Patients were recruited from 7 hospitals in Thailand: 5 in Bangkok (Queen Sirikit National Institute of Child Health, Rajvithi Hospital, Prasat Neurologic Institute, Ramathibodi Hospital, and Phramongkutkao Hospital) and 2 in the southern city of Hat Yai (Hat Yai Hospital, Prince Songkhla University Hospital). The main study physicians in each hospital were specialists in neurology or infectious diseases. During July 2003–August 2005, physicians identified potential study participants who met eligibility criteria and referred them to a study nurse, who obtained written informed consent. The protocol was approved by an institutional review board at the Centers for Disease Control and Prevention (CDC) and the Thailand Ministry of Public Health.

Enrollment criteria for patients of any age included all of the following: 1) acute brain dysfunction requiring hospitalization (new encephalopathy [i.e., altered mental status with or without lethargy], new onset of diffuse or focal central neurologic findings, or new onset of seizures); 2) acute onset of brain dysfunction within 14 days before or 7 days after admission to a study hospital; 3) documented fever (>38°C), history of fever, or hypothermia (<35°C); and 4) clinical indication for lumbar puncture as determined by the patient’s physician. After enrollment, participants were required to meet at least 1 of the following 3 criteria to meet the case definition of acute encephalitis or meningoencephalitis: 1) abnormal findings consistent with encephalitis seen on neuroimages obtained by computed tomography (CT) scan, magnetic resonance imaging (MRI), or cranial ultrasonography; 2) abnormal findings on electroencephalogram (EEG) consistent with encephalitis; or 3) cerebrospinal fluid (CSF) pleocytosis (>15 leukocytes/mm³ for infants <6 weeks of age and >5 leukocytes/mm³ for patients >6 weeks of age). Patients found to have an alternative confirmed diagnosis before discharge that explained their signs and symptoms (e.g., metabolic encephalopathy) were excluded from further study.

For each patient, study physicians completed standardized admission and discharge surveys documenting medical history, signs and symptoms, and neuroimaging and EEG results. Research nurses completed an extensive questionnaire about demographics, medical and exposure history, and laboratory results. Study nurses completed a follow-up questionnaire during each patient’s convalescent-phase visit 3–6 weeks after enrollment. The following specimens were collected: CSF (up to 6.5 mL); acute- and convalescent-phase blood (12.5–22.5 mL from children <5 years of age and 25.5 mL from all others); and oropharyngeal swab, saliva (0.7 mL), urine (10 mL), and fecal (10–20 g) specimens.

Specimens were tested for the presence of >30 pathogens (Table 1). After collection, specimens were immediately separated into portions for clinical testing at hospitals and research testing at reference laboratories (Technical Appendix). Within 24 hours of collection, specimens destined for reference laboratories were aliquoted for distribution to >20 laboratories at the Thailand National Institute of Health and CDC. Any specimens that could not be aliquoted within 24 hours were stored at −20°C and then at −70°C after aliquoting. Specimens from Hat Yai were transported on dry ice to Bangkok every 2–4 weeks. If CSF or serum sample volume was limited, the order of specimen testing was prioritized (Technical Appendix).

Definitions were created to define the etiologic link between identified pathogens and encephalitis; the etiology for each case was classified as confirmed, probable, possible, or unknown (Table 1). Each case could be assigned >1 etiology. The classification system considered the nature of an agent (well-established cause of encephalitis or not) and etiologic significance of a given positive laboratory test result.

Etiology was considered confirmed for cases with positive results for PCR, culture, antigen, or pathogen-specific IgM in CSF for a pathogen considered to be a well-established cause of encephalitis. For arboviruses commonly associated with encephalitis (i.e., JEV or West Nile virus), documentation of acute infection in paired serum samples was considered the diagnostic standard and therefore was considered confirmatory.

Etiology was considered probable for cases with positive results that were strongly suggestive but not considered confirmatory and not clearly established as diagnostic for encephalitis in all situations or cases for which a pathogen not generally established as a cause of encephalitis was detected in CSF. For example, etiology would be considered probable if PCR of CSF was positive for lymphotropic agents (e.g., cytomegalovirus, Epstein-Barr virus [EBV]) or if serum testing results provided evidence of acute infection for pathogens known to be associated with encephalitis, such as Orientia tsutsugamushi, Bartonella henselae, and Mycoplasma pneumoniae.

Etiology was considered possible for cases with no laboratory evidence of CNS involvement but some evidence of acute infection that suggested a potential etiologic role in encephalitis, such as a >4-fold rise in antibody titer for enteroviruses or an oropharyngeal swab sample positive for influenza or parainfluenza viruses. In addition, etiology was considered possible for cases for which information to determine etiology of encephalitis (e.g., human adenoviruses 40 and 41) was insufficient, even if detected in CSF (27). Etiology was considered unknown for cases for which all testing results were negative.

In an effort to differentiate between cases of encephalitis and meningoencephalitis, we defined a subset of patients who met the case definition and specifically had CSF pleocytosis and stiff neck as having meningoencephalitis. To provide a complete description of all patients who met the case definition, we retained the 8 patients for whom a sole confirmed bacterial agent commonly associated with bacterial meningitis was found. Pertinent analyses were performed with and without these 8 patients. For 1 patient for whom a bacterial etiology was confirmed, a viral etiology (JEV) was also confirmed; thus, this patient was not excluded from either analysis.

We present descriptive data with case counts and frequencies. We used the Wilcoxon rank-sum test to compare continuous variables between the confirmed/probable and possible/unknown etiologic groups. The prevalence of categorical variables was compared by using χ² analysis or the Fisher exact test. We considered 2-sided p values <0.05 to be statistically significant. Statistical analyses were performed by using SPSS Statistics for Windows version 21.0 (IBM Corp., Armonk, NY, USA).

Of the 149 cases, etiology was confirmed for 37 (25%), probable for 17 (11%), possible for 44 (30%), and unknown for 51 (34%) (Table 3). Among confirmed etiologies, JEV was identified in 15 patients and dengue virus in 2. In >1 patient, infection with the following was also identified: enteroviruses (n = 6), Cryptococcus spp. (n = 3), Haemophilus influenzae (n = 3), Streptococcus pneumoniae (n = 3), and varicella-zoster virus (VZV; n = 2). Among case-patients whose illness met the definition for confirmed etiology, a bacterial pathogen commonly associated with meningitis was confirmed for 8 patients (H. influenzae for 3, S. pneumoniae for 3, Neisseria meningitidis for 1, and Escherichia coli for 1).

Among patients for whom etiology was probable, O. tsutsugamushi, which causes scrub typhus, was found in 6 patients (5 became ill while in central Thailand and were hospitalized in Bangkok); JEV in 6, EBV in 3 (concurrent with Cryptococcus spp. in 1 and VZV in 1); M. pneumoniae in 3; and Rickettsia conorii, which causes spotted fever, in 3 patients. Among the 54 patients for whom etiology was confirmed/probable, a potentially treatable pathogen was identified for 25 (46%) (C. neoformans, E. coli, herpes simplex virus, H. influenzae, influenza viruses, M. pneumoniae, N. meningitidis, O. tsutsugamushi, R. conorii, S. pneumoniae, and Treponema pallidum). We did not include the patient with confirmed JEV and Salmonella group D infection in the treatable category.

Among the 98 patients for whom at least 1 pathogen was identified, a total of 156 pathogens were detected (Table 3); of 40 (41%) patients for whom >1 pathogen was detected, etiology was classified as possible for most. Of the 54 patients for whom etiology was confirmed/probable, >1 pathogen was classified as confirmed/probable for 6 (11%): (JEV/O. tsutsugamushi/R. conorii, JEV/R. conorii, JEV/Salmonella, Cryptococcus spp./EBV, VZV/EBV, VZV/influenza virus), whereas at least 1 possible etiology was also detected for 28 (52%) (p<0.01). Of 44 patients for whom a possible etiology was identified, >1 possible pathogen was detected for 12 (27%).

Of the 149 patients, 68 (46%) met the definition for meningoencephalitis; for 38 (56%) of these patients, EEG or neuroimaging findings were abnormal, consistent with encephalitis. Of these 38 patients, pathogens considered confirmed/probable etiologic agents were detected for 12 (32%): JEV for 8 (also R. conorii for 2 and O. tsutsugamushi for 1), Cryptococcus for 2, M. pneumoniae for 1, and R. conorii for 1. Of the 30 remaining patients, pathogens that conferred a confirmed/probable etiologic classification were detected for 16 (53%): JEV for 6, O. tsutsugamushi for 4, enterovirus for 3, N. meningitidis for 1, herpes simplex virus 1/2 (not distinguished) for 1, and VZV/EBV (positive for both) for 1.

At admission, very few demographic and clinical differences between those with confirmed/probable and possible/unknown etiologies were found (Table 4). Among patients for whom etiology was confirmed/probable, median temperature was higher, loss of consciousness was less common, and stiff neck was more common. When the 8 patients with bacterial meningitis were excluded from analysis, patients for whom etiology was confirmed/probable had a higher median temperature than those for whom etiology was possible/unknown (37.9 vs. 37.7, p = 0.04) and were more likely to have a stiff neck (54% vs. 31%, p = 0.01). Among the 54 patients for whom etiology was confirmed or probable, no differences were found for demographic characteristics and only 1 difference was found for clinical characteristics at admission (Table 4) between those for whom etiology was or was not treatable; temperature was lower among those for whom etiology was treatable (37.9 vs. 39, p<0.01).

A total of 15 (10%) patients died. The mortality rate was lower, but not significantly, among patients for whom etiology was confirmed/probable than for those for whom etiology was possible/unknown (3.9% vs. 14%, p = 0.09; Table 4). When the 8 patients for whom only a confirmed bacterial etiology was found were excluded from analysis, this difference was similar (4.3% vs. 14%, p = 0.14). Among those who died, etiology was confirmed for 2 (Cryptococcus spp. in a 5-year-old boy and VZV in a 51-year-old HIV-infected woman). Among those who died, 9 (60%) were male and the median age was 34 years of age (1 was <1 year, 1 was 1–4 years, 4 were 5–18 years, 8 were 19–64 years, and 1 was ≥65 years). The median interval between hospital admission and death (or discharge for survivors) was 6 days (vs. 16 days to discharge, p = 0.3), between onset of neurologic symptoms and death was 10 days (vs. 18.5 days to discharge, p = 0.3), and between fever onset and death was 10 days (vs. 20 to discharge, p = 0.2). Among the 134 patients who survived, 43 (32%) remained hospitalized at the time of the convalescent-phase interview, 86 (64%) had returned home, and 5 (3.7%) were lost to follow-up; 5 patients had a seizure after discharge. Of the 86 persons who returned home, 52 (60%) reported complete cognitive recovery; of these, 45 (87%) of these functioned independently or at the same level of care as before hospitalization.