Unplanned Readmissions of Children With Epilepsy in the United States

Abstract

Background:

The burden and characteristics of unplanned readmission after epilepsy-related discharge in children in the United States is not known.

Methods:

We undertook a retrospective cohort study of children aged one to 17 years discharged after a nonelective hospitalization for epilepsy, sampled from the Healthcare Cost and Utilization Project’s 2013 and 2014 Nationwide Readmissions Database. Descriptive statistics and logistic regression models were used to examine the characteristics of initial hospitalization and risk factors for readmission.

Results:

A total of 42,873 admissions for unique patients were identified, with 4470 (10.4%) leading to readmission within 30 days. The most common readmission diagnosis was epilepsy (24.9%). Neurodevelopmental diagnoses including cerebral palsy, intellectual disability, and developmental delay were associated with increased odds of readmission. Longer hospitalization, gastrostomy, and tracheostomy were also associated with readmission, but continuous electroencephalography use was not. Children insured by Medicare had a readmission rate of 34.4%, whereas there were no associations of readmission with other sociodemographic characteristics such as neighborhood, income, and sex.

Conclusions:

Seizures are among the most frequent reasons for hospitalization in children. Establishing a benchmark readmission rate for pediatric epilepsy of 10.4% may be useful to health systems designing quality improvement efforts. Clinical factors were more strongly associated with readmission than demographic characteristics. Interventions to reduce pediatric epilepsy readmissions may have the highest yield when targeting children with neurodevelopmental comorbidities.

Introduction

Epilepsy is a common childhood condition estimated to affect about one in every 100 children.^1–3 Furthermore, childhood epilepsy is a complex disorder associated with a broad range of negative outcomes including greater care needs, worse school outcomes, increased health care use, and increased likelihood of unmet medical and mental health needs.¹ Seizures are consistently among the most frequent reason for hospital admission in children, exceeding admissions for gastroenteritis and similar to admission rates for pneumonia.⁴ Perhaps because of the high cost of readmission to health care systems, hospital readmission rates have been designated as indicators of care quality, first in adults and now in children. The Pediatric Quality Measures Program, a project of the Agency for Healthcare Research and Quality and Centers for Medicare and Medicaid Services established by the Children’s Health Insurance Program Reauthorization Act, has adopted a Pediatric All-Condition Readmission Measure.⁵ However, 30-day readmission rates in children range from 2% to 20% based on the index condition and comorbidities—therefore, understanding the utility of readmission as a care quality metric and reducing readmission rates depends on identifying the risk factors for readmission in specific pediatric patient populations. This study aimed to quantify 30-day unplanned readmission rates in children admitted for epilepsy and to characterize demographic, clinical, and hospital characteristics associated with unplanned readmission in a nationally representative sample.

Methods

Institutional approval and data availability

This study was approved by the University of Pennsylvania Human Protection Research Organization and follows the requirements of the Health Care Utilization Project (HCUP) Data Use agreement. HCUP data are publicly available for use via the Agency for Healthcare Research and Quality: https://www.hcup-us.ahrq.gov/databases.jsp.

Data source

This study used the US HCUP 2013 and 2014 Nationwide Readmissions Database (NRD), a national representative source of information on hospitalizations and readmissions. The NRD is drawn from hospitalization data from 22 states, which contain 51.2% of the US population and represent 49.3% of all US hospitalizations. The NRD is unique in that it is designed to be used with sampling weights, to allow researchers to produce national estimates of hospitalization outcomes. In addition, the NRD contains data on all age groups and from all payers plus the uninsured. Admissions that involve hospital transfers are combined into one continuous record in the NRD so they are not measured more than once in analyses.

There are many distinct characteristics of the NRD that shape which analyses can and cannot be performed using this data source. Unlike other HCUP datasets, the NRD neither contains data on race nor does it provide hospital identifiers that allow for hierarchical analyses at the hospital level, as would be needed for volume-outcome analyses or for identifying specific children’s hospitals. Furthermore, hospitalizations and outcomes for a specific individual cannot be tracked between calendar years in the NRD, so readmission analysis observations must be truncated accordingly. The NRD also cannot track hospitalizations across state lines and therefore identifies nonresident admissions so researchers may exclude them from readmission analyses.

Study design and population

We conducted a retrospective cohort study of hospitalized children with a principal or admitting diagnosis of epilepsy, defined by the International Classification of Disease, Ninth Revision, (ICD-9) diagnosis codes for epilepsy (ICD-9 345.xx) but excluding those with a diagnosis code for convulsion (ICD-9 780.3), a method that has been validated in pediatric hospital admissions for identifying admissions for seizure in children with epilepsy.⁶ We also excluded hospitalizations of children codiagnosed with infantile spasms (345.6x), as this form of epilepsy is known to have unique needs for inpatient care.⁷ Individuals between the ages of one and 17 years were included in the analysis. The lower age limit was chosen because data on hospitalized infants may be unreliable in the NRD.

Index, or qualifying, hospitalizations were defined as those that denoted epilepsy as the admitting diagnosis, were urgent (rather than elective), lasted more than one day, and did not result in inpatient death. We focused on readmission following urgent hospitalizations because unplanned hospitalizations for epilepsy are more likely to reflect acute epilepsy-related health events or complications of epilepsy therapy, whereas elective admissions for epilepsy may consist of electroencephalographic (EEG) monitoring, surgical or diagnostic evaluation, or monitored therapy change. If an individual had more than one qualifying hospitalization in the year, one of those hospitalizations was randomly chosen to be the index encounter. Hospitalizations beginning less than 30 days before the end of the calendar year were excluded due to lack of adequate follow-up time to determine whether a study outcome had occurred. We also excluded patients who were nonresidents of the state in which they were treated, as local readmissions may not have been captured.

Demographic, clinical, and hospital characteristics

We extracted data on demographic characteristics including age, sex, socioeconomic status (categorized as quartiles of median income in the patient’s resident postal code), and expected payer (categorized as Medicare, Medicaid, private insurance, self-pay/no charge/other). Age was evaluated as both a continuous and a categorical variable. Age categories were defined based on epilepsy classification recommendations of International League Against Epilepsy and were grouped as early childhood (ages one to three years), school age (ages four to 12 years), and adolescent (ages 12 to 17 years).⁸

Literature review and clinical expertise were used a priori to determine which codiagnoses and hospital events associated with readmission. Data on the presence of the following conditions or events were collected if documented during the index hospitalization: status epilepticus, continuous EEG (cEEG), intractable epilepsy, intellectual disability, developmental delay, cerebral palsy, Lennox-Gastaut syndrome, gastrostomy tube, tracheostomy, and mechanical ventilation. These data were transformed into indicator variables to be examined as predictors of readmission. We extracted data on adverse events documented using e-codes and categorized these as “adverse events due to medical care or surgical error” and “adverse events due to drugs or therapy.”⁹ We identified the use of critical care services as the presence of an ICD-9 code for apnea (786.03), cardiac or pulmonary arrest (799.1), mechanical ventilation (96.7x), or respiratory failure (518.8x). Hospital characteristics reported in the NRD include teaching status (metropolitan teaching, metropolitan nonteaching, and nonmetropolitan), hospital bed size, and hospital ownership category (government, nonprofit, for-profit).

Outcomes

Our primary study outcome was readmission, defined as the first unplanned admission within 30 days of discharge from an index admission. Secondary outcomes examined for both the index admission and readmission included discharge disposition, length of stay, and hospital charges. The reason for readmission was determined using HCUP clinical classification software (https://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp) to categorize the principal diagnosis for the readmission.

Statistical analysis

Sample weights were applied to create national estimates for all study variables and outcomes. Demographic characteristics were compared using chi-square tests for categorical variables. Multivariable logistic regression models were used to estimate the associations between study variables and our primary and secondary study outcomes. Model variables were initially selected a priori based on clinical knowledge, review of existing literature, and variable availability in the NRD. Adjusted odds ratios reported were from models including all demographic and hospital characteristics, but not all other clinical characteristics due to colinearity. Statistical analyses were performed using SAS software, version 9.4 (SAS Institute, Inc Cary NC, USA).

Results

We identified 42,873 patients with index admissions for epilepsy meeting inclusion criteria. Of these, 4470 (10.4%) were readmitted within 30 days of discharge. Table 1 contains the patient and hospital characteristics of the study sample by readmission status.

TABLE 1.

Characteristics of 42,873 Children Emergently Admitted for Epilepsy in the United States, 2013–2014

Characteristic	N (%)
Age: mean (standard error of mean), median (IQR)	7.6 (0.07), 6.2 (2.8–11.3)
1–3 years	11,608 (27.1)
4–12 years	21,783 (50.8)
12–17 years	9482 (22.1)
Sex
Male	23,423 (54.6)
Female	19,450 (45.4)
Primary payer
Medicare	28 (0.1)
Medicaid	26,369 (61.5)
Private insurance	14,213 (33.2)
Self-pay/no charge/other	2263 (5.3)
Zip code quartile
$1-$37,999	14,153 (33)
$38,000-$47,999	11,759 (27.4)
$48,000-$63,999	9413 (22.0)
$64,000+	7547 (17.6)
Control/ownership of hospital
Government, nonfederal	5842 (13.6)
Private, not-for-profit	35,170 (82.0)
Private, for-profit	1861 (4.3)
Hospital bed size
Small	1737 (4.1)
Medium	10,981 (25.6)
Large	30,155 (70.3)
Teaching status
Metropolitan nonteaching	3355 (7.8)
Metropolitan teaching	38,996 (91.0)
Nonmetropolitan	522 (1.2)
Length of stay (mean, S.D., median, IQR)	2.9 (0.09) 1.4 (0.6–2.5)
0–1 day	17,260 (40.3)
2 days	11,833 (27.6)
3+ days	13,780 (32.1)
Index hospitalization disposition
Home/routine	40,785 (95.1)
Inpatient postacute care	879 (2.0)
Home with home health care	1187 (2.8)
Other	23 (0.1)
Codiagnosis/hospitalization events
Mechanical ventilation/endotracheal intubation	3901 (9.1)
Critical care services	5148 (12.0)
cEEG	8842 (20.6)
Status epilepticus	10,553 (24.6)
Tracheostomy	1975 (4.6)
Gastrostomy	5638 (13.2)
Intellectual disability	3138 (7.3)
Developmental delay	10,338 (24.1)
Cerebral palsy	7151 (16.7)
Intractable epilepsy	7629 (17.8)
Lennox-Gastaut	929 (2.2)
Adverse events
Any adverse event: no	41,062 (95.8)
Any adverse event: yes	1811 (4.2)
Adverse event due to medical care: no	42,511 (99.2)
Adverse event due to medical care: yes	362 (0.8)
Adverse event due to drugs/treatment: no	41,385 (96.5)
Adverse event due to drugs/treatment: yes	1488 (3.5)

Abbreviations:

cEEG = Continuous electroencephalography

IQR = Interquartile range

Index admissions for epilepsy occurred chiefly in large (70.3%), metropolitan-teaching (91.0%) private nonprofit (82.0%) hospitals. Patients were slightly more often male (54.6%) and most often were aged between four and 12 years (50.8%). The most frequent payer was the Medicaid program (61.5%). A diagnosis of intellectual disability or developmental delay was present in about one-third of index admissions, but fewer than one-fifth (17.8%) were diagnosed as having intractable epilepsy. Status epilepticus was diagnosed in 24.6% of patients during the index admission, and 20.6% underwent cEEG monitoring and 12.0% received critical care services. Adverse events were recorded rarely, occurring in 4.2% children hospitalized for epilepsy; medication errors were the most common category of adverse event (3.5%). Disposition after the index admission was home for nearly all patients (95.1%).

Readmission

In this nationally represented sample, 4470, or 10.4%, children were readmitted within 30 days of hospitalization for epilepsy. Table 2 displays the rate of 30-day readmissions for each of the primary study variables. Crude readmission rates were similar across age groups, from 9.3% in those aged 12 to 17 years to 12.1% in those aged one to three years. Males and females were readmitted equally. Children insured by the Medicare program, a group very small in number, had a distinctly high readmission rate (34.4%), whereas private, Medicaid, and uninsured children had near-average readmissions.

TABLE 2.

Index Hospitalization and Patient Characteristics Associated With 4470 Thirty-Day Readmissions of Children With Epilepsy*

Characteristic	Percent Readmitted Within 30 Days
Age
1–3 years	1409 (12.1)
4–12 years	2178 (10.0)
12–17 years	883 (9.3)
Sex
Male	2355 (10.1)
Female	2115 (10.9)
Primary Payer
Medicare	9 (34.4)
Medicaid	2762 (10.5)
Private insurance	1447 (10.2)
Self-pay/no charge/other	252 (11.1)
Zip code quartile
$1-$37,999	1551 (11.0)
$38,000-$47,999	1224 (10.4)
$48,000-$63,999	949 (10.1)
$64,000+	746 (9.9)
Control/ownership of hospital
Government, nonfederal	610 (10.4)
Private, not-for-profit	3707 (10.5)
Private, for-profit	153 (8.2)
Hospital teaching status
Metropolitan nonteaching	328 (9.8)
Metropolitan teaching	4100 (10.5)
Nonmetropolitan	42 (8.0)
Hospitalization length of stay
0–1 day	1328 (7.7)
2 days	1146 (9.7)
3+ days	1996 (14.5)
Hospitalization disposition
Home/routine	4119 (10.1)
Inpatient postacute care	130 (14.8)
Home with home health care	218 (18.4)
Other	2 (10.4)
Hospitalization adverse events
Any adverse event: no	4238 (10.3)
Any adverse event: yes	232 (12.8)
Adverse event due to medical care: no	4420 (10.4)
Adverse event due to medical care: yes	50 (13.9)
Adverse event due to drugs/treatment: no	4288 (10.4)
Adverse event due to drugs/treatment: yes	182 (12.2)
Hospitalization codiagnosis/hospitalization event
Mechanical ventilation/endotracheal intubation	425 (10.9)
Critical care services	650 (12.6)
cEEG	881 (10.0)
Status epilepticus	1106 (10.5)
Tracheostomy	395 (20.0)
Gastrostomy	1029 (18.2)
Intellectual disability	455 (14.5)
Developmental delay	1238 (12.0)
Cerebral palsy	917 (12.8)
Intractable epilepsy	1139 (14.9)
Lennox-Gastaut	88 (9.4)

Abbreviation:

cEEG = Continuous electroencephalography

Longer index hospitalizations had more frequent readmissions, from 7.7% among hospitalizations lasting one or less days to 14.5% among hospitalizations lasting three or more days. Adverse events had 2% to 4% higher readmissions.

Gastrostomy or tracheostomy was associated with a higher-than-average readmission rate (18% to 20%), but status epilepticus and cEEG were not. Individuals with common chronic neurological conditions had mildly increased readmission rates (Table 2). Among nonclinical characteristics, discharge with home health care had the highest readmission rate of 18.4%.

Table 3 displays the results of regression models that examined the associations of patient, clinical, and hospital characteristics with 30-day readmission. In fully adjusted models, Medicare insurance was associated with a nearly 5-fold increase in the odds of readmission when compared with private insurance (adjusted odds ratio [AOR] 4.79, 95% confidence interval [CI] 1.05 to 21.93). Hospital type, patient zip code quartile, patient sex, and index disposition were not associated with readmission.

TABLE 3.

Predictors of 30-Day Readmission Among Children Emergently Admitted for Epilepsy in the United States, 2013–2014

Characteristic	Unadjusted odds of readmission: OR (95% CI)	Adjusted odds of readmission: AOR (95% CI)
Age
1–3 years	REF	REF
4–12 years	0.80 (0.69–0.94)	0.84 (0.72–0.97)
12–17 years	0.74 (0.63–0.88)	0.75 (0.63–0.88)
Sex
Male	REF	REF
Female	1.09 (0.96–1.24)	1.1 (0.96–1.25)
Primary Payer
Medicare	4.64 (0.94–22.82)	4.79 (1.05–21.93)
Medicaid	1.03 (0.89–1.20)	0.96 (0.81–1.13)
Private insurance	REF	REF
Self-pay/no charge/other	1.10 (0.78–1.57)	1.05 (0.74–1.48)
Zip code quartile
$1-$37,999	1.12 (0.92–1.37)	1.13 (0.91–1.39)
$38,000-$47,999	1.06 (0.86–1.31)	1.06 (0.86–1.31)
$48,000-$63,999	1.02 (0.82–1.27)	1.01 (0.82–1.26)
$64,000+	REF	REF
Teaching status
Metropolitan nonteaching	0.92 (0.73–1.16)	1.00 (0.8–1.26)
Metropolitan teaching	REF	REF
Nonmetropolitan	0.74 (0.45–1.21)	0.76 (0.46–1.24)
Index length of stay
0–1 day	REF	REF
2 days	1.29 (1.09–1.53)	1.26 (1.07–1.49)
3+ days	2.03 (1.76–2.35)	1.95 (1.68–2.26)
Index disposition
Home/routine	REF	REF
Inpatient postacute care	1.55 (1.01–2.37)	1.43 (0.93–2.20)
Home with home health care	2.00 (1.51–2.66)	1.67 (1.25–2.23)
Other	1.04 (0.12–9.06)	0.99 (0.10–9.62)
Chronic conditions/index admission characteristics (yes versus no)*
Mechanical ventilation/endotracheal intubation	1.05 (0.83–1.34)	0.81 (0.64–1.04)
Critical care services	1.28 (1.05–1.57)	1.00 (0.82–1.22)
cEEG	0.94 (0.78–1.13)	0.84 (0.70–1.00)
Status epilepticus	1.01 (0.86–1.18)	0.90 (0.77–1.06)
Tracheostomy	2.26 (1.75–2.91)	2.04 (1.58–2.65)
Gastrostomy	2.19 (1.9–2.53)	1.90 (1.62–2.23)
Intellectual disability	1.51 (1.19–1.91)	1.46 (1.13–1.89)
Developmental delay	1.23 (1.07–1.43)	1.17 (1.01–1.35)
Cerebral palsy	1.33 (1.16–1.54)	1.29 (1.11–1.50)
Intractable epilepsy	1.68 (1.42–1.99)	1.53 (1.28–1.82)
Lennox-Gastaut	0.89 (0.6–1.32)	0.96 (0.65–1.42)
Adverse events (yes versus no)
Any	1.28 (1.0–1.63)	1.05 (0.82–1.35)
Medical care	1.39 (0.88–2.2)	1.07 (0.67–1.71)
Drugs	1.2 (0.91–1.6)	1.01 (0.76–1.34)

Abbreviations:

AOR = Adjusted odds ratio

cEEG = Continuous electroencephalography

CI = Confidence interval

OR = Odds ratio

REF = Reference

Bold values are statistically significant (P < 0.05).

^*Adjusted models include sex, age, insurance, zip code income, length of stay, hospital characteristics, and disposition.

Comorbid conditions including intellectual disability, developmental delay, cerebral palsy, tracheostomy, and gastrostomy tube were all associated with increased odds of readmission (Table 3). A diagnosis of intractable epilepsy was also associated with higher readmission (AOR 1.53 95% CI 1.28 to 1.82), although diagnoses of Lennox-Gastaut syndrome (AOR 0.96, 95% CI 0.65 to 1.42) and status epilepticus (AOR 0.9, 95% CI 0.77 to 1.06) were not. We found no statistically significant effect of documented adverse events, cEEG monitoring, mechanical ventilation, or receipt of critical care services during the index admission on odds of readmission in our fully adjusted models.

Readmission diagnoses and outcomes

Using chronic condition classification software, the most common principal diagnosis for readmission was epilepsy (24.9%); pulmonary diagnoses (acute pneumonia, respiratory failure, upper respiratory tract infection, bronchitis, influenza, and aspiration) accounted for 20.2% of readmissions. Medical care-related complications of a device, procedure, or implant were the reasons for 9.9% readmissions.

Median length of stay during readmission was 2.2 days (interquartile range 1.0 to 4.5). Adverse events during readmission were similar to those observed during the index admission (5.2% vs 4.2%). cEEG and mechanical ventilation occurred in 15.9% and 9.6% of repeat hospitalizations. After readmission, discharge to home was slightly lower than on initial admission (88.4%), and 28 of the 4470 patients (0.6%) died in the hospital. Median cost of readmission was $19,733 (interquartile range $9,900 to $42,196).

Discussion

Although a measure for all-cause pediatric readmission⁵ has been established by the Pediatric Quality Measures Program, the threshold over which readmission rates would be considered unacceptably high has not been determined for all causes or specific diagnoses. In a nationally represented sample we found that 10.4% children admitted for epilepsy were readmitted within 30 days of discharge. Having a national benchmark for pediatric epilepsy readmission rates may prove useful for hospitals or health care systems working to improve quality of care for these patients, such as bettering clinical guidelines for inpatient treatment, targeting care of very young patients, and Medicare patients. Furthermore, prevention of readmission represents improved patient care and results in cost savings, especially because some of the readmissions are likely to be prolonged or intensive care unit admissions.¹⁰

Our findings of a 10.4% 30-day readmission rate represents a considerably higher rate than is characteristic of other common chronic diseases with intermittent exacerbations such as asthma (2.5%) as well as of other common reasons for pediatric hospitalization such as pneumonia (7.7%) and gastroenteritis (5.1%).^11–13 Our 10.4% readmission rate is also higher than prior estimates of seizure-related readmission (9%)¹¹ and reflects the need for detailed understanding of the specific disease in designing studies to help contribute evidence for policy decisions. Epilepsy differs from many other common causes of pediatric admission in as much as it has frequent planned (epilepsy monitoring unit) admissions, which would likely alter the readmission rate and were purposefully excluded from our study.

Income, race, and sex are consistent predictors of readmission in adults, including those with epilepsy.^14,15 These sociodemographic characteristics were not associated with readmission in this pediatric study sample. Medicare coverage use was associated with the greatest likelihood of readmission in our adjusted models. Medicare coverage is available to children who are designated as permanently disabled by the social security administration,¹⁶ and the disability required to enroll a child in Medicare likely independently increases hospitalization risk. Taken together, these findings suggest that clinical need, rather than social determinants of health, drive readmission in pediatric epilepsy.

The finding that status epilepticus, Lennox-Gastaut syndrome, and critical care service use were not associated with readmission could be unexpected and bears some consideration. It is possible that clinical guidelines for the outpatient and inpatient management of Lennox-Gastaut syndrome or status epilepticus are effective in minimizing readmissions.¹⁷ Moreover, having a history of Lennox-Gastaut or receipt of critical care services may enhance outpatient follow-up or ensure emergency specialist care during and after hospitalization.

We found that children with other neurological or medical comorbidities, including intellectual disability, developmental delays, and cerebral palsy were at higher risk of readmission and accounted for a substantial proportion of readmissions. Thus, interventions to reduce pediatric epilepsy readmissions may have the highest yield when targeted toward very young children and those with comorbidities, especially neurological comorbidities. Our study confirmed that readmission diagnosis is most commonly the same as the admission diagnosis in epilepsy, as it is across almost all pediatric readmission diagnoses.⁴ Our examination of medication and medical care adverse event codes suggests that an additional proportion (9.9%) of readmissions may also be preventable through enhanced patient safety procedures.

Limitations of this study include the lack of information on hospital or regional variation in readmission rates, which is useful in understanding the likelihood that rates are disease related (in which case rates would look relatively uniform) or have other social and health care delivery-related factors (in which case substantial variability would be expected). This study is also unable to examine in greater detail the clinical factors leading to admission and readmission, which we would assume to be of critical importance, including availability and effectiveness of outpatient follow-up. An additional limitation is the lack of information about children younger than one year, who are known to have more severe forms of epilepsy. Despite these limitations, we present a nationally represented readmission rate after acute hospitalization for pediatric epilepsy as well as data that suggest that continued investment in outpatient care and patient safety may help prevent readmissions in pediatric epilepsy, especially for very young patients, patients with comorbidities, and Medicare patients. Having this benchmark for readmission on a national scale is the first step in bettering outcomes for pediatric patients with epilepsy.

Supplementary Material

Supplementary data

Supplementary data related to this article can be found at https://doi.org/10.1016/j.pediatrneurol.2020.01.010.