Patient Characteristics Associated With Readmission to 3 Neurology Services at an Urban Academic Center

Abstract

Background and Purpose:

Hospital 30-day readmissions in patients with primary neurological problems are not well characterized. We sought to determine patient characteristics associated with readmission across 3 different inpatient neurology services at New York University Langone Hospital.

Methods:

We retrospectively reviewed all 30-day readmissions from the General Neurology, Epilepsy, and Stroke services at NYULH Brooklyn and Manhattan campuses from 2016-2017 and compared them to a random sample of non-readmitted neurology patients. We used univariate analyses to compare demographics, clinical characteristics, disease specific metrics, and discharge factors of non-readmitted and readmitted groups and binomial logistic regression to examine specific variables with adjustment for confounders.

Results:

We included 284 patients with 30-day readmissions and 306 control patients without readmissions matched by discharge location and service. After adjusting for confounders, we found that the following factors were associated with increased readmission risk: a recent hospital encounter increased risk for all services, increased number of medications at discharge, intensive care unit stay, higher length of stay, and prior history of seizure for the General Neurology Service, increased number of medications at discharge for the Epilepsy Service, and active malignancy and higher discharge modified Rankin Scale score for the Stroke Service.

Conclusion:

This study identifies potential risk factors for readmission in patients across multiple neurology services. Further research is needed to establish whether these risk factors hold across multiple institutions.

Introduction

In the early 2000s, about 1-in-5 Medicare beneficiary admissions were readmitted to the hospital within 30 days, costing more than $17 billion annually in potentially avoidable Medicare expenses.¹ Focused on reducing readmissions, the Centers for Medicare and Medicaid Services’ (CMS) linked 30-day readmissions with hospital performance, and in 2012 the Affordable Care Act instituted penalties for excessive readmissions.² Initially focused only on readmissions for acute myocardial infarction, congestive heart failure, and pneumonia, the CMS later expanded metrics to include chronic obstructive pulmonary disease, elective hip or knee replacement, coronary artery bypass graft surgery, and stroke.³ This group is likely to expand in the near future and may include other common neurologic conditions. It is, therefore, important to better understand which patients are at highest risk for readmission in order to tailor interventions to that group.

In neurology, prior studies have focused on readmissions in epilepsy, intracranial hemorrhage, or ischemic stroke.^4-6 Less is known about readmissions in patients with other neurologic conditions.⁷ Also, prior studies have used administrative or national data sets, which may miss patient-level or neurology service-specific characteristics that are potentially important. Therefore, using retrospective medical record review, we sought to determine patient characteristics associated with readmission in the General Neurology, Epilepsy, and Stroke services at our urban, multi-campus, and tertiary care academic center.

Methods

This study and procedures were approved by the New York University Langone Health (NYULH) institutional review board. Using hospital administrative data, we identified all 30-day hospital readmissions (n = 284) within the General Neurology, Epilepsy, and Stroke services at NYULH Manhattan and Brooklyn campuses from 2016-2017. Hospital readmissions were defined as hospitalization to any service at either the Manhattan or Brooklyn campuses within 30 days of index admission after a discharge from a NYULH neurology service. At time of our study, the Manhattan and Brooklyn campuses were the only hospitals within the NYULH system that admitted acute medical-surgical patients. Emergency room visits without hospital admissions and outside hospitalizations were not included. From 4,344 total non-readmitted patients during the same time period, we also randomly selected an equally sized sample of non-readmitted patients matched based on discharge location and service. Only patients who were discharged by a neurology service were included. We excluded those who were deceased or discharged to hospice or another hospital system after index hospitalization. Patients were only included once in any group so that only 1 hospitalization or encounter per patient was included in this study.

Patients admitted to the Stroke service had a primary medical problem of proven ischemic or hemorrhagic stroke or clinical concern for stroke or transient ischemic attack as determined by a trained vascular neurologist. Patients admitted to the Epilepsy service had a primary medical problem of seizure or epilepsy and have had already established care with the outpatient epilepsy center. The General Neurology service admitted all other patients with primary neurologic problems including those with seizure or epilepsy without having had previously established care at the epilepsy center.

All neurology services at both NYULH campuses have similar workflows. Teams are comprised of an attending physician, trainees (residents and fellows), and an advanced practice provider. Daily rounds also include interdisciplinary meetings with case managers, social work, nurse managers, and therapy services (physical, occupational, speech and language). All services use similar discharge workflows through electronic medical record order sets, documentation, and verbal handoff depending on site of discharge. All services have similar methods to scheduling any outpatient follow up.

We performed a retrospective medical record review to collect patient demographics, comorbidities, number of recent hospital visits, admission characteristics such as elective admission vs emergency admission, index hospitalization diagnoses, length of stay, complications such as infections, discharge medications, discharge location, and follow up appointments provided. For each patient, we calculated the Charlson Comorbidity Index (CCI).⁸ Service specific variables of interest were also included. For the epilepsy service, we obtained information about seizure frequency, number of antiepileptics at discharge, history of vagal nerve stimulator (VNS), and prior epilepsy surgery. For the stroke service, these included discharge modified Rankin score (mRS) at discharge, presence of dysphagia, use of intravenous tissue plasminogen (tPA) or mechanical thrombectomy, and NIH Stroke Scale (NIHSS)/intracerebral hemorrhage (ICH) scores on admission.

Data Analysis

We used SPSS (IBM Corp, Armonk, NY) to compare demographics, clinical characteristics, and hospitalization factors between non-readmitted and readmitted groups within each service. For categorical variables, we used Pearson Chi-squared or Fisher’s exact test. We applied post-hoc pairwise comparisons using Bonferroni corrections. For continuous variables, we assessed distribution normality with the Shapiro-Wilk test. Only age was normally distributed. To compare continuous variables, we used independent samples t-test or Mann-Whitney U test. For the variables “at least 1 hospital visit within the last year” and “frequency of >8 seizures per month,” we compared the number of patients in the highest quartile of each service between non-readmitted and readmitted groups using Fisher’s exact test. To adjust for potential confounders, we used binomial logistic regression with non-readmission as reference group. We used a base regression model of adjustment that included age, sex, hypertension, hyperlipidemia and CCI for each service. For the Epilepsy service, the base model additionally included elective admission status and >8 seizures per month. For the Stroke service, the base model additionally included discharge mRS. We then analyzed additional variables of interest individually in the base model of adjustment. The 2-tailed values of p < 0.05 were considered statistically significant.

Results

We included a total of 284 readmitted patients (General [G] = 83, Epilepsy [E] = 59, Stroke [S] = 142) and 306 matched non-readmitted patients (G = 89, E = 79, S = 138) who were discharged from any neurology service at NYULH between 2016-2017. Our neurology 30-day readmission rate of about 6% is comparable to other single center studies.^9,10 By service, the readmission rates for General Neurology, Epilepsy, and Stroke services in 2016 and 2017 respectively are 12.2% and 10.9%, 3.7% and 3.4%, 8.8% and 8.2%. Stratified by readmission, Tables 1 –3 display patient and hospitalization characteristics for the General Neurology, Epilepsy, and Stroke services, respectively.

Table 1.

Characteristics of the General Neurology Service Stratified by Patients With 30-day Readmissions and Matched Control Patients Without Readmission

Characteristic	No readmission (n = 89)	Readmission (n = 83)	p-value
Age, year (sd)	52.7 (20.7)	56.4 (20.1)	0.275
Sex, n female (%)	57 (64.0)	49 (56.6)	0.351
Elective admissions, n (%)	3 (3.4)	4 (4.8)	0.713
Comorbidities, n (%)
Hypertension	27 (30.3)	31 (37.3)	0.339
Hyperlipidemia	16 (18.0)	20 (24.1)	0.353
Diabetes	11 (12.4)	13 (15.7)	0.661
CAD/MI	9 (10.1)	8 (9.8)	1.000
CHF	3 (3.4)	2 (2.4)	1.000
Prior Stroke	5 (5.6)	7 (8.4)	0.556
Premorbid Seizure disorder	6 (6.7)	16 (19.3)	0.021
COPD/Asthma	10 (11.2)	3 (3.6)	0.083
Malignancy	9 (10.1)	9 (10.8)	1.000
Psychiatric	27 (30.3)	20 (24.1)	0.395
Dementia	8 (9.0)	7 (8.4)	1.000
Trach/Peg	1 (1.1)	0 (0)	1.000
Charlson Comorbidity Index, median [IQR]	2 [0-4]	2 [0-4]	0.070
At least 1 hospital visit within last yr, n (%)	38 (42.7)	52 (62.7)	0.010
Index LOS, days [IQR]	4 [2-6]	5 [3-10]	0.027
Index Diagnosis, n (%)			0.006
Seizure/Epilepsy	10 (11.2)*	19 (22.9)*
Intracranial Hemorrhage	12 (13.5)	6 (7.2)
CNS Demyelinating Disease	25 (28.1)*	9 (10.8)*
Primary Headache	7 (7.9)	6 (7.2)
Brain Mass	6 (6.7)*	16 (19.3)*
Myelopathy	3 (3.4)	4 (4.8)
Neuromuscular	3 (3.4)	5 (6.0)
Neuropathy	6 (6.7)	2 (2.4)
Movement/Gait	5 (5.6)	2 (2.4)
Encephalopathy	6 (6.7)	12 (14.5)
Other	6 (6.7)	2 (2.4)
ICU stay, n (%)	4 (4.5)	12 (14.5)	0.034
# Meds at discharge, median [IQR]	6 [2-8]	8 [4-11]	<0.001
Discharge Location, n (%)			0.507
Home	55 (61.8)	44 (53.0)
Rehab	27 (30.3)	31 (37.3)
Other	7 (7.9)	8 (9.6)

sd = standard deviation; CAD/MI = coronary artery disease/myocardial infarction; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; CKD/ESRD = chronic kidney disease/end stage renal disease; IQR = interquartile range; LOS = length of stay; ICU = intensive care unit *p < 0.05 in post-hoc Bonferroni method

Table 2.

Characteristics of the Epilepsy Service by Readmission Stratified by Patients With 30-day Readmissions and Matched Control Patients Without Readmission

Characteristic	No readmission (n = 79)	Readmission (n = 59)	p-value
Age, year (sd)	32.5 (21.4)	33.2 (29.0)	0.857
Sex, n female (%)	43 (54.4)	30 (50.8)	0.732
Elective admissions, n (%)	68 (86.1)	25 (42.4)	<0.001
Comorbidities, n (%)
Hypertension	8 (10.1)	9 (15.3)	0.436
Hyperlipidemia	4 (5.1)	7 (11.9)	0.205
Diabetes	4 (5.1)	2 (3.4)	1.000
Prior Stroke	4 (5.1)	2 (3.4)	1.000
Premorbid Seizure disorder	68 (86.1)	47 (79.7)	0.360
COPD/Asthma	6 (7.6)	7 (11.9)	0.557
Malignancy	5 (6.3)	1 (1.7)	0.238
Psychiatric	25 (31.6)	27 (45.8)	0.111
Dementia	3 (3.8)	6 (10.2)	0.171
Charlson Comorbidity Index, median [IQR]	0 [0-2]	0 [0-2]	0.131
Epilepsy Characteristics, n (%)
Frequency of >8 seizures per mo	14 (17.7)	22 (37.3)	0.041
VNS?	2 (2.5)	7 (11.9)	0.038
Epilepsy Surgery?	4 (5.1)	7 (11.9)	0.205
At least 1 hospital visit within last yr	16 (20.3)	31 (52.5)	<0.001^
Index LOS, days [IQR]	2 [2-5]	4 [2-8]	0.001
Index Diagnosis, n (%)			0.393
Seizure/Epilepsy	75 (94.9)	58 (98.3)
Other	4 (5.1)	1 (1.7)
ICU stay, n (%)	1 (1.3)	5 (8.5)	0.084
Meds at discharge, median [IQR]	4 [2-6]	6 [3-10]	0.001
AEDs at discharge, median [IQR]	1 [1-2]	2 [1-3]	0.015^
Discharge Location, n (%)			0.351
Home	70 (88.6)	47 (79.7)
Rehab	3 (3.8)	4 (6.8)
Other	6 (7.6)	8 (13.6)

sd = standard deviation; CAD/MI = coronary artery disease/myocardial infarction; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; CKD/ESRD = chronic kidney disease/end stage renal disease; IQR = interquartile range; VNS = vagal nerve stimulator; LOS = length of stay; ICU = intensive care unit; AEDs = anti-epileptic drugs ^p < 0.05 when elective admissions excluded (Supplementary Table 1).

Table 3.

Characteristics of the Stroke Service by Readmission Stratified by Patients With 30-day Readmissions and Matched Control Patients Without Readmission

Characteristic	No readmission (n = 138)	Readmission (n = 142)	p-value
Age, year (sd)	67.8 (15.0)	71.3 (17.5)	0.019
Sex, n female (%)	76 (55.1)	64 (45.1)	0.094
Elective admissions, n (%)	0 (0)	0 (0)	n/a
Comorbidities, n (%)
Hypertension	99 (71.7)	116 (81.7)	0.065
Hyperlipidemia	72 (52.2)	90 (63.4)	0.069
Diabetes	47 (34.1)	54 (38.0)	0.534
CAD/MI	32 (23.2)	34 (23.9)	0.889
CHF	19 (13.8)	21 (14.8)	0.865
Prior Stroke	29 (21.0)	33 (23.2)	0.669
Premorbid Seizure disorder	4 (2.9)	12 (8.5)	0.069
COPD/Asthma	17 (12.3)	14 (9.9)	0.570
CKD/ESRD	10 (7.2)	15 (10.6)	0.404
Malignancy	9 (6.5)	26 (18.3)	0.003
Psychiatric	11 (8.0)	17 (12.0)	0.321
Dementia	5 (4.6)	20 (14.1)	0.003
Trach/Peg	1 (0.7)	7 (4.9)	0.067
Charlson Comorbidity Index, median [IQR]	2 [0-5]	4 [3-6]	0.009
Stroke Characteristics
NIHSS, [IQR]	4 [2-8]	4 [1-13]	0.395
ICH score, [IQR]	0 [0 -1]	1 [0-2]	0.382
tPA, n (%)	20 (20.8)	13 (13.5)	0.251
Thrombectomy, n (%)	12 (12.5)	20 (20.8)	0.175
Dysphagia, n (%)	55 (39.9)	80 (56.3)	0.006
At least 1 hospital visit within last yr	37 (26.8)	64 (45.1)	0.002
Index LOS, days [IQR]	3 [2-7]	5 [2-9]	0.002
Index Diagnosis, n (%)			0.004
Ischemic Stroke	96 (69.6)	96 (67.6)
Hemorrhagic Stroke	7 (5.1)*	22 (15.5)*
Transient Ischemic Attack	7 (5.1)	4 (2.8)
Seizure/Epilepsy	1 (0.7)	5 (3.5)
Other	27 (19.6)*	15 (10.6)*
ICU stay, n (%)	23 (16.7)	32 (22.5)	0.232
Discharge mRS, median [IQR]	1 [0-3]	3 [1-4]	<0.001
Meds at discharge, median [IQR]	7 [4-9]	9 [6-12]	<0.001
Discharge Location, n (%)			0.002
Home	89 (64.5)*	62 (43.7)*
Rehab	45 (32.6)*	73 (51.4)*
Other	4 (2.9)	7 (4.9)

sd = standard deviation; CAD/MI = coronary artery disease/myocardial infarction; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; CKD/ESRD = chronic kidney disease/end stage renal disease; IQR= interquartile range; NIHSS = NIH Stroke Scale; ICH = Intracerebral hemorrhage; tPA = tissue plasminogen activator; LOS = length of stay; ICU = intensive care unit; mRS = modified Rankin score *p < 0.05 in post-hoc Bonferroni method

Overall Findings

When comparing readmitted against non-readmitted patients within each service, we found similar distributions of age and sex, except in the stroke service, in which readmitted patients were older (71 vs 68 years, p = 0.019). Elective admissions were minimal (less than 5%) for both the readmitted and the non-readmitted group in the General Neurology service (p = 0.713) and there were no elective admissions in the Stroke service. However as expected, in the Epilepsy service, elective admissions made up a large proportion of admissions and the readmitted group had less electively admitted patients (42.5% vs 86.1%, p ≤ 0.0001). In all services, readmitted patients had a longer length of stay (LOS), were more likely to have at least 1 hospital or emergency department (ED) visit within the year prior, and had more medications at discharge. Readmitted patients in each service had about a 2 day longer index hospitalization and 2 more medications at discharge compared to non-readmitted patients.

General Neurology Service

Regarding service specific characteristics, readmitted patients in the General Neurology service were more likely to have a premorbid seizure disorder (19.3% vs 6.7%, p = 0.021), index diagnoses of seizure or epilepsy (22.9% vs 11.2%, Bonferroni correction p < 0.05) and brain mass (19.3% vs 6.7%, Bonferroni correction p < 0.05), and an intensive care unit (ICU) stay (14.5% vs 4.5%, p = 0.034). The readmitted group was less likely to have index diagnosis of CNS demyelinating disease than the non-readmitted group (10.8% vs 28.1%, Bonferroni correction p < 0.05). In those with seizure disorder in the General Neurology service, we also evaluated seizure frequency, presence of VNS or epilepsy surgery, and number of AEDs at discharge, but these were non-significant.

Epilepsy Service

Within the Epilepsy service, characteristics associated with readmission included greater than 8 seizures per month (37.3% vs 17.7%, p = 0.041), greater AEDs at discharge (median [IQR]: 2 [1-3] vs 1 [1-2], p = 0.015), and prior or current VNS (11.9% vs 2.5%, p = 0.038). Comparing only non-elective admissions (n = 45), readmission patients also were more likely to have at least hospital encounter within the year (61.8% vs 27.3%, p = 0.046) and higher AEDs at discharge (median [IQR]: 1 [1-2] vs 2.5 [1-4], p = 0.037) (denoted by caret in Table 2, and displayed in Supplemental 1). No other Epilepsy service specific differences between readmission groups were found.

Stroke Service

Readmitted patients in the Stroke service were more likely to have an index diagnosis of hemorrhagic stroke (15.5% vs 5.1%, Bonferroni correction p < 0.05), active malignancy (18.3% vs 6.5%, p = 0.003), comorbidity of dementia (14.1% vs 4.6%, p = 0.003), dysphagia as a result of their stroke (56.3% vs 39.9%, p = 0.006), higher discharge mRS (median [IQR]: 3 [1-4] vs 1 [0-3], p < 0.001), and discharge to rehab (51.4% vs 32.6%, p = 0.002). Those with lower admitting suspicion of stroke, such as peripheral vertigo, primary headache, and nonspecific symptoms, were categorized as “other” and less likely to be readmitted. Additionally, unlike other services, readmitted patients in the Stroke service had higher CCI than non-readmitted patients (median [IQR]: 4 [3-6] vs 2 [0-5], p = 0.009). Comparisons of NIHSS, ICH score, administration of IV tPA, or mechanical thrombectomy were not statistically different.

Regression Models

When adjusting for potential confounders (Table 4), many of our findings remain statistically significant. Within every service, having at least 1 hospital or ED visit within the preceding year increased odds for readmission by at least 2-fold. Within the general neurology service, premorbid seizure disorder (Odds Ratio (OR) [95% Confidence Interval (CI)]: 3.23 [1.18-8.81]), higher LOS (OR [95% CI]: 1.08 [1.02 -1.14]), ICU stay (OR [95% CI]: 4.13 [1.25-13.64]), higher number of discharge medications (OR [95% CI]: 1.15 [1.06 -1.25]) were still associated with readmission after adjustment of confounders. In the Epilepsy service, elective readmissions were associated with less readmissions (OR [95% CI]: 0.10 [0.04-0.25]). When included as a confounder, most differences from univariate comparisons were attenuated except for number of discharge medicines (OR [95% CI]: 1.18 [1.04 -1.35]). After excluding Epilepsy service elective admissions, both recent hospital encounters and discharge medicines were still associated with readmission (denoted by caret in Table 4, and displayed in Supplemental Table 2). Within the Stroke service, higher discharge mRS (OR [95% CI]: 1.46 [1.24 -1.73]) and malignancy (OR [95% CI]: 3.15 [1.24-8.02]) remained statistically associated with readmission after adjustment for confounders.

Table 4.

Binomial Logistic Regression Analysis for Readmission Factors

Characteristic	Service
	General (n = 172)			Epilepsy (n = 138)			Stroke (n = 280)
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Base Model
Age	0.993	0.966 -1.021	0.609	0.981	0.955 -1.008	0.159	0.989	0.965 -1.012	0.336
Female Sex	0.716	0.383 -1.338	0.295	1.084	0.480-2.450	0.846	0.664	0.401 -1.097	0.110
Hypertension	0.996	0.446-2.224	0.991	2.052	0.489-8.607	0.326	1.066	0.535-2.123	0.856
Hyperlipidemia	1.068	0.455-2.508	0.880	2.360	0.470-11.85	0.297	1.235	0.704-2.165	0.461
CCI	1.187	0.924 -1.526	0.180	1.234	0.873 -1.744	0.233	1.092	0.930 -1.282	0.284
Elective admission				0.101	0.041-0.252	<0.001
>8 sz/mo				2.446	0.944-6.337	0.066
Discharge mRS							1.463	1.238 -1.728	<0.001
Additional Comorbidities
Prior Stroke	1.125	0.314-4.024	0.856	0.245	0.024-2.531	0.238	0.838	0.447 -1.571	0.582
Seizure disorder	3.227	1.183-8.806	0.022	0.723	0.245-2.135	0.557	3.473	0.998-12.09	0.050
Malignancy	0.773	0.256-2.337	0.649	n/a	n/a	n/a	3.151	1.238-8.021	0.016
Psychiatric	0.800	0.400 -1.602	0.529	1.673	0.692-4.041	0.253	1.771	0.745-4.210	0.196
Dementia	0.640	0.197-2.084	0.459	4.758	0.798-28.38	0.087	2.800	0.949-8.259	0.062
Trach/Peg	n/a	n/a	n/a	n/a	n/a	n/a	2.821	0.321-24.83	0.350
Epilepsy Characteristics, n (%)
VNS?				2.741	0.433-17.35	0.284
Epilepsy Surgery?				2.195	0.460-10.488	0.324
Stroke Characteristics, n (%)
NIHSS							0.991	0.947 -1.036	0.685
ICH score							1.057	0.144-7.753	0.957
tPA							0.770	0.343 -1.726	0.525
Thrombectomy							1.539	0.674-3.514	0.307
Dysphagia							1.162	0.662-2.037	0.601
At least 1 hospital visit within last yr	2.064	1.082-3.938	0.028	3.257	1.356-7.823	0.008^	2.771	1.538-4.993	0.001
LOS days	1.080	1.019 -1.144	0.009	1.067	0.981 -1.162	0.132	1.019	0.974 -1.065	0.419
ICU stay	4.135	1.253-13.641	0.020	2.399	0.147-39.18	0.539	1.086	0.560-2.104	0.808
# Discharge meds	1.152	1.058 -1.254	0.001	1.184	1.039 -1.350	0.011^	1.044	0.977 -1.116	0.200
# AEDs				2.084	0.838-5.187	0.114

Reference category is non-readmitted group. Base model for adjustment includes age, sex, hypertension, hyperlipidemia, and Charlson Comorbidity Index (CCI). Epilepsy service additionally adjusted for: elective admission and number of seizure (sz) > 8/mo Stroke service additionally adjusted for: discharge mRS. mRS = modified Rankin score; VNS = vagal nerve stimulator; NIHSS = NIH Stroke Scale; ICH = Intracerebral hemorrhage; tPA = tissue plasminogen activator; LOS = length of stay; ICU = intensive care unit; AEDs = anti-epileptic drugs ^p < 0.05 when elective admissions excluded (Supplementary Table 2).

Discussion

In this retrospective study, we found common and service-specific factors associated with readmission in neurologic patients. Prior studies on readmissions in neurologic patients have attempted to aid in this, but have mostly examined patients admitted for seizure or stroke, and have used mostly administrative databases which may miss patient-level information.^5,9 Our study adds to the literature by simultaneously examining 3 common neurological service lines and by highlighting and commenting on the potential association with readmission between prior hospital encounters, medication burden, and service unique risk considerations. As inpatient providers are increasingly motivated to decrease readmission rates, knowing these patient factors that place patients at increased risk for readmission is ever more important.

To the best of our knowledge, the relation between prior hospital encounters and readmission in patients with primary neurologic issues has not been well examined. In a single stroke study, a greater number of hospitalizations within the year prior to stroke admission was also found to be an important predictor of 30-day readmission.¹¹ Similar findings have also been described in medical and psychiatric services.^12-14 However, these prior studies did not include emergency room visits which are important as ED visits are increasing and CMS recently introduced measures designed to capture “excess days in acute care” 30 days after discharge.^15-17 Therefore, looking at both prior hospitalizations and ED visits will likely increase in importance when assessing risk factors in neurologic readmissions.

In terms of medication burden, we find that higher number of medicines at discharge in the general neurology and epilepsy services are associated with readmission after multivariable analysis. High medication burden, such as polypharmacy, is variably defined but often set at 5 or more medications.¹⁸ This likely represents a combination of both inappropriately prescribed medicines and burden of chronic conditions. Though, even in context of multi-morbidity, polypharmacy has important clinical implications.¹⁹

In neurology, effects of polypharmacy have mostly been investigated in cognitive impairment and multiple sclerosis where there was found to be an association with adverse effects like hospitalization.^20,21 Although studies in the general geriatric and psychiatry populations both show an association between polypharmacy and readmissions, risks for readmission in neurologic patients with polypharmacy have not been directly studied before.^22,23

In our study, excluding non-readmitted epilepsy patients (who comprise a younger population), all groups had median total discharge medications that met threshold for polypharmacy and every readmitted group had 2 more medications on average than those not readmitted. Although appropriate polypharmacy may be common and a marker of comorbidity burden, there was no difference in total number of comorbidities or in the Charlson Comorbidity Index (CCI) scores between non-readmitted and readmitted groups in our general neurology and epilepsy samples. Further, greater discharge medications in these services were still associated with readmission after adjustment for CCI. This may reflect both relatively lower number of medical comorbidities in the general neurology and epilepsy populations, unlike the Stroke service, and also more refractory disease, especially in epilepsy, which the CCI does not address. Other studies have shown an association with the CCI score and readmission among patients with non-stroke neurological diagnoses, however these studies were larger, older, did not distinguish service lines, or were disease-specific.^24,25 Altogether, in our General Neurology and Epilepsy service populations, high medication burden was common and represented a potential independent marker for readmission risk, to a greater extent than comorbidity burden.

Lastly, many service specific risk factors for readmission were identified which is important given the varied patient populations within each discipline. In the general neurology service, we find ICU stay, higher LOS, and index diagnosis of seizure and/or brain tumor were associated with readmissions which likely represent severity of disease and have been shown in other studies^5,7,25,26 We found those with a primary index diagnosis of CNS demyelinating disease were less likely to be in the readmitted group, but this diagnosis group consisted of younger patients (mean 41 years, sd 2.5), which was previously shown to have less readmissions than older patients.²⁴ Although seizures and epilepsy disorder are a well-known cause of neurologic readmission,⁵ specific seizure or epilepsy characteristics, such as seizure frequency or number of antiepileptics that may also have predictive value in readmissions have not been studied. In our epilepsy service, we were able to assess seizure burden and find that those with the highest seizure burden, classified as >8 seizures per month, had trend but not significance toward readmission in adjusted models. Although medically refractory epilepsy has been shown to be associated with readmissions,²⁷ further quantifying seizure burden may help stratify risk of readmission in patients with epilepsy.

Finally, within our stroke service, we find readmitted patients more often have older age, active malignancy, dementia, dysphagia as a result of stroke, greater discharge mRS, and greater CCI scores on admission than non-readmitted patients. Malignancy and discharge mRS remain significantly different after adjustment for confounders, including CCI. In the literature, readmissions in stroke are well-studied and common. Premorbid chronic conditions, which are often pooled, are increasingly recognized as potential influencers in readmission after ischemic stroke, transient ischemic attack, and coronary artery stenting.^28,29 Specifically, cancer patients with stroke may suffer higher readmission rates at 90-days.^25,30

Additionally, functionality after stroke is an important factor in readmissions. Those with higher stroke severity and worse functional outcomes are at risk for readmission, which are frequently due to infection.³¹ NIHSS and mRS are often cited as markers of stroke severity and post-stroke functionality.³² We did not find a difference in initial NIHSS between readmission groups which may reflect that although NIHSS is a marker of stroke severity, it may not be a strong marker of disability.³³ However, we find higher discharge mRS and discharge to acute rehabilitation was associated with readmission which highlights the association between disability and stroke readmissions.³⁴

Limitations and Conclusions

Limitations of this study include the retrospective design and single academic center population which may limit generalizability. Certain groups had small sample sizes. For instance, we were unable to fully analyze seizure burden in the general neurology group due to the sample size. Not surprisingly, our Epilepsy service had a high proportion of elective admissions. Although certain associations persisted after we adjusted for elective admissions or excluded elective admissions, we acknowledge the limitations the small sample of non-electively admitted patients. While we were able to obtain more granular patient-level detail than administrative databases, our medical record review relied on accurate documentation. Presentation to outside non-index hospitals can complicate readmission risk stratification.³⁵ In our study, information on hospital encounters at outside institutions before or after index hospitalization were not available.

One strength of this study is that most of the variables used were objective and easily found in the medical record, thus not subject to recall bias. Additionally, our inpatient service use of standardized documentation and clinical measures such as seizure frequency in the epilepsy service and premorbid and discharge mRS in the stroke service strengthened our analysis.

From our retrospective study, we have identified potential common and service-specific risk factors for 30-day hospital readmission in neurologic patients. Further studies are needed to validate and refine more detail on these specific possible risk factors. However, we hope this information serves as a framework in assessing patients’ risk of being readmitted in order to guide physicians to further intervene on those most vulnerable.

Next steps could include broadening the patient population in a multi-system approach to further refine and validate variables, using these results to establish clinically useful parameters for assessing readmission risk, and characterizing neurologic readmission preventability. Altogether, understanding these risk factors for readmission in neurologic patients may help tailor service or hospital interventions and inform healthcare policy.