Advertisement

A nationally representative study on discharge against medical advice among those living with epilepsy

  • Parul Agarwal
    Correspondence
    Corresponding author at: Population Health Science & Policy, Department of Neurology, Institute for Health Care Delivery Science, Icahn School of Medicine at Mount Sinai, 1425 Madison Ave. @ East 98th Street, NY, NY 10029, L2-37, Second Floor, USA.
    Affiliations
    Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Institute for Healthcare Delivery Science, Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
  • Huaqing Xi
    Affiliations
    Institute for Healthcare Delivery Science, Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
  • Nathalie Jette
    Affiliations
    Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Institute for Healthcare Delivery Science, Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
  • Jung-yi Lin
    Affiliations
    Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Institute for Healthcare Delivery Science, Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
  • Churl-Su Kwon
    Affiliations
    Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
  • Mandip S. Dhamoon
    Affiliations
    Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
  • Madhu Mazumdar
    Affiliations
    Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA

    Institute for Healthcare Delivery Science, Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA
    Search for articles by this author
Open ArchivePublished:November 30, 2020DOI:https://doi.org/10.1016/j.seizure.2020.11.018

      Highlights

      • Around 2 % admissions in people living with epilepsy resulted in DAMA and these are more frequent in socially-disadvantaged populations.
      • DAMA-related admissions in people with epilepsy significantly increased from 2003 to 2014.
      • Top reasons for DAMA admissions were epilepsy/convulsions, alcohol- and substance-related disorders, and complicated diabetes mellitus.

      Abstract

      Purpose

      Discharges against medical advice (DAMA) are associated with adverse patient outcomes among those with epilepsy. Our goal was to examine trends and factors associated with DAMA among those living with epilepsy.

      Methods

      A retrospective cross-sectional study was performed using the 2003–2014 National Inpatient Sample database. ICD-9-CM diagnosis codes were used to identify admissions of patients with epilepsy. Following outcomes were examined among epilepsy patients: proportion and predictors of DAMA, 12-year DAMA trends and causes of admissions.

      Results

      In 2014, of the 187,850 admissions in patients with epilepsy, 3783 (2.01 %) were DAMA. Male sex, Black race, younger age, lower household income, Medicaid/self-pay/other as primary payer, lower Elixhauser comorbidities index, weekend admission, non-elective admission, hospital in northeast region, and urban nonteaching hospital were all associated with DAMA. There was a significant increase in the proportion of DAMA in people with epilepsy from 2003 to 2014 (1.13 %–2.01 %, p < 0.0001). The top reasons of admissions for epilepsy patients who were DAMA were: epilepsy/convulsion (21.02 %), alcohol- (8.86 %) and substance-related disorders (3.75 %), and diabetes mellitus with complications (3.33 %).

      Conclusions

      Our findings provide opportunities to understand DAMA among those living with epilepsy, which is more prevalent in socially-disadvantaged populations. This study highlight the need to develop electronic medical records-based prediction tools that could be used at the point-of-care to enable the early identification of people at risk for DAMA, since it is often likely preventable. Future mixed methods studies are recommended to identify facilitators of DAMA and strategies for prevention.

      Keywords

      1. Introduction

      Poor management of epilepsy may result in repeated seizures and increased hospital admissions [
      • Tian N.
      • et al.
      Active epilepsy and seizure control in adults—United States, 2013 and 2015.
      ,
      • Holmquist L.
      • Russo C.A.
      • Elixhauser A.
      Hospitalizations for epilepsy and convulsions, 2005, in Healthcare Cost and Utilization Project (HCUP) Statistical Briefs [Internet].
      ]. Around 3 % epilepsy patients leave hospital against medical advice which is strongly related to higher 30-day readmissions [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ]. Proper discharge planning is vital to reduce medication errors, repeated seizures, and other adverse health outcomes. Thus, it is prudent to examine the trends and factors associated with discharges against medical advice (DAMA) among those living with epilepsy to inform future interventions to reduce the occurrence of DAMA. Furthermore, epilepsy is an ambulatory care sensitive condition and constant timely communication with health care providers and continuity of care is essential to improve and maintain general health status and reduce hospital admissions. The Institute of Medicine (IOM) has identified gaps in access to care among those living with epilepsy and is committed to limiting adverse consequences due to epilepsy, enhancing preventative care programs and provision of best quality care based on current research evidence [
      • England M.J.
      • et al.
      Epilepsy across the spectrum: promoting health and understanding.: a summary of the Institute of Medicine report.
      ,
      • Long C.
      • Fureman B.
      • Dingledine R.
      2014 epilepsy benchmarks: progress and opportunities.
      ]. Healthy People 2020 has defined improvement in access to care for epilepsy patients as one of the priority goals [].
      DAMA occurs when a patient leaves the hospital against medical advice and without adequate discharge planning. Although the proportion of those who get DAMA is relatively small (1 %–3 %) [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ,
      • Alfandre D.J.
      "I'm going home": discharges against medical advice.
      ,
      • Ibrahim S.A.
      • Kwoh C.K.
      • Krishnan E.
      Factors associated with patients who leave acute-care hospitals against medical advice.
      ], these patients, when compared to those discharged with medical advice, are at potentially increased risk of adverse outcomes including readmissions within 30-days, increased emergency department visits and healthcare costs [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ,
      • Alfandre D.J.
      "I'm going home": discharges against medical advice.
      ,
      • Kahle C.H.
      • Rubio M.L.
      • Santos R.A.
      Discharges against medical advice: considerations for the hospitalist and the patient.
      ,
      • Southern W.N.
      • Nahvi S.
      • Arnsten J.H.
      Increased risk of mortality and readmission among patients discharged against medical advice.
      ,
      • Glasgow J.M.
      • Vaughn-Sarrazin M.
      • Kaboli P.J.
      Leaving against medical advice (AMA): risk of 30-day mortality and hospital readmission.
      ,
      • Aliyu Z.
      Discharge against medical advice: sociodemographic, clinical and financial perspectives.
      ]. Additionally, DAMA causes financial challenges for health care systems due to escalating health care costs and services utilization as a result of fragmented care [
      • Glasgow J.M.
      • Vaughn-Sarrazin M.
      • Kaboli P.J.
      Leaving against medical advice (AMA): risk of 30-day mortality and hospital readmission.
      ,
      • Aliyu Z.
      Discharge against medical advice: sociodemographic, clinical and financial perspectives.
      ]. Therefore, examining factors associated with DAMA is essential.
      Studies examining predictors of DAMA in those with epilepsy are scarce. A study of neurological patients in Germany found that seizures was one of the three most common presenting symptoms in emergency room patients who were DAMA, and also that younger age was associated with DAMA [
      • Hoyer C.
      • et al.
      Uncompleted emergency department care and discharge against medical advice in patients with neurological complaints: a chart review.
      ]. A recently published US study using the Nationwide Readmission Database noted several factors associated with DAMA in epilepsy patients including younger age, male sex, non-private insurance, lower socioeconomic status, smoking, alcohol history, and drug use [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ]. Our study is different as it uses a nationally representative database over multiple years to highlight the 12-year trends in DAMA among people with epilepsy and to examine causes of admissions in DAMA patients over time. We also characterize factors, e.g. race, associated with DAMA which were not investigated in earlier studies.

      2. Method

      2.1 Study design and data source

      A retrospective cross-sectional study was conducted using the National Inpatient Sample (NIS) database for the years 2003–2014. The NIS is part of the Healthcare Cost and Utilization Project (HCUP), maintained by the Agency for Healthcare Research and Quality (AHRQ). It is one of the largest, publically available, nationally representative, all-payer, inpatient stay database in the US. The NIS is a 20 % stratified sample from the State Inpatient Database, which includes all inpatient discharges across 44 States and the District of Columbia, representing more than 96 % of the U.S. population [

      2014 Introduction to the NIS. Healthcare Cost and Utilization Project (HCUP). December 2016 [cited 2019 September, 25]; Available from: https://www.hcup-us.ahrq.gov/db/nation/nis/NIS_Introduction_2014.jsp.

      ]. The 2014 NIS includes more than 7 million inpatient records from 4411 US community hospitals, excluding rehabilitation and long-term acute care hospitals [

      2014 Introduction to the NIS. Healthcare Cost and Utilization Project (HCUP). December 2016 [cited 2019 September, 25]; Available from: https://www.hcup-us.ahrq.gov/db/nation/nis/NIS_Introduction_2014.jsp.

      ]. The NIS is a discharge-level database, which implies that we cannot identify repeat hospitalizations among patients from this database as a new identification number is assigned to each hospital discharge.
      Participating hospitals in the NIS database are stratified based on bed size, teaching status, hospital control/ownership, location and census divisions. Large sample size, stratified design and availability of discharge weights facilitates national estimates and analysis of special populations such as the uninsured. The NIS contains variables including patient demographic information, hospital characteristics, expected payment source, discharge status, length of stay (LOS), severity and comorbidity measures and 30 diagnostic and 15 procedural variables, defined using the International Classification of Diseases, Ninth revision, Clinical Modification (ICD-9-CM) codes [

      INTRODUCTION TO THE HCUP NATIONAL INPATIENT SAMPLE (NIS) 2014. 2016 [cited 2019 September, 25]; Available from: https://www.hcup-us.ahrq.gov/db/nation/nis/NISIntroduction2014.pdf.

      ]. All data are de-identified.

      2.2 Study cohort

      The 2014 NIS was used to identify potential DAMA predictors, while the 2003–2014 NIS were used to assess the 12-year trend of DAMA and causes of hospital admissions in patients with and without DAMA. Hospitalizations with epilepsy in any diagnostic position were identified for all ages using previously validated ICD-9 CM codes 345.xx (excluding status epilepticus: 345.2x and 345.3x, if coded without an epilepsy code) [
      • Jette N.
      • et al.
      How accurate is ICD coding for epilepsy?.
      ,
      • Mbizvo G.K.
      • et al.
      The accuracy of using administrative healthcare data to identify epilepsy cases: a systematic review of validation studies.
      ]. The following admissions were excluded: those who died during hospitalization and cases with a missing or invalid value for age, sex, race, median household income, expected primary payer, LOS, total charges, disposition status, admission day is/not on a weekend and elective/non-elective admission status. Fig. 1 shows the process of cohort selection in the 2014 NIS. The study cohort for the years 2003–2013 was similarly selected.
      Fig. 1
      Fig. 1Diagram for the cohort selection in 2014.

      2.3 Measures

      2.3.1 Dependent variable

      The primary outcome was discharge disposition dichotomized as DAMA vs. non-DAMA. DAMA was defined as discharge of patients who left against medical advice or discontinued care. All patients with other discharge dispositions (routine discharge by hospital, transfer to short-term hospital, other transfers to skilled nursing facility, intermediate care facility, another type of facility, home health care, or discharge alive but destination unknown) were classified as non-DAMA admissions. Proportions and factors associated with DAMA were examined. The secondary outcomes were 12-year trend of DAMA proportions and causes of admissions defined using clinical classifications software (CCS) categories for diagnoses.

      2.3.2 Independent variables

      Sociodemographic variables included sex (female, male), age, race/ethnicity (White, Black, Hispanic, other), median household income for patient's ZIP code (Q1: $1 - $39,999, Q2: $40,000 - $50,999, Q3: $51,000 - $65,999, Q4: $66,000+) and expected primary payer (Medicare, Medicaid, private insurance, self-pay, other). We also extracted admission day (admitted Monday-Friday, admitted Saturday-Sunday), admission status (non-elective admission, elective admission), length of stay (LOS), total charges (USD), and AHRQ Elixhauser comorbidity index developed to predict hospital readmissions [
      • Moore B.J.
      • et al.
      Identifying increased risk of readmission and in-hospital mortality using hospital administrative data: the AHRQ elixhauser comorbidity index.
      ]. AHRQ Elixhauser comorbidity index is a weighted comorbidity index that was estimated using the Elixhauser comorbidity software and ICD-9-CM codes [
      • Moore B.J.
      • et al.
      Identifying increased risk of readmission and in-hospital mortality using hospital administrative data: the AHRQ elixhauser comorbidity index.
      ]. Comorbidities that were included are presented in appendix Table A2. Hospital-level variables included bed size (small, medium, large), hospital location/teaching status (rural, urban nonteaching, urban teaching), and region of hospital (northeast, mid-west, south, west).

      2.4 Statistical analysis

      To identify risk factors associated with DAMA, we first looked into the demographic and clinical characteristics of the cohort in 2014, stratified by DAMA status. Categorical data were reported as frequency and percentage, and continuous variables were reported as mean with standard error (SE) or median with range. Cell sizes less than or equal to 10 are expressed as not reportable (NR) according to HCUP data agreement [
      Publishing with HCUP data. Healthcare cost and utilization project (HCUP).
      ].
      Univariable logistic models were fitted to assess covariate difference (sex, age, race/ethnicity, median household income, expected primary, admission day, admission, Elixhauser comorbidity, hospital bed size, hospital location/teaching status, and region of hospital) between the DAMA and non-DAMA group. Multivariable logistic regression analysis was then performed to determine predictors of DAMA risk and included covariates from the univariate logistic models that were significant (p < 0.05).
      For the hospitalization trend analysis, we aggregated the data from 2003 to 2014 and calculated the yearly weighted frequency of DAMA among those with epilepsy. The Cochran–Armitage trend test was conducted to perform a temporal trend analysis of the proportion of DAMA for all races, primary payer type and type of admissions. The multiple comparisons adjustment method of Bonferroni was used for the trend analysis. Major causes of admission for the epilepsy cohort in the DAMA and non-DAMA group were described with frequencies and percentages. The statistical significance level was set as 0.05 throughout. All analyses accounted for the complex sample weighted design of the survey data in NIS and were conducted using Statistical Analysis Software version 9.4 (SAS Institute Inc., Cary, NC).

      2.5 Data availability statement

      HCUP data are available publically. The data user agreement limits release of data, and any requests should be made directly with HCUP. All of our analyses comply with HCUP regulations.

      2.6 Standard protocol approvals, registrations and patient consent

      The Mount Sinai Institutional Review Board has reviewed and approved this project and waived need for human consent. The study was approved by the Mount Sinai Program for the Protection of Human Subjects under IRB-IF2356842 and all analyses comply with the HCUP data use agreement.

      3. Results

      3.1 Baseline characteristics

      Out of 7,071,762 admissions in 2014, we identified 187,850 eligible admissions among epilepsy, 3783 of whom (2.01 %) got DAMA. Table 1 presents the patient and hospital characteristics. Among all hospitalizations, epilepsy patients who were DAMA as compared to non-DAMA were younger (mean age 44.42 vs. 52.38 years), more likely to be males (58.21 % vs. 47.21 %), more likely to be Black (26.04 % vs. 20.20 %), in the lowest quartile of median income (45.65 % vs. 34.84 %), and more likely to be Medicaid insured (43.91 % vs. 25.20 %) or self-pay (11.08 % vs. 3.25 %). They were also more likely to be admitted on the weekend (26.86 % vs. 22.34 %), had a lower proportion of elective admissions (5.84 % vs. 15.25 %), a lower Elixhauser comorbidity index score (mean 5.03 vs 8.36), more likely to be at an urban nonteaching hospital (29.58 vs 25.42), less likely to be at an urban teaching hospital (62.68 vs. 66.24), and were more likely to be admitted in the Northeast (26.22 vs. 21.17) and less likely in the Midwest (17.08 vs. 20.57) of the US. Patients who were DAMA had lower LOS (median 1 vs. 3 days) and overall hospital charges (median $15,668 vs. $27,655).
      Table 1Characteristics of epilepsy inpatients in NIS in 2014 stratified by DAMA and non-DAMA groups.
      DAMA
      DAMA: Discharge against medical advice.
      Non-DAMATotalOdds ratio/Percent change in LOS and total charge95 % Confidence intervalp-value
      NWt-N
      Wt N: Weighted N; weighted values are supplied by NIS to calculate estimates at national level and are calculated as the number of nationally hospital discharges divided by the number of hospital discharges in the sampling frame.
      %NWt-N%NWt-N%
      Total378318,9152.01184,067920,33597.99187,850939,250100.00
      Age, mean (SE
      SE: Standard error.
      )
      44.42(0.23)52.38(0.35)52.22 (0.34)0.99(0.98,0.99)<.0001
      Sex<.0001
      In addition to p-values for each category, type III p-values were calculated and reported for categorical variables to indicate whether the covariate was significantly related to DAMA or not.
      Male220211,01058.2186,889434,44547.2189,091445,45547.43Ref
      Ref: referent level of the categorical variables in the univariable logistic regression models.
      Female1581790541.7997,178485,89052.7998,759493,79552.570.64(0.60,0.69)<.0001
      Race<.0001
      White223511,17559.08120,204601,02065.30122,439612,19565.18Ref
      Black9854,92526.0436,967184,83520.0837,952189,76020.201.43(1.31,1.57)<.0001
      Hispanic36618309.6717,94089,7009.7518,30691,5309.751.10(0.95,1.26)0.1978
      Other1979855.21895644,7804.87915345,7654.871.18(0.96,1.46)0.1142
      Median household income for patient's ZIP Code<.0001
      Q1:1–39,9991727863545.6564,121320,60534.8465,848329,24035.052.12(1.86,2.42)<.0001
      Q2: 40,000–50,999986493026.0649,947249,73527.1450,933254,66527.111.56(1.36,1.78)<.0001
      Q3: 51,000–65,999668334017.6638,337191,68520.8339,005195,02520.761.37(1.20,1.57)<.0001
      Q4: 66,000+402201010.6331,662158,31017.2032,064160,32017.07Ref
      Expected primary payer<.0001
      Medicare1138569030.0893,645468,22550.8894,783473,91550.46Ref
      Medicaid1661830543.9146,390231,95025.2048,051240,25525.582.95(2.70,3.22)<.0001
      Private insurance420210011.1032,642163,21017.7333,062165,31017.601.06(0.93,1.20)0.3800
      Self-pay419209511.08598329,9153.25640232,0103.415.76(5.09,6.53)<.0001
      Other1457253.83540727,0352.94555227,7602.962.21(1.81,2.70)<.0001
      Admission day is on a weekend
      Admitted Monday-Friday276713,83573.14142,942714,71077.66145,709728,54577.57Ref
      Admitted Saturday-Sunday1016508026.8641,125205,62522.3442,141210,70522.431.28(1.18,1.38)<.0001
      Elective versus non-elective admission
      Non-elective admission356217,81094.16156,005780,02584.75159,567797,83584.94Ref
      Elective admission22111055.8428,062140,31015.2528,283141,41515.060.35(0.29,0.42)<.0001
      Length of stay(LOS), median (range)1 (0, 296)3 (0, 358)3 (0, 358)−51.29 %(−52.49 %,−50.06 %)<.0001
      Total charges, median (range)15,668 (674, 920,339)27,655 (101, 4.374.478)27,351 (101, 4,337,478)−44.36 %(−45.97 %,−42.71 %)<.0001
      Elixhauser Comorbidity Index for Mortality score, mean (SE)5.03(0.16)8.36(0.05)8.29 (0.05)0.96(0.96,0.97)<.0001
      Hospital bed size0.1492
      Small697348518.4230,729153,64616.6931,426157,13116.73Ref
      Medium1048524027.7052,282261,41028.4053,330266,65028.390.88(0.77,1.02)0.0843
      Large203810,19053.87101,056505,28054.90103,094515,47054.880.89(0.78,1.01)0.0719
      Hospital location/teaching status0.0002
      Rural29314657.7515,35076,7508.3415,64378,2158.33Ref
      Urban nonteaching1119559529.5846,793233,96525.4247,912239,56025.511.25(1.08,1.46)0.0032
      Urban teaching237111,85562.68121,924609,62166.24124,295621,47666.171.02(0.88,1.18)0.7982
      Region of hospital
      Northeast992496026.2238,961194,80521.1739,953199,76521.27Ref
      Midwest646323017.0837,857189,28520.5738,503192,51520.500.67(0.57,0.79)<.0001
      South1481740539.1575,105375,52540.8076,586382,93040.770.77(0.67,0.89)0.0004
      West664332017.5532,144160,72017.4632,808164,04017.470.81(0.69,0.96)0.0140
      f NR: Not reportable; cell sizes ≤10 are expressed as NR according to the HCUP data agreement.
      a DAMA: Discharge against medical advice.
      b Wt N: Weighted N; weighted values are supplied by NIS to calculate estimates at national level and are calculated as the number of nationally hospital discharges divided by the number of hospital discharges in the sampling frame.
      c SE: Standard error.
      d In addition to p-values for each category, type III p-values were calculated and reported for categorical variables to indicate whether the covariate was significantly related to DAMA or not.
      e Ref: referent level of the categorical variables in the univariable logistic regression models.

      3.2 Potential predictors of DAMA: multivariable analysis results

      The statistically significant predictors in the univariable analysis remained significant in the multivariable model (Table 2). Blacks had higher odds of DAMA than Whites (AOR = 1.12, 95 % CI = 1.02, 1.22, p = 0.02). As compared to those with highest median household income of 66,000 and above, admission in those with epilepsy of lower median household income i.e. $39,999 and below were associated with higher odds of DAMA (adjusted OR (AOR) = 1.90; 95 % CI = 1.65, 2.18; p < 0.0001). Compared to admissions in epilepsy patients with Medicare, Medicaid (AOR = 2.14; 95 % CI = 1.93, 2.37; p < 0.0001) and self-pay (AOR = 3.99; 95 % CI = 3.47, 4.59; p < 0.0001) admissions were associated with higher odds of DAMA. Those with admissions on weekend had higher odds of DAMA (AOR = 1.17; 95 % CI = 1.08, 1.26; p < 0.0001) as compared to those admitted on weekdays. The following factors were associated with lower DAMA odds: age (AOR = 0.991; 95 % CI = 0.989, 0.993; p < 0.0001), AHRQ Elixhauser comorbidity index (AOR = 0.97; 95 % CI = 0.96, 0.97; p < 0.0001), female sex (AOR = 0.67; 95 % CI = 0.62, 0.71; p < 0.0001), Hispanics (AOR = 0.74; 95 % CI = 0.64, 0.86; p < 0.0001), elective admissions (AOR = 0.34; 95 % CI = 0.28, 0.42; p < 0.0001), and hospitals located in the Midwest (AOR = 0.57; 95 % CI = 0.49, 0.69; p < 0.0001) compared to the Northeast US.
      Table 2Predictors of DAMA in NIS in 2014 for epilepsy inpatients from multivariable logistic regression.
      Odds ratio95 % Confidence intervalp-value
      Age in years at admission0.991(0.989, 0.993)<.0001
      Elixhauser Comorbidity Index for Mortality score0.97(0.96, 0.97)<.0001
      Sex<.0001
      In addition to p-values for each category, type III p-values were calculated and reported for categorical variables to indicate whether the covariate was significantly related to DAMA or not.
      MaleRef
      Ref: referent level of the categorical variables in the multivariable logistic regression models.
      Female0.67(0.62, 0.71)<.0001
      Race<.0001
       WhiteRef
       Black1.12(1.02, 1.22)0.0150
       Hispanic0.74(0.64, 0.86)<.0001
       Other0.96(0.79, 1.16)0.6547
      Median household income for patient's ZIP Code<.0001
       Q1: 1–39,9991.90(1.65, 2.18)<.0001
       Q2: 40,000–50,9991.50(1.30, 1.72)<.0001
       Q3: 51,000–65,9991.33(1.17, 1.52)<.0001
       Q4: 66,000+Ref
      Expected primary payer<.0001
       MedicareRef
       Medicaid2.14(1.93, 2.37)<.0001
       Private insurance0.97(0.85, 1.11)0.6414
       Self-pay3.99(3.47, 4.59)<.0001
       Other1.71(1.40, 2.09)<.0001
      Admission day is on a weekend<.0001
       Admitted Monday-FridayRef
       Admitted Saturday-Sunday1.17(1.08, 1.26)<.0001
      Elective versus non-elective admission<.0001
       Non-electiveRef
       Elective admission0.34(0.28, 0.42)<.0001
      Region of hospital<.0001
       NortheastRef
       Midwest0.57(0.49, 0.66)<.0001
       South0.62(0.54, 0.71)<.0001
       West0.70(0.60, 0.82)<.0001
      Hospital location/teaching status<.0001
      RuralRef
      Urban nonteaching1.26(1.08, 1.47)0.0032
      Urban teaching0.87(0.75, 1.01)0.0661
      a Ref: referent level of the categorical variables in the multivariable logistic regression models.
      b In addition to p-values for each category, type III p-values were calculated and reported for categorical variables to indicate whether the covariate was significantly related to DAMA or not.

      3.3 Temporal trend analysis

      There was a substantial increase in DAMA proportions among epilepsy patients from 2003 to 2014 (1.13 %–2.01 %, p < 0.0001; Fig. 2A). An increased proportions of Whites (0.97 %–1.83 %, p < 0.0001), Blacks (1.40 %–2.60 %, p = 0.01), and other (1.50 %–2.15 %, p < 0.0001) races were DAMA between 2003 and 2014 (Fig. 2B). However, a similar trend was not observed for Hispanic patients with epilepsy (1.60 %–2.00 %, p = 0.67). Those with Medicaid (1.86 %–3.46 %, p < 0.0001), and private insurance (0.49 %–1.27 %, p < 0.0001) had significantly increasing trend of DAMA from 2003 to 2014. Proportion of DAMA among self-pay patients did not significantly increase over time (3.19 %–6.54 %, p = 1.00), however proportion of DAMA in this group was generally much higher as compared to other groups (Fig. 2C). Medicare beneficiaries (1.00 %–1.20 %, p = 0.05) had a stable trend for DAMA. Percent of DAMA among non-elective admissions of epilepsy patients increased between 2003 and 2014 (1.03 %–1.90 %, p < 0.0001), while that of electively admitted patients had a relatively stable trend from 2003 to 2014 (0.10 % to 0.12 %, p < 0.0001) (Fig. 2D).
      Fig. 2
      Fig. 2Temporal trend of proportion of DAMA in epilepsy patients from 2003 to 2014. (A) Trend in all epilepsy patients. (B) Trend stratified by race. (C) Trend stratified by expected primary payer. (D) Trend stratified by elective admission.

      3.4 Causes of admission

      Fig. 3 shows the top 5 causes of admissions in epilepsy patients in 2014. The top causes of admissions in other years are listed in Appendix (Table A1). In 2014, the most common cause of admission was epilepsy with convulsions in both DAMA (21.02 %) and non-DAMA (16.8 %) groups. Among DAMA patients, alcohol-related (8.86 %), substance-related (3.75 %) disorders and diabetes mellitus with complications (3.33 %) were the top reasons of admission. Septicemia (except in labor) was a common cause of hospitalization in both DAMA and non-DAMA groups.
      Fig. 3
      Fig. 3Incidence of top five causes of admissions in 2014.

      4. Discussion

      This study describes proportion and factors associated with DAMA among those living with epilepsy, 12-year temporal trends of DAMA and causes of admissions. We found that around 2 % of the hospital admissions resulted in DAMA in those with epilepsy, and that DAMA was more common in socially-disadvantaged groups. Epilepsy patients with DAMA had lower comorbidity index, but had a higher proportion of those with comorbid conditions including alcohol and drug abuse disorders as well as mood disorders. There was a substantial increase in the proportion of admissions resulting in DAMA from 2003 to 2014. The most common causes of admissions were epilepsy/convulsions, alcohol-related and substance-related disorders.
      DAMA patients were younger, were more likely to be male, Black, admitted on a weekend, and had lower socioeconomic status, a non-elective admission, and comorbidities including drug and alcohol abuse, psychoses, paralysis and liver disease. These findings are consistent with other studies in other patient populations as described in prior systematic reviews [
      • Alfandre D.J.
      "I'm going home": discharges against medical advice.
      ,
      • Alfandre D.
      Reconsidering against medical advice discharges: embracing patient-centeredness to promote high quality care and a renewed research agenda.
      ] as well as in a recent study in people with epilepsy [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ]. Other studies have identified novel findings in the general population related to factors associated with DAMA that we could not examine here due to limitations of the dataset e.g. lack of access to a primary care physician, presence of financial limitations, being the caregiver of a sick family member, and the preconceived belief that the hospital admission is only short term [
      • Alfandre D.J.
      "I'm going home": discharges against medical advice.
      ,
      • Alfandre D.
      Reconsidering against medical advice discharges: embracing patient-centeredness to promote high quality care and a renewed research agenda.
      ,
      • Jeremiah J.
      • O’Sullivan P.
      • Stein M.D.
      Who leaves against medical advice?.
      ,
      • Green P.
      • Watts D.
      • Dhopesh V.
      Why patients sign out against medical advice (AMA): factors motivating patients to sign out AMA.
      ] ; nevertheless, these are yet to be studied in those living with epilepsy. Early identification of subgroups of epilepsy patients at-risk for DAMA will facilitate timely intervention to prevent DAMA and the subsequent reduction in adverse outcomes. Improved understanding of the factors associated with DAMA beyond those available in large secondary population-based datasets such as the NIS may also help inform efforts to improve continuity of care. For example, further more detailed exploration of DAMA risk factors may find that outreach by social workers, public health agencies and other community-based partners may reduce DAMA among those living with epilepsy [
      • Alfandre D.
      Improving quality in against medical advice discharges—more empirical evidence, enhanced professional education, and directed systems changes.
      ,
      • Alfandre D.
      Advancing the science of discharges against medical advice: taking a deeper dive.
      ].
      Our finding of 2 % DAMA among those living with epilepsy is similar to earlier studies in the general population or in epilepsy that reported DAMA proportions between 1–3 % [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ,
      • Alfandre D.J.
      "I'm going home": discharges against medical advice.
      ,
      • Ibrahim S.A.
      • Kwoh C.K.
      • Krishnan E.
      Factors associated with patients who leave acute-care hospitals against medical advice.
      ]. However, it is noteworthy that a direct comparison of these proportions is implausible as we found only one epilepsy specific study [
      • Raja A.
      • Trivedi P.D.
      • Dhamoon M.S.
      Discharge against medical advice among neurological patients: characteristics and outcomes.
      ]. After further analysis of the NIS database, we found that there were fewer DAMA dispositions in people without epilepsy (1.06 %) compared to people with epilepsy in 2014. This remained consistent throughout the study period (2003–2014: epilepsy = 1.13 %–2.01 %, no epilepsy = 0.90 %–1.06 %). One notable finding of our study is that the proportion of admissions that were DAMA increased considerably among those with Medicaid insurance (1.86 % in 2003 to 3.46 % in 2014; p < 0.0001). It implies that socioeconomic status is a strong predictor of DAMA among those living with epilepsy. Future research, including mixed methods studies, with a qualitative component, is warranted to better understand the reasons for DAMA in low-income subgroups of people with epilepsy and the challenges faced by healthcare providers when patients decide to leave against medical advice. Indeed, a qualitative study of pediatric residents found that DAMA can make decision-making challenging for physicians as well as sometimes caregivers request earlier discharges due to missed wages or the perceived notion that the patient is well [
      • Macrohon B.C.
      Pediatrician’s perspectives on discharge against medical advice (DAMA) among pediatric patients: a qualitative study.
      ]. A similar study exploring attitudes and perception regarding DAMA among neurologists would provide a deeper understanding about these issues.
      Alcohol-related disorders were among the top 5 causes of admission in people with epilepsy who left against medical advice. This was seen consistently from 2003 to 2014 while substance-related disorders, nonspecific chest pain, and diabetes mellitus with complications appeared as one of the top causes of admission only since 2006. Indeed, other studies have estimated that alcohol and substance related disorders are most prevalent in DAMA admissions [
      • Spooner K.K.
      • et al.
      Discharge against medical advice in the United States, 2002–2011.
      ,
      • Kenne D.R.
      • Boros A.P.
      • Fischbein R.L.
      Characteristics of opiate users leaving detoxification treatment against medical advice.
      ]. Affective disorders/mood disorders were also common reasons of admission for both DAMA and non-DAMA groups in our study. Those living with epilepsy are already suffering from a complex debilitating condition. Comorbid mood disorders, being relatively common in people with epilepsy, can overwhelm patients and further worsen quality of life [
      • Tian N.
      • et al.
      Active epilepsy and seizure control in adults—United States, 2013 and 2015.
      ,
      • Kanner A.M.
      Depression and epilepsy: a new perspective on two closely related disorders.
      ,
      • Siarava E.
      • et al.
      Depression and quality of life in patients with epilepsy in Northwest Greece.
      ]. Implementation of appropriate targeted interventions and counseling strategies are imperative to provide optimal discharge planning and reduce DAMA among epilepsy patients with comorbid mood disorders, alcohol-related and substance- related disorders.
      To the best of our knowledge, this is the first comprehensive long-term (12-year) study examining DAMA and causes of admissions in people with epilepsy. The sample size was very large and the application of discharge weights enables us to generalize the study findings to the US population. However, there are also study limitations. First, the NIS is an encounter-level rather than a patient-level database, thereby limiting identification of repeated visits and total number of epilepsy patients who left against medical advice. Due to unavailability of patient notes and qualitative data in our study, we could not identify the reasons for DAMA among those living with epilepsy. Although ICD-9-CM codes for epilepsy are not validated in the NIS database, we used a case definition that has been previously validated in other US databases to capture cohorts of people with epilepsy. We excluded those who had only status epilepticus without concurrent epilepsy diagnosis to reduce false positives and exclude provoked cases [
      • Jette N.
      • et al.
      How accurate is ICD coding for epilepsy?.
      ]. Coding accuracy for epilepsy subtypes (four-digit codes) are not optimal and were not specifically addressed in this study.
      DAMA is a significant issue in those living with epilepsy. The characteristics and causes of admissions of DAMA subgroup differ significantly from non-DAMA subgroup and in particular demonstrates how socially-disadvantaged groups may be more prominently affected by this issue. There is a need to better understand the reasons behind DAMA that will enable healthcare providers and public health practitioners to provide appropriate support to at-risk patients. Our study may inform the design of future interventions to mitigate any barriers in continuity of care. It also highlights the need to develop electronic medical records-based prediction tools that could be used at the point-of-care to enable the early identification of people with epilepsy at risk for DAMA, since it is often likely preventable. Few studies have developed models that can be implemented in the EMR system to predict discharge disposition in advance [
      • Ballester N.
      • et al.
      An early warning tool for predicting at admission the discharge disposition of a hospitalized patient.
      ,
      • Salimi M.
      • Rozovskaya V.
      Predicting discharge disposition using patient complaint notes in electronic medical records.
      ,
      • Peck J.S.
      • et al.
      Predicting emergency department inpatient admissions to improve same‐day patient flow.
      ]. However more work is required to develop tools that are specifically designed for those living with epilepsy followed by their real-time application.

      Ethical publication statement

      We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

      Declaration of Competing Interest

      Dr. Jette receives grant funding paid to her institution for grants unrelated to this work from NINDS ( NIH U24NS107201 , NIH IU54NS100064 ) and PCORI . She receives an honorarium for her work as an Associate Editor of Epilepsia. Other authors report no disclosures.

      Acknowledgements

      Dr. Jette is the Bludhorn Professor of International Medicine at the Icahn School of Medicine at Mount Sinai.

      Appendix A. Supplementary data

      The following is Supplementary data to this article:

      References

        • Tian N.
        • et al.
        Active epilepsy and seizure control in adults—United States, 2013 and 2015.
        MMWR Morb Mortal Wkly Rep. 2018; 67: 437
        • Holmquist L.
        • Russo C.A.
        • Elixhauser A.
        Hospitalizations for epilepsy and convulsions, 2005, in Healthcare Cost and Utilization Project (HCUP) Statistical Briefs [Internet].
        Agency for Healthcare Research and Quality (US), 2008
        • Raja A.
        • Trivedi P.D.
        • Dhamoon M.S.
        Discharge against medical advice among neurological patients: characteristics and outcomes.
        Health Serv Res. 2020;
        • England M.J.
        • et al.
        Epilepsy across the spectrum: promoting health and understanding.: a summary of the Institute of Medicine report.
        Epilepsy Behav. 2012; 25: 266-276
        • Long C.
        • Fureman B.
        • Dingledine R.
        2014 epilepsy benchmarks: progress and opportunities.
        Epilepsy Curr. 2016; 16: 179-181
      1. Healthy people 2020: disability and health.
        2020 ([cited 2020 July]; Available from:)
        • Alfandre D.J.
        "I'm going home": discharges against medical advice.
        Mayo Clin Proc. 2009; 84: 255-260
        • Ibrahim S.A.
        • Kwoh C.K.
        • Krishnan E.
        Factors associated with patients who leave acute-care hospitals against medical advice.
        Am J Public Health. 2007; 97: 2204-2208
        • Kahle C.H.
        • Rubio M.L.
        • Santos R.A.
        Discharges against medical advice: considerations for the hospitalist and the patient.
        Hosp Med Clin. 2015; 4: 421-429
        • Southern W.N.
        • Nahvi S.
        • Arnsten J.H.
        Increased risk of mortality and readmission among patients discharged against medical advice.
        Am J Med. 2012; 125: 594-602
        • Glasgow J.M.
        • Vaughn-Sarrazin M.
        • Kaboli P.J.
        Leaving against medical advice (AMA): risk of 30-day mortality and hospital readmission.
        J Gen Intern Med. 2010; 25: 926-929
        • Aliyu Z.
        Discharge against medical advice: sociodemographic, clinical and financial perspectives.
        Int J Clin Pract. 2002; 56: 325-327
        • Hoyer C.
        • et al.
        Uncompleted emergency department care and discharge against medical advice in patients with neurological complaints: a chart review.
        BMC Emerg Med. 2019; 19: 52
      2. 2014 Introduction to the NIS. Healthcare Cost and Utilization Project (HCUP). December 2016 [cited 2019 September, 25]; Available from: https://www.hcup-us.ahrq.gov/db/nation/nis/NIS_Introduction_2014.jsp.

      3. INTRODUCTION TO THE HCUP NATIONAL INPATIENT SAMPLE (NIS) 2014. 2016 [cited 2019 September, 25]; Available from: https://www.hcup-us.ahrq.gov/db/nation/nis/NISIntroduction2014.pdf.

        • Jette N.
        • et al.
        How accurate is ICD coding for epilepsy?.
        Epilepsia. 2010; 51: 62-69
        • Mbizvo G.K.
        • et al.
        The accuracy of using administrative healthcare data to identify epilepsy cases: a systematic review of validation studies.
        Epilepsia. 2020;
        • Moore B.J.
        • et al.
        Identifying increased risk of readmission and in-hospital mortality using hospital administrative data: the AHRQ elixhauser comorbidity index.
        Med Care. 2017; 55: 698-705
      4. Publishing with HCUP data. Healthcare cost and utilization project (HCUP).
        2020 ([cited 2020 January]; Available from:)
        • Alfandre D.
        Reconsidering against medical advice discharges: embracing patient-centeredness to promote high quality care and a renewed research agenda.
        J Gen Intern Med. 2013; 28: 1657-1662
        • Jeremiah J.
        • O’Sullivan P.
        • Stein M.D.
        Who leaves against medical advice?.
        J Gen Intern Med. 1995; 10: 403-405
        • Green P.
        • Watts D.
        • Dhopesh V.
        Why patients sign out against medical advice (AMA): factors motivating patients to sign out AMA.
        Am J Drug Alcohol Abuse. 2004; 30: 489-493
        • Alfandre D.
        Improving quality in against medical advice discharges—more empirical evidence, enhanced professional education, and directed systems changes.
        J Hosp Med. 2017; 12: 59-60
        • Alfandre D.
        Advancing the science of discharges against medical advice: taking a deeper dive.
        Heart. 2019; 105: 268-269
        • Macrohon B.C.
        Pediatrician’s perspectives on discharge against medical advice (DAMA) among pediatric patients: a qualitative study.
        BMC Pediatr. 2012; 12: 1-9
        • Spooner K.K.
        • et al.
        Discharge against medical advice in the United States, 2002–2011.
        Mayo Clin Proc. 2017; 92: 525-535
        • Kenne D.R.
        • Boros A.P.
        • Fischbein R.L.
        Characteristics of opiate users leaving detoxification treatment against medical advice.
        J Addict Dis. 2010; 29: 383-394
        • Kanner A.M.
        Depression and epilepsy: a new perspective on two closely related disorders.
        Epilepsy Curr. 2006; 6: 141-146
        • Siarava E.
        • et al.
        Depression and quality of life in patients with epilepsy in Northwest Greece.
        Seizure. 2019; 66: 93-98
        • Ballester N.
        • et al.
        An early warning tool for predicting at admission the discharge disposition of a hospitalized patient.
        Am J Manag Care. 2018; 24: e325-e331
        • Salimi M.
        • Rozovskaya V.
        Predicting discharge disposition using patient complaint notes in electronic medical records.
        Proceedings of the BioNLP 2018 Workshop. 2018;
        • Peck J.S.
        • et al.
        Predicting emergency department inpatient admissions to improve same‐day patient flow.
        J Acad Emerg Med Case Rep. 2012; 19: E1045-E1054