Secular Trends in the Incidence of Encephalitis and Outcomes of Postencephalitic Epilepsy in Taiwan: A Nationwide Population-Based Study

There are diverse clinical manifestations and numerous potential etiologies of acute encephalitis pose significant diagnostic and management challenges, often resulting in severe illness requiring hospitalization. Over the past decade, substantial progress has been made in identifying infectious and autoimmune causes of encephalitis [1‒4]. However, cases with unknown etiologies still account for 36–50% of all encephalitis cases [2, 5‒9]. The clinical presentation of encephalitis may include fever, headache, cognitive decline, psychobehavioral changes, and seizures, with potential outcomes ranging from mortality to long-term morbidity, including epilepsy and permanent neurological disability [6]. Encephalitis affects individuals across all age groups, with the highest incidence observed in children and older adults [10, 11]. The reported incidence rates are 10.3–13.8 per 100,000 per year in the pediatric population, 2.2 per 100,000 per year in adults, and 6.34–8.66 per 100,000 per year across all age groups [12, 13]. In the USA, the estimated incidence of encephalitis-related hospitalizations is 7.3 per 100,000 population [11]. In Italy, the average annual rate of encephalitis hospitalization from 1999 to 2005 was 5.88 per 100,000 population [5]. In Korea, the annual incidence of encephalitis in 2015 was 7.5 per 100,000 person-years [14]. Despite advances in diagnosis and treatment, encephalitis continues to result in mortality rates of 5.6–39.3%, as reported in various countries [2, 13, 15‒17]. In England, the mortality rate from viral encephalitis is 7% [6]. In one center in Canada, death caused by acute encephalitis occurred in 9.5% of patients with viral encephalitis, 9.1% of those with autoimmune encephalitis, and 8.5% of those with encephalitis of other/unknown etiologies [17]. Furthermore, the incidence of postencephalitic epilepsy (PE) is highest during the first 5 years following the initial illness but remains elevated for up to 15 years [18]. Encephalitis also contributes significantly to disability-adjusted life years [19] and negatively impacts cognitive outcomes [20].

In 2013, the International Encephalitis Consortium established simplified diagnostic criteria for encephalitis and encephalopathy of presumed infectious or autoimmune origin [21]. According to these criteria, a diagnosis of encephalitis requires an altered mental status that persists for over 24 h without an alternative cause, along with additional minor criteria. Confirmed cases of encephalitis require hospitalization [6, 12].

In this study, we estimated the incidence rate and secular trend of encephalitis in Taiwan, using data from the linkage between the National Health Insurance (NHI) Database and the National Death Registry. Additionally, we examined outcomes of epilepsy in cases of encephalitis in Taiwan.

The Institutional Review Board of Chang Gung Memorial Hospital approved this study (Approval No. CORPG3G0481). The Ministry of Health and Welfare anonymized all data used in this study, thereby waiving the need for patient consent.

The primary data source for this study was Taiwan’s NHI system. We identified individuals registered in Taiwan’s NHI between January 1, 1998, and December 31, 2017. The NHI system, established in 1995, mandates participation for all residents of Taiwan, with current enrollment exceeding 99% of the population. The total eligible population in our study consisted of 23,890,640 beneficiaries (50.02% men) registered in the NHIRD in 2017. The NHI Research Database (NHIRD) includes comprehensive registration information and original claims for reimbursement, encompassing data on demographics, clinical visit dates, medical diagnoses, expenditures, prescriptions, examinations, and procedures. Data associated with outpatient and inpatient events are summarized in two distinct records: one detailing expenditure, including the medical professional involved, diagnosis, surgery, and specific fees for each procedure; and another summarizing all prescriptions, tests, examinations, and procedures related to each event. Each individual’s records within the NHI system are linked via a unique personal identification number, ensuring continuity and traceability. The validity, representativeness, and clinical consistency of this database have previously been documented [22]. The denominator data are derived from the Registry of Beneficiaries, a component of the NHI database that contains records of all beneficiaries’ demographics, insurance status, residence, and socioeconomic data. Since January 1, 2016, the diagnostic coding system in the NHI database has adhered to the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM). For inpatient diagnosis in Taiwan, trained and certified coding specialists assign corresponding ICD diagnosis codes based on patients’ medical records, so it is relatively accurate.

To ensure confidentiality, identification numbers were de-identified before being released for research purposes. However, these identifiers remain unique for valid internal linkages and, with assistance from the Ministry of Health and Welfare, for external linkages to other government-held data, including death records.

Incident encephalitis cases were defined as those with an encephalitis diagnosis corresponding to any of the following ICD-9-CM codes: 323.0–323.9, 036.1, 046.2, 046.3, 049.9, 054.3, 055.0, 072.2, or 062.0–064.9, and the following ICD-10-CM codes: A86, A879, A89, G039, G040, G048, or G049 requiring hospitalization [23]. A proportional analysis of etiology in encephalitis (ICD codes are listed in online supplementary Table S1; for all online suppl. material, see https://doi.org/10.1159/000543853) was performed in which the etiology of encephalitis was stratified as bacterial, viral, acute disseminated encephalomyelitis (ADEM), and others (including autoimmune conditions) by patient age and by year.

The trends in incidence were analyzed by computing the age- and sex-standardized incidence rates for each calendar year from 2003 to 2017. The population structure of 2017 served as the reference.

Incident cases of encephalitis were defined as individuals with no prior history of encephalitis as recorded in their hospitalization records during the 3 years before January 1 of each calendar year, but with a recorded diagnosis of encephalitis during that year. Individuals with at least a 1-year registration period before January 1 of each calendar year were included. For these incident cases, the at-risk cohort was defined as individuals without a history of encephalitis for that year, including all individuals registered with NHI during that period. Follow-up was extended from January 1 of each year to the earliest date of incident encephalitis, death, or December 31 of the specified year. Crude and standardized incidence rates of encephalitis were estimated per 100,000 person-years for each calendar year. Incidence was calculated by dividing the number of incident cases in a calendar year by the total person-years in the at-risk population during that same year. Subgroup analyses were conducted to estimate incidence by sex and age.

For encephalitis, a Charlson Comorbidity Index (CCI) score > 2 was considered a risk factor predicting inpatient mortality [24]. The CCI measures individual health status and comorbidity, controlling for confounding factors in its prediction of mortality [25]. While not without limitations, it is a suitable tool for assessing medical costs in pediatric patients [26] and for predicting mortality, excluding patients under 1 year of age [27]. The CCI was developed in medical centers in 1987 and includes 17 categories of comorbid diseases, weighted based on the initial cohort of hospitalized patients who died within 1 year of follow-up [28] (online suppl. Table S2). The CCI score is calculated as the resultant weighted sum, representing individual health status [29].

Given the complexity of encephalitis presentation and disease course, many patients may exhibit various neuropsychiatric behaviors prior to hospitalization. Selected neuropsychiatric diseases were identified using ICD-9 and ICD-10 codes (online suppl. Table S3) before diagnosis. Preexisting conditions were defined based on physician diagnosis, requiring identification from at least two outpatient clinical records separated by at least 30 days, or at least one hospitalization discharge diagnosis.

The cumulative probability of a 5-year survival rate by age at encephalitis diagnosis was estimated using Kaplan-Meier curves to plot survival based on all-cause mortality.

As PE is a significant comorbidity that requires ongoing anti-seizure medication, a diagnosis of epilepsy can serve as a proxy for a major comorbidity in encephalitis patients. We followed patients diagnosed using ICD-9 code 345 and ICD-10 code G40, requiring at least three outpatient records or at least one hospitalization record [30], to estimate the incidence of newly diagnosed comorbid epilepsy after incident encephalitis. The cumulative probability of 5-year comorbid PE was calculated.

The Joinpoint Regression Program version 4.2.0.2 (National Cancer Institute, Bethesda, MD, USA) [31] was used to estimate temporal trends in crude and standardized incidence rates. The Bayesian information criterion was employed to generate different numbers of “joinpoints” over time, reflecting significant changes in the prevalence of encephalitis and incidence trends, to best fit the data. Annual percentage changes (APCs) and 95% CIs for each segment were calculated based on the assumption of a Poisson distribution for the observed numbers of incident cases. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Between January 1, 1998, and December 31, 2017, 39,798 individuals were admitted to acute inpatient hospital care in Taiwan with a diagnosis of encephalitis. The mean age of these patients was 35.97 years (standard deviation: 27.64 years), with a slight male predominance (55.6%). Stratification by age revealed that the highest incidence was observed in children under 10 years, accounting for 11,460 cases (28.8%). In this age group, the highest number of incident cases was in children 1–4 years of age (n = 5,938, 14.92% of all encephalitis cases), followed by those aged 5–10 years (n = 4,066, 10.72%). The second-highest incident age group was 50–59 years, with 4,435 cases (11.14%) (Table 1).

Approximately 55% of the individuals resided in urban areas, with the majority of cases diagnosed in northern Taiwan (43.59%), likely reflecting the region’s higher population density. There were no significant differences in income levels among the cases. Notably, 29.40% of the incident encephalitis patients had a CCI score >2 prior to diagnosis. The most prevalent preexisting neuropsychiatric comorbidities included mood disorders (18.08%), sleep disorders (15.86%), epilepsy (12.46%), anxiety (10.97%), and depression (6.88%).

The diagnosis was most frequently made in June and January. Specifically, June (the transition from spring to summer in Taiwan) was the most common month of diagnosis, followed by January. January represents the coldest month in Taiwan, encompassing the traditional Chinese solar terms of “slight cold” and “great cold,” while June includes the “summer solstice,” marking the northernmost migration of the sun and the longest day of the year in the northern hemisphere (online suppl. Fig. S1).

In 2017, the NHIRD recorded 2,477 incident cases of encephalitis (males: 53.37%) requiring hospitalization, with a total follow-up of 23,692,010 person-years. The standardized incidence rate was 10.46 (95% CI: 10.04–10.87) per 100,000 person-years. Males exhibited a higher standardized incidence (11.22 per 100,000 person-years; 95% CI: 10.61–11.82) compared to females (9.70 per 100,000 person-years; 95% CI: 9.14–10.26).

Across age groups, the incidence was highest among patients aged > 70 years (22.20 per 100,000 person-years) and children aged 1–4 years (21.53 per 100,000 person-years). The lowest incidence was observed in patients aged 30–49 years (7.33 per 100,000 person-years) (online suppl. Fig. S2).

The overall incidence of encephalitis increased from 5.95 (95% CI: 5.61–6.28) per 100,000 person-years in 2003 to 10.46 (95% CI: 10.04–10.87) per 100,000 person-years in 2017, corresponding to an average APC of 1.6% (95% CI: 1.1–2.2; p < 0.05) (Table 2). Males consistently had a slightly higher incidence than females throughout the study period (Fig. 1a). The incidence trend increased sharply from 2003 to 2005, with an average APC of 25.4 (95% CI: 14.6–37.1). This was followed by a gradual decline in the trend from 2005 to 2013 (average APC: −2.0; 95% CI: −3.0 to −1.0), with a subsequent rise peaking in 2016 (annual incidence: 11.49; 95% CI: 11.05–11.92; average APC: 7.9; 95% CI: 5.6–10.4; Table 2; online suppl. Table S4).

To evaluate the trends in incidence rates from 2003 to 2017, individuals with encephalitis were divided into two age groups: < 18 years and ≥ 18 years. In the Taiwanese medical system, pediatric physicians or pediatric neurologists care for patients from birth to 18 years of age, and neurologists care for those ≥ 18 years. The incidence of encephalitis from 2003 to 2017 was consistently higher in the population < 18 years than in that ≥18 years (Fig. 1b). According to age stratification, the incidence distribution of encephalitis was highest in the 0–4 age group (∼20% in each calendar year) (Fig. 2).

Among patients < 18 years, the age- and sex-standardized incidence rate surged in 2003 and reached a peak in 2005, with an average APC of 25.4 (95% CI: 10.6–42.1). The trend then declined until 2014 (average APC: −5.7; 95% CI: −7.1 to −4.3), followed by a sharp increase peaking in 2016 (average APC: 14.3; 95% CI: 6.0–23.2) (online suppl. Tables S4 and S5).

For patients ≥ 18 years, the incidence rate increased from 2003 to 2005, remained stable between 2005 and 2013, and then gradually increased, peaking in 2016. From 2003 to 2017, the average APC was 2.6 (95% CI: 2.1–3.0; p < 0.05) (online suppl. Tables S4 and S6). Based on age stratification, the incidence distribution of encephalitis was highest in the 50–70 years old age group (15–20% in each calendar year) (Fig. 2).

The etiology of encephalitis was categorized as bacterial, viral, ADEM, or others. During each calendar year from 2003 to 2017, the others etiology was predominant (55.8–67.4%). In practice, autoimmune encephalitis or encephalopathy is classified into other encephalitis subgroups by certified coding specialists. All-cause viral encephalitis during the study period accounted for 12.6–24% of the cases, with higher proportions in the years 2016 and 2017. All-cause bacterial encephalitis accounted for 9.8–27.5% of cases during 2003–2017, but the proportions decreased in 2016 and 2017 (Fig. 3).

In the age group 0–4 years, the infectious etiology (including bacterial and viral) was as high as 50% in 2013 and then gradually decreased (the lowest rate was 32.4%, in 2011 and the 2nd highest rate was 46.8% in 2014). In the age 0–4-year group, bacterial encephalitis predominated from 2003 to 2015, whereas viral encephalitis predominated in 2016 and 2017. In the 5–10-year and 11–17-year age groups, an infectious etiology accounted for about 25% of cases, with a viral etiology predominating. Overall, the etiology in nearly 70% of the cases in the age group 5–17 years was others (Fig. 3).

An infectious etiology accounted for 35–40% of cases in the age group ≥ 18 years, with a viral etiology predominating. During the period 2003–2015, a bacterial etiology accounted for a higher proportion of cases than did a viral etiology, while the latter became the dominant form of infectious encephalitis in 2016 and 2017 (Fig. 3).

Figure 4 illustrates the 5-year survival rates stratified by various age subgroups. Each line represents a specific age group, ranging from 1 to 4 years to > 70 years, depicting survival rates over the 5-year period. Younger age groups, particularly those under 18 years, demonstrated consistently high survival rates with minimal decline over time. Conversely, older age groups exhibited a more pronounced decline in survival rates, reflecting the impact of age on long-term outcomes, with younger populations demonstrating greater resilience and more favorable survival rates.

The cumulative probability of developing PE reached 16.36% by year 1. Subsequently, the probability continued to increase, albeit at a slower rate, reaching 19.58% by year 5, with males exhibiting a higher risk compared to females (Fig. 5a). Stratification by age subgroup (Fig. 5b) revealed that children under 1 year old had a significantly higher cumulative probability of developing PE compared to older children, indicating that very young children are at a significantly higher risk of this complication.

To the best of our knowledge, this study provides the first population-based estimate of encephalitis-related hospital admissions in Taiwan. Over the 20-year period from 1998 to 2017, Taiwan reported 39,798 cases, with a predominance in children under 10 years of age and a higher incidence in males (55.6%). The highest incidence was observed in urban areas and northern Taiwan, likely reflecting the population density in these regions. The standardized incidence rate increased from 5.95 to 10.46 per 100,000 person-years between 2003 and 2017, with significant peaks in children aged 1–4 years and adults aged 70 years and older. Preexisting neuropsychiatric conditions requiring clinical management were common among these patients. The cumulative probability of developing PE reached 19.58% by 5 years, with the highest risk observed in infants under 1 year of age. Seasonal peaks in encephalitis incidence occurred in June and January, corresponding with Taiwan’s climatic patterns.

Our results highlight the significant influence of age on both the incidence and outcomes of encephalitis. The highest incidence rates were observed in children under 10 years, particularly those aged 1–4 years, and in elderly individuals (≥70 years), while middle-aged adults (30–49 years) had the lowest incidence rates. Survival outcomes also varied with age; younger age groups, particularly those under 18 years, demonstrated higher survival rates, whereas older age groups experienced a more pronounced decline in survival. Furthermore, the risk of developing PE was notably higher in infants under 1 year compared to slightly older children, highlighting age as a critical factor in determining both the risk of encephalitis and the likelihood of long-term complications.

In 2017, the incidence of encephalitis in Taiwan was 10.46 per 100,000 person-years, with significantly higher rates in children (21.53 per 100,000) and the elderly (22.20 per 100,000). These figures are comparable to international incidence rates of 6.34–13.8 per 100,000 person-years [5, 11‒13, 15] and the higher burden in children aligns with previous studies [10, 11].

This study also provides insights into the temporal and demographic patterns of encephalitis incidence, along with regional variation within Taiwan. Notably, a significant peak in incidence was observed in 2016, which coincided with a major seasonal influenza A/H3N2 outbreak in Taiwan during the 2015–2016 season. This outbreak led to over 2,000 confirmed influenza cases with intensive care unit admissions and 163 deaths [32]. The correlation between seasonal influenza and encephalitis incidence implies that specific infectious pathogens may play a role in encephalitis epidemiology. In Taiwan, the influenza season usually begins in December and peaks in January and February of the following year [32]. Taiwan’s seasonal influenza mass-vaccination program, directed and coordinated by the Taiwan CDC, was initiated in 1998 under the Ministry of Health and Welfare. The number of doses purchased for this program in Taiwan increased steadily from 180,000 in 1998 to 6.1 million in 2016 and 2017 [33]. Taiwan’s influenza vaccination program may have contributed to the observed trends in encephalitis incidence. This finding is consistent with a study from England, where the introduction of the combined measles-mumps-rubella vaccine in the 1980s effectively reduced encephalitis cases [23]. Beyond infectious causes, the recognition of immune-mediated encephalitis has increased in Taiwan over the past two decades, partly due to the availability of commercialized assays for autoantibodies, including anti-NMDA receptor antibodies [34].

The cumulative incidence of PE in our study was approximately 16% in the first year, increasing to 19.58% by the fifth year. This is consistent with a previous pediatric cohort study from northern Taiwan (1984–2000), which reported a 16.4% incidence of epilepsy over a mean follow-up period of 6 ± 4.6 years [35]. By comparison, a US pediatric cohort study (2004–2011) reported a higher incidence of PE at 24% [36]. Studies from Sweden and China have reported lower incidence rates of 9% and 11% at 2 years and 1.5 years, respectively [37, 38]. In adults, a US cohort study estimated a 30% incidence of PE within 1–7 years of follow-up [18]. PE affects 16–30% of all encephalitis patients, with varying prevalence rates across countries. In addition, early seizure presentation is associated with the risk of PE, with a 20-year risk of 22% in pediatric viral encephalitis patients presenting with early seizures compared to 10% for cases without early seizure [19]. In addition, abnormal magnetic resonance imaging findings can be used to predict PE [18]. We further identified very young children as being particularly vulnerable to PE compared to older children and adults.

This study had several limitations. First, the case definitions were based on diagnostic codes for encephalitis, which may have led to underdiagnosis. Subacute cases that were not hospitalized or misdiagnosed as psychiatric conditions before the availability of autoantibody testing for autoimmune encephalitis may have been missed. Additionally, super-acute cases that resulted in death before hospitalization may not have been captured in our data. However, the reliability of our incidence estimates is bolstered by being based on physician-diagnosed cases translated into a standardized diagnostic classification system. Second, the proportion of cases due to autoimmune encephalitis could not be specified due to limitations in the ICD9 – and even the new ICD10 – coding methods. For example, anti-NMDAR encephalitis is not categorized as limbic system dysfunction if it is not accompanied by a neoplasm and does not have a specific ICD code. Third, we were unable to include all relevant variables due to the lack of data linkage in the NHI database. Consequently, we could not detail all etiological factors of encephalitis, which are essential for understanding its causes and tailoring preventive measures. Finally, while we focused on incidence, mortality, and PE, our study lacked comprehensive data on other long-term sequelae and the impact on quality of life.

The strength of this study was the large number of patients analyzed, allowing for a comprehensive examination of this condition at the national level. Additionally, the long period covered by the dataset provided valuable insights into long-term trends in mortality and PE. Further research incorporating detailed clinical data and a broader range of outcomes is needed to enhance our understanding of encephalitis and inform better management strategies.

We conducted a population-based study and the results present the current epidemiology of encephalitis in Taiwan. Overall, the results underscore age as crucial for identifying individuals at the greatest risk for encephalitis and predicting their outcomes, thereby informing more focused prevention and treatment approaches.

This study protocol was reviewed and approved by the Institutional Review Board of Chang Gung Memorial Hospital at Linkou branch approved this study (Approval No. CORPG3G0481). The Ministry of Health and Welfare anonymized all data in this study, so patient consent was waived.

The authors have no conflicts of interest to declare.

This work was funded by Chang Gung Memorial Hospital, Linkou branch (projects CMRPG3K1941 and CMRPG3M1571). The funder had no role in the design and conduct of the study, data collection, management, analysis, and interpretation, manuscript preparation, review, or approval, or decision to submit the manuscript for publication.

I-J. Chou and Y.-S. Wang designed the research. I-J. Chou and C.-F. Kuo performed all statistical analyses and prepared figures and tables. Y.-S. Wang, C. Fan, and I-J. Chou drafted the manuscript. I-J. Chou, K.-L. Lin, and C.-F. Kuo prepared grant applications. J.-Y. Hou, Y.-T. Cheng, Y.-H. Liu, C.-Y. Kuo, and J.-J. Lin provided clinical input, helped to interpret the data from the medical point of view, and critically revised the intellectual contents of the manuscript. All authors commented on the manuscript and contributed to the writing process. I-J. Chou had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. The corresponding author (I-J. Chou) attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Yi-Shan Wang and Chi Fan contributed equally to this work.

The source data of this study are the National Health Insurance database which has limited access to investigators with ethical and administrative approval. The lead author (I-J. Chou) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; no important aspects of the study have been omitted. Any discrepancies from the study as planned have been omitted and explained. The code is available on request from the corresponding author.