US Hospitalizations and 60-Day Readmission Rates Associated With Herpes Simplex Virus Encephalitis: Analysis of All Cause Readmissions and Encephalopathy Associated Readmissions

Abstract

Background

Herpes simplex encephalitis (HSE) is the most common cause of encephalitis hospitalizations. We sought to describe and analyze features associated with all cause readmissions and encephalopathy associated readmissions amongst HSE cases.

Methods

HSE hospitalizations and 60-day rehospitalizations were assessed in a retrospective cohort using linked hospitalizations from the Healthcare Utilization Project (HCUP) National Readmission Database (NRD) from 2010 through 2017. Risk factors for all-cause readmissions and encephalopathy associated readmissions were assessed with a weighted logistic regression model.

Results

There were 10 272 HSE cases in the US between 2010 and 2017, resulting in a national rate of 4.95 per 100 000 hospitalizations. A total of 23.7% were readmitted at least once within 60-days. Patients that were readmitted were older (mean age 62.4 vs 57.9, P < .001), had a greater number of procedures at the index hospitalization (adjusted odds ratio [aOR] 1.03, P < .001) and have a higher Charlson comorbidity score (aOR 1.11, P < .001). Among those readmitted, 465 (16.5%) had an encephalopathy related diagnosis. Over 8 years, the rate of encephalopathy associated readmissions increased from 0.12 to 0.20. Encephalopathy specific readmissions were found to be associated with greater age (mean age 65.9 vs 61.7, P = .004) and findings of cerebral edema at index hospitalization (aOR 2.16, P < .001).

Conclusions

HSE readmissions are relatively common, particularly among older and sicker individuals. However, early signs and symptoms of neurological disease at index were correlated with encephalopathic specific readmissions.

Readmissions occurred in approximately 1 of every 4 HSE cases that survived and were often associated with an encephalopathic diagnosis at readmission (12%–20%). Greater age, clinical complexity, and neurological symptoms at index admission were key risk factors for readmission.

Encephalitis remains one of the most challenging medical conditions to diagnose and treat, largely in part to the numerous possible etiologies associated with the disease and the unique pathophysiology they each possess [1]. However, herpes simplex virus remains the most common known cause of encephalitis and accounts for 30% to 40% of all cases [2, 3]. Herpes simplex encephalitis (HSE) is associated with significant disease burden requiring intensive care in over half of all cases [4]. Prior to antiviral therapy, mortality associated with HSE was reported as high as 70% [5, 6]. Over the past several decades, mortality outcomes have improved drastically; however, this has also led to an increasing prevalence of disease morbidity, long-term neurological symptoms, and even full relapse of encephalopathic disease requiring repeat hospitalizations [7, 8].

Although this disease is challenging to treat in the hospital initially, its ongoing sequelae after discharge remain an important consideration to improve treatment and prevent further disease burden. HSE related readmissions have been gaining increasing attention in the literature, particularly between HSE and recurrent encephalopathy [8]. This can include recurrent HSE, new onset encephalitis from a different etiology, and cases with no known etiology. With better diagnostic tools and increased recognition, recurrent encephalopathy has become more apparent and concerning. To date, this has been largely described through case studies and case series [9–13]. With a relatively rare disease such as HSE, studies using administrative data sets are highly effective ways to better understand HSE outcomes and their associated risk factors. The objective of our study was to use a national administrative dataset to describe HSE hospitalizations and rehospitalizations. Our study had 2 aims: first to describe all-cause 60-day readmission rates and associated risk-factors for HSE rehospitalizations for any medical reason, and second, to characterize and identify risk factors associated with encephalopathy specific rehospitalizations amongst all HSE cases.

METHODS

Study Design

Our study was designed as a retrospective cohort of a longitudinal national administrative dataset from the Healthcare Utilization Project (HCUP) known as the Nationwide Readmissions Database (NRD) from 2010 to 2017. The NRD is a publicly available database that consists of hospitalizations and rehospitalizations within each year that are linked via a specific patient identifier. In 2010 there were 18 states included in the database and gradually increased to 26 states by 2017. The database is constructed as a weighted survey design allowing stratification for national estimates [14]. No patient identifiable information is included in the HCUP data sets; nevertheless, the project was submitted and deemed exempt from Baylor College of Medicine’s Institutional Review Board. HCUP data access was approved to the authors by the Agency for Healthcare Research and Quality (AHRQ) data use agreement.

In order to make valid comparisons we followed the recommended data management protocol outlined by HCUP [14]. All baseline hospitalizations were restricted to those associated with patients greater than or equal to 1 year of age and occurring between months of January and October to allow accurate capture of 60 day readmissions (cases do not cross over each calendar year). According to Armangue et al, recurrent neurological symptoms and relevant immunological markers have been shown to occur up to 5 weeks after HSE hospitalizations [15]. Hence, it was decided to use a 60-day readmission window. Also, the coding mechanisms used in the HCUP data sets were updated in October of 2015. This required 2015 to be divided as two separate data sets and led to the exclusion of August and September from the first 2015 data set and the removal of November and December in the second 2015 data set. From this cohort, HSE cases were identified by using International Classification of Disease, ninth (ICD-9) and tenth edition codes (ICD-10), that were listed as the primary diagnosis (Figure 1). Of these cases, those that did not die during hospitalization were flagged as “index cases” and were matched with subsequent hospitalizations if they occurred within 60 days. It is important to note that the analysis and associated results are based on these specific index discharges as units of measurement and not individual patients. Because some individuals had multiple visits with a primary diagnosis of HSE, there is a slightly higher number of index cases than unique individuals.

Figure 1.

Number of unweighted and weighted cases. Abbreviations: HSE, herpes simplex encephalitis; NRD HCUP, Nationwide Readmissions Database—Healthcare Cost and Utilization Project.

Additional diagnoses, symptoms, and procedures were similarly identified using ICD clinical and procedural codes as described below. Preestablished variables provided by the NRD were used for patient specific characteristics (ie, gender, age, payor, length of stay, residence status, time of admission, disposition at discharge, and associated zip-code income quartile) and hospital level characteristics (ie, hospital location, size, and teaching status).

Definitions

HSE cases were defined by searching the primary listed diagnoses for the appropriate ICD-9 (“054.3”) or ICD-10 (“B004” and “B100*”) diagnosis codes. Cases were restricted to the primary diagnostic code only to help improve accuracy of valid HSE cases. All listed diagnostic codes were used to assess patient level disorders and calculate the Charlson Comorbidity Index value or specific flags for symptoms known to be associated with poor HSE outcomes that were chosen a priori (ie, fever, aphasia, dysphagia, abnormal gait, seizures, cerebral edema, thrombocytopenia, leukocytosis, and leukocytopenia) [2, 7]. Similarly, ICD-9 and ICD-10 procedure codes were used to identify documentation of intubation and tracheotomy. Diagnoses at readmission encounters were collected and summarized into clinically related categories. Associated codes for all diagnoses, symptoms, and procedures are listed in Appendix A.

Statistical Analysis

Unless specified, numeric count values were reported as their weighted values and column percents. Composite values of continuous variables were reported as means and associated standard deviations. Dichotomized comparisons were evaluated for statistically significant differences using χ ², ANOVA, or Student t test where appropriate. A weighted logistic regression model was used to derive the unadjusted and adjusted odds ratios associated with each risk factor and their correlation with readmission amongst all HSE cases, as well as encephalopathy specific readmissions among HSE index cases. A multinomial model was used to compare all risk factors that had a P-value < .2 in the unadjusted models for each comparison (ie, all readmissions and encephalopathy specific readmissions). Analysis was performed using R statistical software (version 4.0.3, “stats” package) [16, 17], and R studio environment [18].

RESULTS

HSE Hospitalizations

From 2010 through 2017, there were 104 897 595 hospitalizations in the NRD data set that were assessed and filtered for eligibility (≥1 year of age and occurring within the appropriate month ranges) for a final total of 95 068 178 eligible cases representing a weighted total of 207 468 667 hospitalizations within the United States (Table 1). Of these hospitalizations, 4577 were identified with a primary diagnosis of HSE (10 272 weighted cases, rate of 4.95 per 100 000 hospitalizations). HSE specific hospitalizations varied between ages groups (18–44 = 2.73 and 45–64 = 6.23, per 100 000 in their respective age groups), had higher rates of mortality, and transfer to outside facilities or home health as compared to all hospitalizations (death = 17.81 and transfer/home health = 10.06, per 100 000). Other demographic features remained relatively consistent with rates of all other US hospitalizations.

Table 1.

Demographic Values and Rates for All Herpes Simplex Encephalitis (HSE) (Primary Diagnosis Only) Hospitalizations in the United States per Year

	All Hospitalizationsa			HSE Primary Diagnosis b			Rate of HSE Cases per 100 000
	Total n	Weighted n	%	HSE n	Weighted n	%
	95 068 178	207 468 667	100	4577	10 272	100	4.95
Gender
Male	39374914	85 922 943	41.4	2230	5032	49.0	5.86
Female	55 693 264	121 545 724	58.6	2347	5242	51.0	4.31
Age (continuous)
Age, mean	56.0	…	…	59.9	…	…	…
Age (categorical)
1–17	3 034 721	9 174 432	4.4	139	421	4.1	4.59
18–44	26 462 071	57 808 211	27.9	736	1576	15.3	2.73
45–64	26 777 643	57 777 828	27.8	1569	3598	35.0	6.23
≥65	38 793 743	82 708 199	39.9	2133	4677	45.5	5.65
Primary payorc
Medicare/Medicaid	61 289 069	132 303 789	63.8	2741	6034	58.7	4.56
Private	26 105 284	58 053 657	28.0	1453	3369	32.8	5.80
Uninsured/other	76 58 215	17 081 738	8.2	383	871	8.5	5.10
Discharge dispositiond
Home or self care	64 324 591	141 658 473	68.3	1572	3488	34.0	2.46
Transfer, home health	27 215 115	58 469 192	28.2	2616	5883	57.3	10.06
Against medical advice	1 318 245	2 652 412	1.3	45	85	0.8	3.20
Other	40 752	74 241	0.0	4	7	0.1	9.43
Death	2 120 875	4 513 341	2.2	337	804	7.8	17.81
Hospitalization
Length of stay, mean	4.9	…	…	13.5	…	…	…

^aHospitalizations and HSE cases exclude the last 60 days of each data set and are ≥1 year of age.

^bHSE as primary diagnosis indicated by the first listed diagnosis, either “543” for International Classification Diagnosis Code, ninth edition or “B004,” “B100*” for the tenth edition. HSE cases in this table are not the same as index cases.

^cPayor groups: Medicare/Medicaid = includes both fee-for-service and managed care patients; private = includes commercial carriers and private health maintenance organizations (HMOs)/preferred provider organizations (PPOs); uninsured/other = includes worker’s compensation, CHAMPUS, CHAMPVA, Title V, and other government programs.

^dTransfer includes short-term hospitals, skilled nursing facility, or intermediate care facility.

Of the 10 272 weighted cases, there were 804 inpatient deaths (7.8%) and 9468 survivors (92.2%) at discharge. Among those that survived, several had multiple HSE admissions or multiple readmissions, resulting in a total number of 10 009 index hospitalizations (Table 2). Index hospitalizations were on average 59 years of age, had a slight female predominance (51.3%), and were more likely discharged to an outside facility or requiring home health (62.8%). Index hospitalizations lasted on average 13.2 days (SD = 15.6) with an average of 13 diagnoses and 3 procedures per case. Individuals were found to have a mean Charlson Comorbidity Index score of 1.12 (SD = 1.2) and most commonly presented with fever (32%), seizures (27%), and aphasia (12.5%). Intubation was recorded in 15.8% of index cases, and a tracheotomy was performed in 13.6% of index cases. Nearly all tracheotomy cases had an intubation (99.2%), although only 85% of intubations were associated with tracheotomies.

Table 2.

Comparison of Demographic and Clinical Characteristics Between Index Cases With No Readmissions and Cases Readmitted Within 60 Days

	HSE Cases Survive (Index Cases)a		HSE at Index With No Readmits		HSE at Index With 60-day Readmission
	nb	%	nb	%	nb	%	P-value
All hospitalizations	10 009	(100)	7206	(100)	2815	(100)	…
Gender
Male	4877	(48.7)	3504	(48.7)	1373	(48.9)	.5
Female	5132	(51.3)	3699	(51.4)	1434	(51.1)
Age (continuous)
Age, mean (SD)	59	(19.23)	57.95	(19.6)	62.41	(17.1)	<.001
Age (categorical)
1–17	417	(4.2)	351	(4.9)	66	(2.4)	<.001
18–44	1621	(16.2)	1310	(18.2)	310	(11.1)
45–64	3650	(36.5)	2609	(36.2)	1041	(37.1)
≥65	4322	(43.2)	2933	(40.7)	1389	(49.5)
Primary payorc
Medicare/Medicaid	5782	(57.8)	3957	(55)	1824	(65)	<.001
Private	3382	(33.8)	2547	(35.3)	835	(29.7)
Uninsured/other	846	(8.5)	698	(9.7)	148	(5.3)
Discharge dispositiond
Discharge home or self care	3625	(36.2)	2887	(40.1)	738	(26.2)	<.001
Transfer, home health care	6286	(62.8)	4243	(58.9)	2043	(72.9)
Against medical advice	91	(0.9)	65	(0.9)	26	(0.9)
Other	7	(0.1)	7	(0.1)	0	(0)
Death	0	(0)	0	(0)	0	(0)
Zip code income quartilee
0–25th percentile	2513	(25.5)	1723	(24.2)	789	(28.6)	.28
26–50th percentile	2539	(25.7)	1787	(25.1)	752	(27.24)
51–75th percentile	2510	(25.4)	1876	(26.4)	634	(22.9)
76–100th percentile	2305	(23.4)	1720	(24.2)	586	(21.2)
Clinical features at index, mean (SD)
Length of stay	13.24	15.6	12.72	14.9	15.41	17.7	<.001
Number of diagnoses	13.66	7.5	13.19	7.4	15.46	7.4	<.001
Number of procedures	3.02	2.9	2.91	2.7	3.45	3.3	<.001
Charlson comorbidity score	1.12	1.2	1.04	1.2	1.44	1.3	<.001
Procedures at index
Intubated	1577	(15.8)	1113	(15.4)	464	(16.5)	.25
Trach placement	1359	(13.6)	970	(13.5)	388	(13.8)	.61
Signs and symptoms at index
Fever	3202	(32)	2288	(31.7)	914	(32.5)	.79
Seizures	2702	(27)	1884	(26.1)	818	(29.1)	.14
Aphasia	1250	(12.5)	921	(12.8)	329	(11.7)	.19
Cerebral edema	906	(9.1)	606	(8.4)	300	(10.6)	.16
Dysphagia	804	(8)	509	(7.1)	295	(10.5)	.001
Leukocytosis	558	(5.6)	390	(5.4)	169	(6.0)	1.00
Thrombocytopenia	463	(4.6)	318	(4.4)	145	(5.1)	.23
Abnormal gait	449	(4.5)	338	(4.7)	111	(3.9)	.27
Leukocytopenia	73	(0.7)	45	(0.6)	27	(1.0)	.20

Abbreviations: HSE, herpes simplex encephalitis; SD, standard deviation.

^aIndex cases only included hospitalizations with primary diagnosis of HSE; hospitalization did not occur in the last 60 days of the data set, ≥1 year of age, and did not die during hospitalization.

^bValues are surveyed weighted counts.

^cPayor groups: Medicare/Medicaid = includes both fee-for-service and managed care patients; private = includes commercial carriers and private health maintenance organizations (HMOs)/preferred provider organizations (PPOs); uninsured/other = includes worker’s compensation, CHAMPUS, CHAMPVA, Title V, and other government programs.

^dTransfer includes short-term hospitals, skilled nursing facility, or intermediate care facility.

^eBased on estimated median household income of residents in the patient’s ZIP code; approximate dollar ranges: 0–25% = 1–40 000; 26–50% = 41 000–50 000; 51–75% = 51 000–65 000; 76–100% >65 000.

All Cause 60-Day Readmissions of HSE Index Hospitalizations

Of those that survived, 28.1% would be readmitted at least once within 60 days. Compared to HSE cases with no readmissions (Table 2), those readmitted within 60 days were slightly older (mean of 62.4 vs 57.9, P < .001), had a higher proportion insured by Medicare/Medicaid (65% vs 55%, P < .001), and were more likely to be transferred or discharged with home health at index (72.9% vs 58.9%, P < .001). Index hospitalizations for those readmitted were found to have a longer length of stay (15.4 vs 12.7 days), higher number of diagnoses (15.5 vs 13.2), higher number of procedures (3.5 vs 2.9), and higher Charlson comorbidity index score (1.4 vs 1.0) (P < .001, all). Clinically, only dysphagia symptoms were more common at index amongst readmitted cases (10.5% vs 7.1%, P = .001). Logistic modeling showed that after adjustment, age groups above 18–44 were all associated with an increased odds of readmission (ages 45–64: adjusted odds ratio [aOR] 1.28, P < .001; ages ≥65: aOR 1.18, P = .04) (Table 3). Disposition at discharge also played a significant role in predicting readmission where those who were documented as leaving against medical advice had an aOR of 1.58 (P = .05) and those transferred or discharged with home health had an aOR of 1.35 (P < .001). The 2 highest income quartiles were associated with decreased odds of readmission compared to the lowest quartile (aOR 0.81, P < .001, each). Clinically, a higher number of procedures, a higher Charlson comorbidity score, and cerebral edema at index were all found to increase odds of readmission (aOR 1.03, P = .004; aOR 1.11, P < .001; and aOR 1.15, P = .05, respectively). Using all clinical diagnosis codes listed during the visit, the 4 most common diagnoses at readmission were found to be renal disease, pulmonary disease, encephalopathy, and diabetes (Figure 2). Over the course of 8 years, the rate of encephalopathy associated readmissions increased from a proportion of 0.12 to 0.20 (mean of 16.5%).

Table 3.

Weighted Logistic Regression Models for All-Cause 60-day Readmissions and Those Specifically With Encephalopathy Diagnoses

	All-cause 60-day Readmits Among All HSE Indexa Hospitalizations		Encephalopathyb Specific 60-day Readmits Among All HSE Index Hospitalizations
	OR (95% CI) P-value	aOR (95% CI) P-value	OR (95% CI) P-value	aOR (95% CI) P-value
Gender
Male	Reference		Reference
Female	0.99 (.91–1.06) .85		1.05 (.86–1.28) .62
Age (categorical)
1–17	0.81 (0.61–1.08) .16	0.80 (0.60–1.07) .14	0.69 (.21–1.78) .48	0.79 (.24–2.08) .66
18–44	Reference	Reference	Reference	Reference
45–64	1.52 (1.33–1.75) <.001	1.28 (1.11–1.49) <.001	1.85 (1.24–2.86) .004	1.80 (1.19–2.81) .007
≥65	1.71 (1.49–1.96) <.001	1.18 (1.01–1.39) .04	2.26 (1.53–3.45) <.001	1.86 (1.23–2.89) .004
Primary payorc
Medicare/Medicaid	1.30 (.24–.28) <.001	1.10 (.98–1.23) .11	0.87 (.70–1.08) .20
Private	Reference	Reference	Reference
Uninsured/other	0.69 (.57–.84) <.001	0.72 (.58–.87) .001	0.66 (.38–1.08) .12
Discharge dispositiond
Home or self care	Reference	Reference	Reference	Reference
Transfer, home health care	1.64 (1.49–1.80) <.001	1.35 (1.21–1.49) <.001	2.2 (1.39–2.88) <.001	1.98 (1.49–2.66) <.001
Against medical advice	1.42 (0.89–2.18) 0.13	1.58 (0.98–2.16) .05	NA	NA
Zip income quartilee
0–25th percentile	Reference	Reference	Reference	Reference
26–50th percentile	0.93 (.86–1.05) .23	0.97 (.86–1.09) .61	1.38 (1.06–1.81) .02	1.47 (1.11–1.95) .007
51–75th percentile	0.78 (0.70–0.88) <.001	0.81 (0.72–0.92) <.001	1.22 (0.91–1.62) .18	1.32 (.98–1.77) .06
76–100th percentile	0.79 (0.70–0.89) <.001	0.81 (.71–.92) .001	1.16 (.86–1.56) .32	1.11 (.82–1.51) .50
Clinical features at index
Length of stay	1.01 (1.00–1.02) <.001	1.00 (.99–1.01) .27	1.00 (.99–1.01) .09
Number of diagnoses	1.03 (1.03–1.04) <.001	1.01 (.99–1.01) .19	1.02 (1.00–1.03) .02	0.99 (.98–1.01) .52
Number of procedures	1.06 (1.04–1.07) <.001	1.03 (1.01–1.05) .004	1.03 (1.00–1.06) .09
Charlson comorbidity score	1.21 (1.17–1.25) <.001	1.11 (1.07–1.16) <.001	1.04 (0.96–1.13) .30
Procedures at index
Intubated	1.05 (.93–1.17) .42		1.12 (.86–1.45) .39
Trach placement	1.01 (.89–1.13) .94		0.98 (.73–1.31) .91
Signs and symptoms at index
Abnormal gait	0.87 (.70–1.07) .18		0.85 (.48–1.41) .55
Aphasia	0.93 (0.82–1.06) 0.30		1.21 (.90–1.62) .20
Cerebral edema	1.21 (1.05–1.39) 0.007	1.15 (0.99–1.33) 0.05	2.16 (1.64–2.84) <.001	2.16 (1.61–2.88) <.001
Seizures	1.13 (1.03–1.24) .01	1.10 (0.99–1.21) .06	1.09 (.88–1.36) .42
Dysphagia	1.41 (1.22–1.63) <.001	1.12 (0.96–1.30) .15	1.19 (.87–1.61) .27
Leukocytosis	1.10 (.92–1.32) .29		1.19 (.79–1.75) .39
Leukocytopenia	1.36 (.86–2.11) .17		1.53 (.57–3.55) .35
Fever	1.03 (.94–1.12) .54		1.19 (.97–1.47) .10
Thrombocytopenia	1.11 (.92–1.33) .27		1.03 (.66–1.56) .88

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; HSE, herpes simplex encephalitis; NA, not applicable; OR, odds ratio.

^aIndex cases only included hospitalizations with primary diagnosis of HSE; hospitalization did not occur in the last 60 days of the data set, >1 year of age, and did not die during hospitalization.

^bEncephalopathy encompasses “encephalopathy nos.” and “encephalopathy other” via International Classification Diagnosis Code, ninth edition (34830, 34839) and tenth edition (G9340, G9341).

^cPayor groups: Medicare/Medicaid = includes both fee-for-service and managed care patients; private = includes commercial carriers and private health maintenance organizations (HMOs)/preferred provider organizations (PPOs); uninsured/other = includes worker’s compensation, CHAMPUS, CHAMPVA, Title V, and other government programs.

^dTransfer includes short-term hospitals, skilled nursing facility, or intermediate care facility.

^eBased on estimated median household income of residents in the patient’s ZIP code; approximate dollar ranges: 0–25% = 1–40 000; 26–50% = 41 000–50 000; 51–75% = 51 000–65 000; 76–100% >65 000.

Figure 2.

Four most common diagnosis groups listed at the 60-day readmission. Among all cases that were readmitted, the most common diagnosis groups at readmission were diabetes, pulmonary disease, renal disease, and encephalopathy diagnosis codes. ICD 9 codes associated with each diagnosis group are found in Appendix A.

Encephalopathy Specific 60-Day Readmissions of HSE Index Hospitalizations

Out of all readmissions, 465 (20%) were found to have an associated encephalopathic diagnosis (Appendix B). There were no statistically significant differences of index cases with encephalopathic readmissions and all cause readmissions in regard to gender, insurance status, or clinical features. However, encephalopathic readmits were more likely to be older (mean of 65.9 vs 61.7, P < .001), at the 26–50th zip income percentile (30.9% vs 26.5%, P = .03), clinically present with cerebral edema at index (18% vs 9.2%, P = .03), and be transferred or discharged with home health care at index (84.4% vs 70.5%, P < .001). In the final adjusted model, age groups above 18–44 were all associated with an increased odds of encephalopathy at readmission (ages 45–64: aOR 1.80, P = .007; ages ≥65: aOR 1.86, P = .004) (Table 3). Transfer and home health at discharge were also found associated with encephalopathy (aOR 1.98, P < .001), however, only the 26–50th zip income percentiles were statistically more likely to have encephalopathy symptoms at readmission (aOR 1.47, P = .007). Finally, cerebral edema symptoms at index hospitalization were found to be correlated with encephalopathy at readmission with an aOR of 2.16 (P < .001).

DISCUSSION

HSE hospitalizations were found to impact older individuals, lead to much longer hospitalizations, and result in higher mortality and morbidity as compared to the average of all hospitalizations in the United States. We found that 7.8% of all HSE hospitalizations lead to inpatient death. Of those that survived, over 1 in 4 were readmitted within 60 days of discharge. Furthermore, 4.6% of HSE survival cases were readmitted with an associated encephalopathy diagnosis (16.5% of all HSE readmissions between 2010 and 2017). Encephalopathy-related and all-cause readmissions were both correlated with greater age, longer hospitalizations, and more clinically complex patients. Encephalopathy specific readmissions were on average 4.5 years older, more likely to have been originally discharged to an outside facility/home health, and had evidence of cerebral edema at index hospitalization as compared to other HSE all-cause readmissions. Zip-code associated income levels had a contradictory correlation, where higher zip-code quartiles were associated with a lower likelihood of readmissions overall but had increased likelihood of encephalopathy specific readmissions. Over 8 years, encephalopathy-associated readmissions have steadily increased and is one of the most common diagnoses associated with HSE readmissions, trailing only behind diabetes, renal and pulmonary associated diagnosis codes. Overall, these results provide new core knowledge to the clinical course and natural history of HSE. They also reaffirm and clarify the suspected correlations between HSE and recurrent encephalopathy.

The rate of HSE cases occurred in 4.9 out of every 100 000 hospitalizations. This notably excludes all cases in children <1 year of age. This rate appears to be slightly more than expected from prior research using the HCUP data set where the rate of all cause encephalitis was found to be approximately 7.3 per 100 000 hospitalizations [2]. Mortality was significantly higher amongst HSE cases compared to all hospitalizations (7.8% vs 2.2%). This was reflective of results summarized by Modi et al, where inpatient mortality rates in other HSE studies had a range between 4% and 11% [18]. The high rate of morbidity in HSE was also represented with a much longer average length of stay (13.5 days) and over half (57.3%) of all HSE cases being transferred to alternative care facilities or being discharged with home health. Additionally, readmissions can be considered a form of morbidity and an indication of disease severity, representing a significant burden on the patients, as well as the healthcare system [19]. Thus, understanding and preventing readmissions has begun to take a vital role in the management plan of serious illnesses. In HSE we found that 28.1% were readmitted within 60 days. For comparison, rate of all cause 60-day readmission for heart failure is approximately 33.3%, 16.6% for acute exacerbation of chronic obstructive pulmonary disease, and 16.3% for stroke related hospitalizations [20–22]. This shows that HSE is on the higher end of the readmission spectrum and is an important consideration in the care of patients afflicted with the infection. It appears that older individuals and those with more complicated index hospitalizations are at highest risk. Both of these factors have been shown to be associate with all cause readmission among various other diseases and does not appear to be unique to encephalitis [23–25]. Alternatively, those from more affluent zip codes were less likely to have a readmission. It is reasonable to speculate that it could be a reflection of the population’s overall health or a feature related to their access to outpatient medical care allowing them to avoid preventable hospitalizations. However, we also found that those without insurance or traditional payor mechanisms had a lower risk of readmission as well, which conflicts with the latter hypothesis. Overall, it is difficult to say if these protective factors are associated with individuals avoiding readmission due to healthcare access, cost aversion, or some other factor we were not able to account for.

Further exploration of each readmission hospitalization showed a pattern of specific diagnoses. Diagnosis codes consistent with neurological disease were quite common. Collectively they were associated with 1 of every 6 readmissions. This observation has also been reflected in other studies. HSE has been correlated with recurrent encephalopathy, seizures, and most recently, autoimmune encephalitis [26–30]. Although these outcomes are relatively rare, their clinical significance is immense. Early awareness and rapid treatment have always been a part of encephalitis management and these readmission cases are no different. Hence, the frequency and unique recurrence of these neurologic symptoms after HSE have become incorporated into clinical management of several guidelines [31, 32]. Although it is difficult to identify the exact etiology for all post-HSE encephalopathy cases, the autoimmune mechanism has become a leading area of interest, specifically in regards to antibodies against N-methyl-D-aspartate receptors (NMDAR) [33, 34]. However, evidence is still premature, and most information has been derived from case studies and case series [35–38]. We believe that our study is one of the first to explore and describe this correlation on a population level. Unfortunately, with the nature of clinical coding, particularly with ICD-9, we were unable to specifically identify NMDAR diagnoses at readmission. However, we were able to capture relevant symptoms that are likely reflective of this unique population as a whole. The overall characteristics of the encephalopathy specific readmission group show that it occurs in all ages but is generally more common in older individuals. These cases were also more likely to have cerebral edema at their index HSE hospitalization. Also, it was found to be more common among individuals within the middle zip code income quartiles but had no statistically significant correlation with insurance or payor type.

Although HSE may be a relatively rare condition, the large sample size provided by the NRD HCUP data set has allowed for a detailed accounting of the disease and associated outcomes. The reliable linkage of cases is also a unique and powerful feature of the data that allows us to explore readmission of HSE cases and obtain a high-level perspective of the long-term consequences of this complex disease. Administrative data sets are particularly dependent upon diagnostic codes for accurate capture of a disease population. For HSE, ICD codes have been shown to have a moderate positive predictive value but very high sensitivity [39]. We were acutely aware of the potential for misclassification bias because of this and attempted to mitigate the risk by limiting our inclusion criteria of HSE cases to only the primary diagnosis listed during the index encounter. Similarly, many descriptive factors used to describe clinical features at index and follow up hospitalizations were dependent upon ICD codes. Although their accuracy is less of a concern, there is little information about the reliability of their documentation during encounters and potentially introduces additional information bias. As with most studies that involve administrative data sets, more investigation is needed but the context and scope of the disease is significantly clearer. Many of our results are consistent with prior research and elucidates new knowledge that can assist with pushing HSE-related research forward.

CONCLUSION

HSE is a rare but serious disorder with a high rate of rehospitalization. Many of these readmissions are associated with seizures and encephalopathy related diagnosis codes. These results demonstrate long-term disease morbidity of HSE and recurrent, or possibly induced, encephalopathic disease.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgment

Financial support. This research was supported by the Health Resources and Services Administration, an agency of the US Department of Health and Human Services (grant number T32 HP10031) (Michael A. Hansen’s postdoctoral funding). The funding source/study sponsor had no role in the writing of this manuscript, or decision to submit the manuscript for publication.

Potential conflicts of Interest. R. H. has received research support and fees from Biofire®. M. A. H. reports no potential conflicts. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.