Inpatient Sample (NIS) is a publicly available database, which can be obtained upon request from the Healthcare Cost and Utilization Project (HCUP) website (https://www.hcup-us.ahrq.gov/nisoverview.jsp)

Statistical Analysis

All analyzes were performed using Stata ver.14 software (StataCorp LP College Station, TX). We used Chi-square tests and the Analysis of Variance (ANOVA) test to compare variables by length of stay (0-1 day, 2-6 days, and 7+ days) for categorical and continuous variables. All variables but the CCI were categorical variables. Odds ratios (ORs) and respective 95% confidence intervals (CI) of independent factors associated with prolonged LOS (≥ 2 days) were computed using a logistic regression model. We used an alpha level of 0.05, and variables for which statistical tests yielded p < 0.05 were considered significant and discussed in this paper.

Results

General Characteristics

Demographic characteristics and mortality, and length of stay

Of the 57,832 patients included in this analysis, 6,133 (10.65%) had a LOS from 0-1 days, 27,327 (47.25%) had a LOS from 2-6 days, and 24,372 (42.09%) had a LOS of 7 or more days (Table 1). There was a significant difference in LOS across age brackets (p < 0.001). A greater percentage of older patients had longer LOS than younger patients, with 53.33% of patients aged 65-84 and 51.47% of patients aged 85+ spending 7+ days in the hospital, while just 28.96% of patients aged 18-44 years spent 7+ days in the hospital (p < 0.001). In the same vein, 17.92% of patients aged 18-44 spent 0-1 days in the hospital, while only 4.77% of patients aged 85+ spent 0-1 days (p < 0.001). With respect to gender, female patients had longer stays in the hospital than male patients (44.2% of females spent 7+ days, while 40.3% of males spent 7+ days; p < 0.001). We also identified a statistically significant difference across race/ethnic groups (p < 0.001). This difference was most prominent for Black patients, where 47.28% spent 7+ days, relative to White (40.74%) or Hispanic patients (39.93%). Patients who died during hospitalization were significantly more likely to have had a longer LOS (p < 0.001), as 52.13% had a LOS over 7 days compared to 8.6% for 0-1 days and 39.26 for 2-6 days Table 1.

Table 1.

Comparisons of Hospital Length of Stay for Generalized Convulsive Status Epilepticus by Sociodemographic and Clinical Characteristics.

Variables	0-1 day (%)	2-6 days (%)	7+ days (%)	p-value
N(n)	10.65	47.25	42.09
Age category				<0.001
Age 18-44	17.92	53.12	28.96
Age 45-64	8.96	47.35	43.69
Age 65-84	5.89	40.78	53.33
Age 85+	4.77	43.76	51.47
Gender				<0.001
Male	11.47	48.23	40.30
Female	9.72	46.08	44.2
Race/ethnicity				<0.001
White	11.07	48.19	40.74
Black	8.96	43.76	47.28
Hispanic	10.86	49.21	39.93
Others	10.11	45.45	44.44
Died during hospitalization	8.6	39.26	52.13	<0.001
Continuous video EEG monitoring	3.21	35.47	61.31	<0.001
Primary payer				<0.001
Medicare	7.6	44.59	47.81
Medicaid	12.22	48.9	38.88
Private	12.8	48.31	38.9
Self-pay/no charge/others	17.25	53.35	29.4
Hospital bed size				<0.001
Small	14.98	51.04	33.98
Medium	12.01	48.93	39.07
Large	9.49	45.96	44.56
Urban-teaching status				<0.001
Rural	19.55	53.76	26.69
Urban nonteaching	11.71	50.18	38.11
Urban teaching	8.69	44.30	47.01
Admission day				<0.001
Weekday	10.77	45.81	43.42
Weekend	10.37	50.90	38.73
Median household income for patient’s ZIP code				<0.001
Quartile 1	10.56	46.51	42.93
Quartile 2	11.13	48.34	40.53
Quartile 3	10.37	48.37	41.26
Quartile4	10.36	45.93	43.71
Hospital census region				<0.001
Northeast	9.18	43.72	47.10
Midwest	11.31	48.51	40.18
South	9.93	47.35	42.72
West	12.91	48.88	38.21
Charleson Co-morbidity Index (CCI), mean (SE)	0.89 (0.021)	1.32 (0.012)	1.82 (0.015)	<0.001
Year category
Year 2003/06	26.27	34.46	39.27	0.1310
Year 2007/10	25.44	34.46	40.11
Year 2011/14	24.58	34.44	40.98

N, weighted sample size; n, unweighted sample size; %, weighted percentage.

Hospital and healthcare system related characteristics and length of stay

LOS was different by insurance status (p < 0.001). For example, patients on Medicare had a longer stay (47.81% spending 7+ days) than non-Medicare patients (patients on Medicaid, private insurance, and those paying out of pocket or means outside of public or private insurance). In the same line, LOS differed by hospital bed sizes (p < 0.001). Patients hospitalized in larger hospitals spent more time in the hospital (44% of patients in large beds stayed for 7+ days compared to 39.07% for those in medium bed size hospitals and 33.98% in small beds). Similarly, LOS was different by urban-teaching status of the hospitals (p < 0.001), as patients seen at urban hospitals, particularly urban teaching hospitals, stayed longer in the hospital (7+ days: 26.69% for rural hospitals compared to 38.11% for urban nonteaching and 47.01% for urban teaching hospitals). Patients admitted on weekends were also more likely to have a prolonged hospitalization than those admitted during weekdays (the proportions of patients admitted during the weekend vs. weekday spending 7+ days in the hospital were 43.42% and 38.73%, respectively, p < 0.001). We also saw a significant difference in LOS across census regions (p < 0.001), with the largest difference in patients who had a LOS of at least 7 days noted between the Northeast (47.10%) and West (38.21%).

Medical comorbidities/continuous video EEG monitoring

Patients who had a greater number of medical comorbidities as assessed by the CCI were more likely to spend more days in the hospital than those with fewer medical comorbidities (p < 0.001). Of note, LOS did not vary by year between 2003 and 2014 (p = 0.131). Continuous video EEG monitoring was associated with longer stays (p < 0.001) as 61.31% of patients receiving it spent 7+ days, compared to 3.21%, spending 0-1 days and 35.47% spending 2-6 days.

Factors Associated With Prolonged LOS

Demographic variables

After adjusting for confounders, the following demographic variables were independently associated with an increased likelihood of prolonged length of hospitalization: being a female patient (OR for female patients vs. male patients: 1.11, 95% CI: 1.05-1.19, p < 0.001), advanced age (OR for 45-64 years vs. 18-44 years: 1.98, 95% CI: 1.85-2.12, p < 0.001; OR for 65-84 years vs. 18-44 years: 2.56, 95% CI: 2.31-2.82, p < 0.001; OR for 85 and more years vs. 18-44 years: 3.4, 95% CI: 2.77-4.17, p < 0.001) (Table 2). Using non-Hispanic Whites as a reference group, Hispanic patients (OR: 1.19, 95% CI: 1.06-1.34, p < 0.001) had an increased probability of a lengthier hospitalization. Although Black patients were more likely than their Whites counterparts to stay longer in the hospital, the difference was marginally significant (OR: 1.09 (95% CI: 1.00-1.18), p = 0.05) Table 2.

Table 2.

Logistic Regression Model: Adjusted Odds-Ratio for Prolonged Length of Stay Among Adults With Generalized Convulsive Status Epilepticus.

Variables	Odds ratio (95% CI)	p-value
Gender
Male (Ref.)
Female	1.11 (1.05-1.19)	<0.001
Age category
Age 18-44 (Ref.)
Age 45-64	1.98 (1.85-2.12)	<0.001
Age 65-84	2.56 (2.31-2.82)	<0.001
Age 85+	3.40 (2.77-4.17)	<0.001
Race/ethnicity
Non-Hispanic White (Ref.)
Non-Hispanic Black	1.09 (1.00-1.18)	0.050
Hispanic	1.19 (1.06-1.34)	<0.001
Others	1. 18 (1.04-1.35)	0.013
Primary payer
Medicare (Ref.)
Medicaid	0.92 (0.84-1.01)	0.067
Private	0.78 (0.72-0.86)	<0.001
Self-pay/no charge/others	0.64 (0.57-0.70)	<0.001
Hospital bed size
Small (Ref.)
Medium	1.28 (1.15-1.42)	<0.001
Large	1.74 (1.58-1.92)	<0.001
Urban-teaching status
Rural (Ref.)
Urban nonteaching	1.99 (1.78-2.23)	<0.001
Urban teaching	2.75 (2.46-3.07)	<0.001
Admission day
Weekday (Ref.)
Weekend	1.09 (1.02-1.17)	<0.001
Median household income for patient’s ZIP code
Quartile 1 (Ref.)
Quartile 2	1.02 (0.93-1.11)	0.704
Quartile 3	1.02 (0.93-1.21)	0.623
Quartile4	0.99 (0.89-1.09)	0.768
Hospital census region
Northeast (Ref.)
Midwest	0.91 (0.81-1.02)	0.107
South	1.09 (0.99-1.21)	0.085
West	0.80 (0.72-0.90)	<0.001
Charleson Co-morbidity index (CCI), mean	1.18 (1.15-1.22)	<0.001
Year Category
Year 2003/06 (ref)	--	--
Year 2007/10	0.94 (0.87-1.04)	0.257
Year 2011/14	0.89 (0.82-0.97)	0.01

Hospital and healthcare system related variables

A positive association with prolonged LOS was also evident for patients admitted to large hospitals compared to those in small bed size hospitals (OR: 1.74, 95% CI: 1.58-1.92, p < 0.001). Compared to those admitted in rural hospitals, patients staying at hospitals in urban settings were also more likely to have a longer stay (OR for urban non-teaching hospitals: 1.99, 95% CI: 1.78-2.23; OR for urban teaching hospitals: 2.75, 95% CI: 2.46-3.07). Patients admitted on weekends were more likely (OR: 1.09, 95% CI: 1.02-1.17) to have an increased LOS compared to those admitted on weekdays. Patients hospitalized in the South (based on census regions) had a 9% increased odds of staying the hospital (OR: 1.09, 95% CI: 0.99-1.21) for a lengthy hospitalization relative to patients hospitalized in the Northeast. Presence of medical comorbidities, which were assessed using the mean Charleson Comorbidity Index, conferred an increased likelihood of prolonged hospital stay (OR: 1.18, 95% CI: 1.15-1.22).

Conversely, the odds of prolonged LOS were lower for certain variables. Relative to patients on Medicare, patients on Medicaid had 8% lower odds of prolonged hospitalization (OR: 0.92, 95% CI: 0.84-1.01), patients with private insurance had 22% lower odds of staying in the hospital for a lengthy hospitalization (OR: 0.78, 95% CI: 0.72-0.86), and those paying outside of traditional public/private insurance methods had 36% lower odds of prolonged hospitalization (OR: 0.64, 95% CI: 0.57-0.7). Similarly, patients hospitalized in the West (based on census regions) had 20% reduced odds of staying in the hospital (OR: 0.8, 95% CI: 0.72-0.9) relative to those hospitalized in the Northeast and patients hospitalized in the Midwest had 9% reduced odds of prolonged hospitalization (OR: 0.91, 95% CI: 0.81-1.02) relative to those hospitalized in the Northeast.

Medical comorbidities/continuous video EEG monitoring

The current analysis also revealed that compared to patients hospitalized from 2003-2006, patients hospitalized from 2011-2014 and from 2007-2010 were 11% (OR: 0.89, 95% CI: 0.82-0.97) and 6% (OR: 0.94, 95% CI: 0.87-1.04) less likely to spend prolonged time in the hospital, respectively.

Discussion

Our nationwide analysis confirms that most patients with GCSE have a prolonged hospital LOS and extends findings from previous studies of this topic to uncover the independent associations of select patient-level and system-level factors with prolonged hospitalization for GCSE in the United States.

The findings of this study are in accord with previous observations that the LOS is prolonged in patients with status epilepticus. Using a national administrative database in Germany, Strzelczyk et al.¹⁶ reported a median total length of stay of 11 days. In our study, a little over 2 out of 5 patients with a discharge diagnosis of GCSE spent at least 7 days in the hospital. This prolonged length of stay denotes an important utilization of healthcare resources by patients with GCSE, which is also reflected in the observation that more than 60% of those who were on continuous video EEG monitoring spent 7 days or more in the hospital.

We examined general patient-specific and healthcare system-related independent factors associated with LOS. Older age was associated with an increased LOS, mirroring previous observations on the relationship between age and LOS in patients with GCSE¹⁷ and mirroring previous observations on the relation between age and length of hospitalization for patients with other medical conditions, such as transient ischemic attacks¹¹ and trauma.¹⁸ Altogether, the positive relationship between age and prolonged LOS as observed in this study and other acute medical conditions may be explained by the fact that elderly people are more severely affected by acute illnesses possibly due to pre-existing medical comorbidities, unfavorable physiology, and perhaps late presentation to the hospital; the latter likely resulting from social isolation and decreased sense of urgency in this vulnerable population¹⁹.

In this analysis, female patients had increased odds of longer hospitalization than their male counterparts, corroborating previous reports that women were more likely to stay in the hospital than men after a generalized convulsive status epilepticus.^17,20 The underlying pathophysiologic mechanisms explaining this difference have not been completely elucidated. It has been purported that hormonal differences between men and women could account for the differential in disease (including GCSE) severity and therefore the LOS between female and male participants. With regard to epilepsy, estrogens, for example, increase dendritic spine density, therefore allowing for greater synaptic transmission, and increasing excitatory glutamatergic activity through increasing NMDA receptor activity.²¹ Meanwhile, androgens may increase the strength of inhibitory GABA-activated currents.²¹ Factors pertaining to the healthcare system could also account for the gender disparity in hospital LOS. For instance, there may be more difficulties in coordinating discharges of female patients as more women live alone and may need to go to long term care facilities as observed after hospitalization for other medical conditions, such as acute myocardial infarction.²² Although the logistic regression analysis adjusted for comorbidities, it is still possible that some medical comorbidities not included in the Charleson comorbidity index had a more pronounced impact on women than in men in our study, such as gynecologic cancers. Nonetheless, women only had an 11% increased likelihood of staying in the hospital longer than males, and it is unclear what the clinical and economic implications of this apparent small difference are.

LOS was shorter for non-Hispanic Whites. We were not able to explain this difference. Whether non-Hispanic Whites have a smoother discharge process as a result of a more favorable social capital is a possible explanation that deserve further exploration. Admissions to large urban hospitals were associated with a prolonged hospital stay, an observation paralleling the fact that the sickest patients requiring complex and multidisciplinary treatment and more difficulties in coordinating and planning hospital discharges are seen in those medical facilities. Privately insured and self-pay patients with GCSE were 24 to 36% less likely to have a prolonged hospital stay. We suspect that a combination of at least 2 factors accounts for this observation. First, patients in this insurance category may be healthier and younger as the current study covered a large period before January 1, 2014. After January 2014, insurances could not deny coverage to people with preexisting conditions and were mandated to provide affordable premiums, adjusted only for age, tobacco use, and geographic area.²³ Second, this observation likely brings to light disparities in smoothly processing discharges, with those privately insured being more likely to have an expedited and well discharge coordination plans.

Another important observation was the prolonged length of stay among patients admitted during the weekend. The “weekend effect,” a theory that outcome is worse among patients treated during the weekend compared to those treated during weekday has been studied in other neurological emergencies, including stroke.²⁴ The current study suggests that this effect extends to generalized convulsive status epilepticus. Although we could not specifically retrieve information in the NIS to explain the prolonged length of stay during the weekends, certain factors such as shortage of staff, a less smooth transfer, and delay in early recognition of ongoing subclinical status epilepticus after the end of generalized tonic/clonic activities could account for this finding.

Length of stay was longer for patients with a discharge diagnosis of GCSE in the Northeast. This is not specific for GCSE as length of stay are longer for all diagnoses in the Northeast.⁹

Our study has limitations. The NIS provides a sample of discharges from American community hospitals but does not include data from rehabilitation and long-term acute care hospitals. However, it is unlikely that patients with GCSE would receive care in the latter healthcare facilities. We identified cases using ICD-9 code 345.3 which has not been specifically validated in the NIS. Further, granular data, which could influence LOS, such as the duration of status epilepticus, antiepileptic drug use and clinical response to treatment, whether intubation or mechanical ventilation were required, the etiologies of status epilepticus, and whether patients had a pre-existing diagnosis of epilepsy (refractory or not) were not available in the NIS. In addition, status epilepticus survivors have been shown to have impairments in their quality of life and daily life activities²⁵. Although an important aspect to analyze, the relationship between length of stay and outcome, follow-up, and quality of life following GCSE could not be assessed due to the nature of the database used. Nonetheless, our study uses a large data size and several variables to help provide a better understanding of the factors associated with length of hospitalization for GCSE.

In conclusion, nearly 9 in 10 patients with generalized convulsive status epilepticus spend ≥2 days in the hospital with patients and healthcare system specific factors associated with a lengthier hospital stay. Multi-level approach strategies may be the most efficient to reduce the LOS in patients with GCSE. Further studies will evaluate granular patient-specific factors associated with prolonged LOS.