Hospital Readmission Rates in Patients With Neuromyelitis Optica Spectrum Disorder

Neuromyelitis optica spectrum disorder (NMOSD) is an uncommon antibody-mediated demyelination disorder of the central nervous system. Initially considered a variant of multiple sclerosis (MS), in 2004 it was classified as a separate disease when the specific causative IgG antibodies against aquaporin 4 were discovered.¹  The clinical course is recurrent debilitating relapses affecting the spinal cord (transverse myelitis) and the brain (optic neuritis or brainstem syndromes).²  Early diagnosis and treatment is crucial because relapses result in severe damage to the central nervous system and negatively affect quality of life. In the past several years, the Food and Drug Administration has approved novel maintenance therapies for NMOSD.³ 

Flares in patients with NMOSD occur more frequently than flares in patients with MS and usually result in longer lasting and more debilitating deficits.⁴  Treatment of flares usually requires hospitalization and subsequent immunomodulatory therapy with corticosteroids, intravenous immunoglobulin (IVIg), or plasma exchange (PLEX). In certain situations, patients with NMOSD may be given immunomodulatory infusions such as rituximab and eculizumab to prevent further relapse activity.⁵  The range of recovery after a flare is variable: some patients have no improvement, and a minority have full recovery.⁴  Known risk factors that can help predict an aggressive disease course include Black race, older age, and male sex.^6–8  These factors help clinicians stratify patients according to risk of aggressive disease. Patients with aggressive disease may require more frequent hospital readmissions for continued immunomodulatory therapy. Therefore, analysis of patient rehospitalization will help clinicians by identifying additional common features of high-risk patients.

During the past several years, significant emphasis has been placed on hospital readmission rates to gauge health care quality with direct implications on compensation.⁹  There have been multiple studies examining the hospital readmission rates of other autoimmune diseases.^10,11  Analyzing readmission rates will help identify patients with therapy-resistant NMOSD flares, which may help physicians triage patients with NMOSD accordingly. The purpose of this study was to identify characteristics of patients at high risk for 30-day readmission to the hospital.

The NMOSD readmissions were identified through discharge data from the Nationwide Readmissions Database (NRD) of the Healthcare Cost and Utilization Project (HCUP) of the Agency for Healthcare Research and Quality for 2017.¹²  This database contains approximately 18 million admissions to US hospitals (weighted estimate to ~36 million discharges). The weighted estimates are used to produce the national estimates of occur-rences. The NRD recommends using weighted estimates for assessments of the entire US population. Each patient is tracked with a deidentified unique number that provides information about patient readmissions.

Clinical Classifications Software Refined/International Classification of Diseases, Tenth Revision (ICD-10) codes were used to define NMOSD (TABLE S1, available online at IJMSC.org). All patients with a primary diagnosis of NMOSD with index hospitalizations in 2017 were pooled. We excluded patients who were admitted under observation, were evaluated in the emergency department or urgent care without admission, died during the index hospitalization, were admitted in December 2017 (due to lack of adequate 30-day follow-up data), or had incomplete data. The primary outcome variable was the first nonelective readmission within 30 days of discharge from the index hospitalization. If a patient had multiple readmissions within 30 days of discharge, only the first readmission was used in the statistical analysis. The study was approved with exempt status by the University of South Alabama institutional review board.

Several baseline patient and hospital characteristics were collected: age, sex, true cost of hospitalization, length of stay, weekend admission, insurance status, quartile of median household income, All Patient Refined Diagnosis Related Groups (APR-DRGs): likelihood of dying, APR-DRG: loss of function, hospital bed size, hospital teaching status, discharge location, control/ownership of hospital, hospital designation, patient comorbidities, and treatment interventions during the index hospitalization (TABLE S2). A cost analysis was performed to determine health care costs, multiplying hospital charges by the Agency for Healthcare Research and Quality’s all-payer cost-to-charge ratios. The effects of several patient comorbidities and immunomodulatory therapies (ie, PLEX, IVIg, immunomodulatory infusions, including any chemotherapeutic agents or monoclonal antibody treatment) on readmission rates were analyzed. Typical agents used during NMOSD flares are rituximab and eculizumab.¹³  Etiology of readmission was determined by the first-listed Clinical Classifications Software Refined/ICD-10 diagnosis code (TABLE 1). All statistical testing was performed using weighted admissions as recommended by HCUP. Statistical analysis was performed as described previously using IBM SPSS Statistics for Windows, version 1.0.0.118 (IBM Corp).^10,14  Continuous variables were analyzed with a 2-tailed t test, and categorical variables were analyzed using the Pearson χ² test. Standard errors were calculated using HCUP’s alternative method to account for individual hospitals without NMOSD admissions in the NRD (hcup-us.ahrq.gov/tech_assist/standar derrors/508/508course_2016.jsp#importance). Multivariable logistic regression analysis was performed to determine the predictors of readmission (TABLE 2).

In 2017, there were 2447 weighted (1350 unweighted) admissions with NMOSD as the primary admission diagnosis; of these, 292 (169 unweighted) patients were readmitted within 30 days of discharge. The 30-day readmission rate for NMOSD was 11.9% (95% CI, 10.6%-13.3%) (FIGURE S1). The readmission frequency was highest in the first days after discharge, with half of all readmissions occurring within 13 days of discharge from the index hospitalization.

The baseline patient and hospital characteristics for patients readmitted and those not readmitted are shown in Table S2. Older patients (P < .001) and men (P = .014) were readmitted more frequently. Patients with longer length of index hospitalization stay (P < .001) and higher cost of index hospitalization (P < .001) were significantly more likely to be readmitted. Before adjustment for other variables (eg, age), patients with private insurance were less likely to be readmitted than patients with Medicare (P < .001). No other insurance variable differed significantly from Medicare variables. No readmission differences were seen between groups for household income quartile, hospital bed size, hospital designation, and hospital ownership. Patients with a higher risk of mortality (APR-DRG: likelihood of dying) and morbidity (APR-DRG: loss of function) were more likely to be readmitted in a severity-dependent manner (P < .001). Patients admitted to teaching metropolitan hospitals were more likely to be readmitted than patients at nonteaching metropolitan hospitals (P < .001). Patients discharged home with self-care were less likely to be readmitted than those discharged to all other locations (P < .001).

The effect of patient comorbidities and treatments during the index hospitalization on readmission rates is shown in Table S2. A higher frequency of readmissions was seen among patients with congestive heart failure, acute kidney injury, chronic obstructive pulmonary disease, chronic kidney disease, hypertension, peripheral vascular disease (PVD), anemia, cancer, neurocognitive disorders, and neurologic blindness. None of the analyzed comorbidities resulted in lower readmissions rates. There were no significant differences in readmission rates with any of the index hospitalization treatments (ie, IVIg, PLEX, and immunomodulatory treatments).

The primary readmission ICD-10 diagnosis is shown in Table 1. The 3 recurring organ system complications necessitating readmission were neurologic (50.2%), infectious (15.3%), and respiratory (6.0%). The most frequent diagnosis for readmission was NMOSD (25.6%). The most frequent nonneurologic diagnosis for readmission was septicemia (12.0%).

Multivariate logistic regression analysis was performed to identify predictors of 30-day readmission (Table 2). Patients aged 65 to 74 years (odds ratio [OR], 1.954; P = .042) had an increased rate of readmission, and all other age groups had a similar rate of readmission. Accounting for other comorbidities, sex had no effect on readmission odds. The only insurance status protective of readmission was private insurance (OR, 0.528; P < .001) relative to the Medicaid population. Hospital teaching status affected readmission odds, as both teaching metropolitan hospitals (OR, 0.535; P < .001) and nonmetropolitan hospitals (OR, 0.425; P = .042) were protective of readmission relative to nonteaching metropolitan hospitals. Several comorbidities were shown to be predictive of readmission: respiratory failure (OR, 2.952; P < .001), PVD (OR, 2.401; P = .005), neurocognitive disorders (OR, 3.944; P < .001), and neurologic blindness (OR, 2.159; P < .007). Only lipid disorders resulted in decreased odds of readmission (OR, 0.624; P = .016). Whereas IVIg and immunomodulatory infusions did not affect odds of readmission, PLEX increased odds of readmission (OR, 1.483; P = .025).

This retrospective study showed that approximately 11.9% (95% CI, 10.6%-13.3%) of patients with NMOSD are readmitted to the hospital within 30 days of discharge. This is similar to readmission rates for multiple sclerosis (11%)¹⁰  and other nervous system diseases (14.0%).¹⁵  A stronger female predominance has been reported for NMOSD than for MS, ranging from 66% to 88%.¹⁶  In the present study, women comprised 78.5% of all NMOSD hospitalizations (Table S2), and although a lower proportion of female patients were readmitted (Table S2), it is not statistically significant after accounting for comorbidities (Table 2). This contrasts with MS, where female sex is found to be protective of readmission.¹⁰  Although this may be partly due to the smaller sample size in this study, male sex has been shown to result in an aggressive phenotype of NMOSD.⁶  This may also be due to increased comorbidities seen in men resulting in readmission. Although age is usually an important factor in hospital readmission, after stratification of patients by age, the oldest group (>75 years) and the youngest group (<44 years) have lower odds of readmission than the middle 2 groups (45-64 years and 65-74 years). The group aged 65 to 74 years had statistically significantly higher odds of readmission relative to the oldest group (>75 years) (Table 2). This contrasts with MS, where age is not predictive of readmissions.¹⁰  This finding may suggest that disease activity in NMOSD peaks in the sixth or seventh decade of life and then begins to decrease. All insurance types had similar odds of readmission except for private insurance, which had decreased odds. A similar trend is seen in the general population¹⁵  and in patients with MS.¹⁰  Interestingly, hospital teaching status also affects the odds of readmission. Patients are more likely to be read-mitted after treatment at a nonteaching metropolitan hospital than after treatment at a teaching metropolitan hospital or a nonmetropolitan hospital. This contrasts with MS, where both nonteaching metropolitan hospitals and teaching metropolitan hospitals had similar odds of readmission.¹⁰  This finding may be due to the decreased familiarity with the treatment of NMOSD flares vs MS flares among general neurologists. Neurologists specializing in autoimmune diseases are more likely to be employed by teaching metropolitan hospitals, and although there are overlaps between treatment of MS and NMOSD, NMOSD flares frequently require more aggressive management.³ 

Several patient comorbidities were analyzed to determine predictive value for rehospitalization. After accounting for age, sex, insurance status, and hospital teaching status, several comorbidi-ties increased the odds of readmission: respiratory failure, PVD, neurocognitive disorders, and neurologic blindness (Table 2). Respiratory failure and PVD have a similar effect in MS,¹⁰  possibly resulting from the increased oxidative stress and inflammation seen with both conditions.¹⁷  Neurocognitive disorders increase the odds of readmission, possibly due to the decreased cerebral cognitive reserve and poorer baseline function seen with these conditions. Neurologic blindness results in increased odds of readmission. Optic neuritis is one of the classical NMOSD flares and results in rapid loss of vision. Interestingly, neither neurocognitive disorders nor neurologic blindness affected readmission in MS.¹⁰  The only comorbidity resulting in a decreased OR of readmission was lipid disorders. This is unexpected because hyperlipidemia also increases oxidative stress such as PVD. However, hyperlipidemia did not affect readmission or worsen functional status in MS as did other conditions, such as hypertension, diabetes, and obstructive lung disease.¹⁸  These findings suggest that hyperlipidemia may have a unique role in demyelinating diseases, and further research is needed to elucidate the effect of hyperlipidemia in NMOSD.

Immunomodulatory therapies such as IVIg and PLEX are frequently used for NMOSD flares, and for relapse prevention, immunomodulatory infusions of monoclonal antibodies are started during the hospitalization.^19,20  Treatment with PLEX increased the odds of readmission, and IVIg and immunomodulatory infusions showed no significant effect on readmission. There was no effect seen with IVIg or PLEX in MS.¹⁰  PLEX is the most frequent treatment modality in the study, and it is likely that it is statistically significant due to the higher sample size. Another possibility is that PLEX has inherent risks, such as thrombosis and infection, that may result in readmission; however, previous studies have not shown increased readmissions with PLEX, suggesting that this may be less likely. More research is needed to determine the optimal immunomodulatory therapy for the initial NMOSD flare and subsequent maintenance therapy during the initial hospitalization.

Analysis of hospital readmission etiologies revealed that the most common cause of readmissions was neurologic complications (50.2%) (Table 1). The most common readmission diagnosis was NMOSD (25.6%). Other recurrent etiologies of readmission were infectious (15.3%) and respiratory (6.0%) complications (Table 1). These are similar to etiologies for readmission for MS.¹⁰  Immunomodulatory therapies for NMOSD flares can suppress the immune system and lead to an increased infection rate. A high proportion of readmissions due to neurologic etiologies suggests incomplete initial NMOSD flare treatment and that stronger initial immunomodulatory therapies may decrease readmission. However, the risk and benefit of each immunomodulatory therapy must be weighed before making such a decision.

Although this study successfully determined characteristics to identify patients with NMOSD at high risk for hospital readmission within 30 days of discharge, there are certain limitations to this type of study. As a retrospective analysis, only correlation can be inferred, and no statement of causation can be made. The NRD stores only admission data, so any patient who is seen in the emergency department or urgent care after discharge or admitted to the hospital for observation would not be recorded by the database. The database also does not include medications taken by the patient. Currently, there are several medications approved by the Food and Drug Administration for NMOSD, and these medications may affect readmission rates, but that information is unavailable. Furthermore, misclassification bias is present in this study. Although the goal of ICD-10 codes is to improve diagnostic accuracy, patients may be misdiagnosed as having NMOSD. This is evidenced by a small proportion of patients readmitted with the diagnosis of MS (Table 1). Another limitation is that other cofactors not tested in this study may impact readmission rates. There may also be synergistic roles between some risk factors that have not been explored. The NRD includes approximately 60% of all hospital admissions in the United States, and diagnostic inaccuracies are expected. The results of this study can be incorporated into emergency medical records, and use of artificial intelligence and machine learning can help develop algorithms to predict patients at high risk for readmission. This can help redistribute health care resources to result in better patient outcomes.

Currently, NMOSD is undergoing a pharmacologic revolution, with a variety of different therapies soon to be available for patients. Improved understanding and predictors of patients at high risk for disease activity will allow for better stratification. This study identified several characteristics that increased patient risk of hospital readmission. Future efforts could include the development of higher-efficacy immunomodulatory therapies for high-risk patients to decrease hospital readmissions and overall disease burden.