Etiological spectrum, risk factors, and outcomes of pediatric acute encephalitis: a 17-year pediatric intensive care unit retrospective cohort study

Abstract

Background

Acute encephalitis is a serious neurological emergency in children, characterized by altered consciousness and brain dysfunction. Prompt recognition and management are essential to reduce morbidity and mortality.

Aims

This study aimed to determine the prevalence and identify clinical and paraclinical factors associated with mortality and neurological sequelae among children with acute encephalitis admitted to a pediatric intensive care unit (PICU).

Methods

A retrospective cohort study was conducted over 17 years (2008–2024) in the PICU of Hédi Chaker University Hospital, Sfax, Tunisia. Cases were defined according to International Encephalitis Consortium criteria.

Results

Of 2,240 PICU admissions, 58 children (2.6%) met inclusion criteria, representing a mean of 3.4 cases per year over the 17-year study period in this tertiary care PICU. Median age was 5 years; 55.2% were male. Common presentations included seizures (82.8%), altered consciousness (69.0%), and fever (82.8%). Among the 46 patients who underwent MRI, neuroimaging abnormalities were detected in 35 (76.1%). Etiology was identified in 81.0% of cases (47/58). Mortality was 20.7% (12/58), and among the 46 survivors, 9 (19.6%) developed neurological sequelae. On univariate analysis, abnormal movements (p = 0.010) and need for supplemental oxygen (p = 0.027) were significantly associated with sequelae, while age under 1 year, seizures at admission, abnormal CT findings, and requirement for advanced resuscitation measures were associated with mortality.

Conclusion

Pediatric acute encephalitis carries substantial mortality and morbidity. Early diagnosis, comprehensive etiological investigation, and timely intensive supportive care are essential to improve outcomes. Prevention through immunization and enhanced surveillance should be prioritized.

Introduction

Acute encephalitis in children is a severe neurological emergency characterized by non-suppurative inflammation of the cerebral parenchyma, leading to acute neurological dysfunction [1]. Its rapidly progressive course and high morbidity and mortality necessitate intensive care management. The global incidence in children is estimated at 3.5–7.4 cases per 100,000 per year, reflecting a significant but variable disease burden [2, 3].

Pathophysiologically, pediatric acute encephalitis is dichotomized into two principal etiological categories: infectious (primary) and immune-mediated. Infectious encephalitis results from the direct invasion of neural tissue by pathogens, whereas immune-mediated encephalitis (IME) arises from a dysregulated host immune response targeting cerebral antigens. The latter category encompasses distinct entities such as Acute Disseminated Encephalomyelitis (ADEM) and Immune-mediated encephalitis, the latter being specifically defined by the presence of pathogenic autoantibodies against neuronal surface or synaptic proteins [4]. Notably, the clinical and diagnostic delineation between these categories is often complex, as post-infectious inflammatory mechanisms can blur this distinction, and the two processes may coexist [5].

Despite considerable advances in diagnostic modalities, including refined molecular biological techniques and expanded panels for autoantibody detection, establishing a definitive etiology remains elusive in a substantial proportion of cases [6]. This diagnostic uncertainty complicates prognostic stratification and targeted therapeutic intervention. The clinical presentation is notoriously polymorphic and non-specific, commonly manifesting as encephalopathy, fever, seizures, focal neurological deficits, and signs of meningeal irritation [5]. Consequently, outcomes are variable, with survivors facing a considerable risk of persistent cognitive, motor, and neuropsychological sequelae, highlighting the critical imperative for early recognition and a multidisciplinary approach to care [7].

Given these challenges, a thorough understanding of the epidemiological patterns and determinants of severe disease is essential for optimizing clinical management and resource allocation in critical care settings. Therefore, the primary aims of this study are to determine the prevalence of acute encephalitis among children admitted to a Pediatric Intensive Care Unit (PICU) and to identify specific clinical and laboratory factors associated with mortality and neurological sequelae in this population, thereby contributing to improved prognostic assessment and risk stratification.

Patients and methods

Study design and setting

This was a single-center, retrospective cohort study conducted in the Pediatric Intensive Care Unit (PICU) of Hédi Chaker University Hospital, Sfax, Tunisia, over a 17-year period (January 1, 2008 – December 31, 2024).

Study population

All patients aged 1 month to 14 years diagnosed with acute encephalitis and admitted to the PICU were included. Cases met the 2013 International Encephalitis Consortium criteria [7]: encephalopathy (altered consciousness ≥ 24 h) plus at least two of: fever ≥ 38 °C, new-onset seizures, focal neurological deficits, CSF pleocytosis (≥ 5 leukocytes/mm³), EEG findings consistent with encephalitis, or neuroimaging suggestive of encephalitis. Excluded were patients ≥ 14 years and those with alternative diagnoses (metabolic, anoxic-ischemic, toxic, neoplastic encephalopathies, vasculitis, brain abscess, or subdural empyema).

Etiological classification and certainty levels

Etiology was classified as infectious (viral/bacterial) or immune-mediated (ADEM, autoimmune encephalitis). All cases were retrospectively reviewed by two investigators (AE, FS) using predefined rules; disagreements were resolved by a third (MB). Certainty levels for infectious etiologies were [8]:

• **Certain: ** Pathogen isolated from CSF/brain tissue, or specific IgM in CSF

• **Probable: ** Strong serological evidence or isolation from another sterile site (e.g., blood) **Possible: ** Suggestive but inconclusive serology, or compatible clinical/radiological findings without microbiological confirmation

Cases without microbiological, serological, or autoantibody evidence were classified as “undetermined etiology.”

Bacterial encephalitis definition

Bacterial encephalitis required: (a) clinical criteria (encephalopathy plus ≥ 2 minor criteria per International Encephalitis Consortium) AND (b) at least one of: certain (pathogen in CSF), probable (pathogen in blood with compatible CSF), or possible (compatible clinical/radiological findings with antibiotic response, without microbiological confirmation).

Data collection

Data were extracted from electronic and physical records, including: demographics, vaccination status (per national schedule), prodromal symptoms, clinical presentation, Glasgow Coma Scale (GCS), status epilepticus occurrence, neuroimaging (CT/MRI), EEG, laboratory parameters (complete blood count, CRP, CSF analysis), microbiological testing (blood/CSF cultures, PCR for HSV/enteroviruses, serology), and autoimmune markers (anti-NMDAR, anti-MOG, anti-GAD, anti-VGKC) when indicated.

Outcomes

Short-term outcomes included PICU length of stay, in-hospital mortality, and unfavorable evolution (worsening neurological status, new seizures). Long-term outcomes were assessed during post-discharge follow-up, documenting motor, cognitive, or behavioral sequelae.

Statistical analysis

Qualitative variables are presented as numbers (percentages). Quantitative variables are expressed as mean (SD) or median (range). Associations were examined using Pearson’s chi-square or Fisher’s exact test (categorical) and t-test or Mann-Whitney U test (continuous). Significance threshold was p < 0.05. Multivariate logistic regression was not performed due to small sample size (n = 58) and low outcome events (12 deaths, 9 sequelae), which would risk overfitting. This corresponded to an insufficient number of events per variable (EPV), below the recommended threshold for reliable multivariate modeling.

Results

Over 17 years, 58 children met inclusion criteria, representing 2.6% of 2,240 PICU admissions (mean 3.4 cases/year) (Fig. 1).

Fig. 1.

Annual incidence of acute encephalitis cases against the backdrop of total pediatric ICU admissions from 2008 to 2024

Descriptive study

Baseline characteristics

Male predominance (32/58, 55.2%; sex ratio 1.2). Median age 5.4 years (range 2 months–13.7 years); age distribution: <1 year (20/58, 34.5%), 1–3 years (10/58, 17.2%), 3–9 years (23/58, 39.7%), > 9 years (5/58, 8.6%). Most patients resided in Sfax governorate (35/58, 60.3%). Seasonal peak: spring/summer (33/58, 57.2%). All were up-to-date with vaccinations; one had recent vaccination. Prodromal symptoms: flu-like (13/58, 22.4%), rash (9/58, 15.5%).

Clinical presentation

Mean prodrome-to-encephalitis interval: 6.4 days (SD 6.2). Presenting features: seizures 48/58 (82.8%), altered mental status 40/58 (69.0%), coma 31/58 (53.4%), status epilepticus 14/58 (23.9%), fever 48/58 (82.8%), meningeal signs 10/58 (17.2%). Generalized tonic-clonic seizures predominated (97.8% of seizure cases). Focal deficits were less common (< 7% each).

Paraclinical Investigations

Paraclinical investigations revealed a median white blood cell count of 10,480/µL, with hyperleukocytosis (> 10,000/µL) in 51.7% of patients and leukopenia (< 4,000/µL) in 3.4%, while the median C-reactive protein level was 2.75 mg/L, elevated (> 10 mg/L) in 32.7% of cases; cerebrospinal fluid analysis, performed in 49 patients (84.5%), showed pleocytosis (≥ 5 WBC/µL) in 30 (61.2%), a median cell count of 10 cells/µL, elevated protein (> 0.4 g/L) in 12 (24.5%), and hypoglycorrhachia in 3 (6.1%), with a positive culture in one case (2.0%); microbiological testing identified HSV PCR positivity in 6 of 56 CSF samples (10.7%), a positive blood culture in 1 of 49 cases (2.0%), and positive Mycoplasma pneumoniae serology in 1 of 6 tested patients (16.7%); neuroimaging demonstrated abnormal findings in 35 of 46 brain MRIs (76.1%) and in 23 of 50 brain CT scans (46.0%); and neurophysiological assessment via electroencephalogram, conducted in 39 patients (67.2%), was abnormal in 35 (89.7%).

Etiological distribution

An etiology was identified in 47 of 58 pediatric patients (81.0%). Infectious encephalitis was the most frequent category, affecting 31 patients (53.4%), with viral causes in 18 (31.0%) and bacterial causes in 13 (22.4%). Immune-mediated encephalitis accounted for 16 cases (27.6%), predominantly post-viral acute disseminated encephalomyelitis (ADEM) in 13 (22.4%). One case (1.7%) of limbic encephalitis was documented. In 10 patients (17.2%), the cause remained undetermined despite investigation (Fig. 2).

Fig. 2.

Etiological distribution of the 58 pediatric acute encephalitis cases diagnosed during the study period

**Certainty levels for etiological diagnoses.** Among the 47 cases with an identified etiology, the distribution by certainty level was as follows. Of the 18 viral encephalitis cases, 9 (50.0%) were classified as certain (including 6 HSV PCR-positive in CSF), 4 (22.2%) as probable, and 5 (27.8%) as possible. Of the 13 bacterial encephalitis cases, only 1 (7.7%) was classified as certain (CSF culture positive for *Streptococcus pneumoniae*), none as probable, and the remaining 12 (92.3%) as possible. All 15 ADEM cases were diagnosed using established clinical and radiological pediatric criteria [9] rather than microbiological certainty levels. The single limbic encephalitis case was autoantibody-positive (certain). Overall, 11 cases (23.4% of identified etiologies) met criteria for certain diagnosis, 4 (8.5%) for probable, and 17 (36.2%) for possible; certainty levels were not applicable for the 15 ADEM cases (31.9% of identified etiologies).

Some of these undetermined cases may have involved encephalopathy secondary to systemic infections, such as sepsis-associated encephalopathy, reflecting the challenges of retrospective diagnostic differentiation.

Therapeutic management

Empirical acyclovir (50/58, 86.2%; median 6 days). Empirical antibiotics (51/58, 87.9%; mainly cefotaxime/vancomycin). Corticosteroids (26/58, 44.8%), IV immunoglobulins (9/58, 15.5%). Invasive mechanical ventilation (38/58, 65.5%), sedation (33/58, 56.9%), vasoactive drugs (7/58, 12.1%). Anticonvulsants (45/58, 77.6%).

Outcomes

The median length of intensive care unit (ICU) stay was 6 days (mean 7.2 days, standard deviation 5.5). Due to the retrospective design, neurological outcomes were assessed using predefined clinical criteria rather than standardized scales such as the Pediatric Cerebral Performance Category (PCPC) or Glasgow Outcome Scale (GOS).

**Patient disposition.** Of the 58 patients in the total cohort, 12 (20.7%) died during hospitalization. The remaining 46 patients (79.3%) were discharged alive as survivors.

**Follow-up status among survivors.** Among the 46 survivors, 35 patients (76.1% of survivors, 60.3% of the total cohort) had documented follow-up visits with outcome ascertainment. The remaining 11 survivors (23.9% of survivors, 19.0% of the total cohort) were lost to follow-up; their neurological outcome status could not be determined.

**Outcomes among those with follow-up.** Of the 35 patients with documented follow-up, 26 (74.3% of followed patients, 44.8% of the total cohort) had a favorable outcome without neurological sequelae. The remaining 9 patients (25.7% of followed patients, 15.5% of the total cohort) developed neurological sequelae.

**Estimated sequelae range for all survivors.** When calculating the sequelae rate among all 46 survivors (including those lost to follow-up), the true proportion depends on assumptions about the 11 patients without follow-up. Assuming none of the lost patients had sequelae yields a minimum estimate of 19.6% (9/46). Assuming all lost patients had sequelae yields a maximum estimate of 43.5% (20/46). The true rate likely lies between these extremes.

**Note on outcome assessment standardization.** Neurological outcomes were assessed using predefined clinical criteria documented during outpatient consultations (including evaluation of motor, cognitive, and behavioral function) rather than validated standardized scales such as the Paediatric Cerebral Performance Category (PCPC) or Glasgow Outcome Scale (GOS). This limits comparability with studies that used these instruments and may introduce assessment bias.

Among the nine survivors with documented sequelae, neurosensory deficits were most common (7 patients, 12.0% of the total cohort), followed by motor deficits (2 patients, 3.4%). Follow-up brain MRI was performed in only three children (5.2% of the cohort); all three had normal findings. Due to the very small sample size and likely selection bias (patients with poor outcomes may have been less likely to undergo follow-up imaging), this finding should not be interpreted as evidence of favorable radiological outcomes in the majority of survivors. The clinical, paraclinical, etiological, and outcome characteristics of the 58 cases of acute encephalitis admitted to the Pediatric, Emergency, and Pediatric Resuscitation departments between 2008 and 2024 are detailed in the study, summarizing demographics, presentation, diagnostic findings, etiology, and outcomes (Table 1).

Table 1.

Clinical, paraclinical, and outcome characteristics of pediatric acute encephalitis cases (2008–2024)

Observation number	Age	Sex	Clinical signs	MRI	EEG	Etiology	Sequels	Death
1	< 1 year	Male	AMS-Coma-Seizure	Abnormal	Normal	Indeterminate origin		Yes
2	3–9 years	Female	AMS-Coma-Status Epilepticus-Fever	Abnormal	Abnormal	Indeterminate origin	Motor sequels	No
3	< 1 year	Male	AMS-Coma-Status epilepticus	Abnormal	Abnormal	Indeterminate origin	No sequels	No
4	3–9 years	Male	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Post-viral ADEM	No sequels	No
5	< 1 year	Male	AMS-Coma-Status Epilepticus-Fever	Abnormal	Abnormal	Bacterial	Neurosensory sequels	No
6	3–9 years	Male	AMS-Coma-Fever	Normal		Post-viral ADEM		Yes
7	3–9 years	Male	AMS-Coma	Abnormal	Abnormal	Post-viral ADEM	No sequels	No
8	1–2 year	Male	AMS-Seizure-Fever	Abnormal	Normal	Viral	No sequels	No
9	< 1 year	Male	AMS-Coma	Abnormal		Post-vaccinal ADEM		Yes
10	3–9 years	Male	AMS-Fever	Normal	Abnormal	Viral	No sequels	No
11	10–14 years	Female	AMS-Coma-Status Epilepticus-Fever	Abnormal	Abnormal	Bacterial	No sequels	No
12	3–9 years	Female	AMS-Coma-Seizure-Fever			Post-viral ADEM	No sequels	No
13	3–9 years	Male	Status epilepticus-Fever	Abnormal		Post-viral ADEM	No sequels	No
14	< 1 year	Female	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Bacterial		Yes
15	< 1 year	Female	Seizure-Fever			Post-viral ADEM	No sequels	No
16	3–9 years	Female	AMS-Coma-Seizure-Fever	Normal	Abnormal	Viral	No sequels	No
17	< 1 year	Female	AMS-Fever			Bacterial		Yes
18	3–9 years	Female	AMS-Coma-Status Epilepticus-Fever	Abnormal		Viral	Neurosensory sequels	No
19	10–14 years	Female	AMS-Coma-Status Epilepticus-Fever	Normal		Post-viral ADEM	No sequels	No
20	< 1 year	Female	Seizure-Fever	Abnormal	Abnormal	Viral	Neurosensory sequels	No
21	3–9 years	Female	AMS-Coma-Status Epilepticus-Fever	Abnormal	Abnormal	Indeterminate origin	Neurosensory sequels	No
22	10–14 years	Male	AMS-Coma-Seizure-Fever	Normal	Abnormal	Indeterminate origin	No sequels	No
23	3–9 years	Female	AMS-Seizure-Fever	Abnormal	Abnormal	Viral	No sequels	No
24	3–9 years	Female	AMS-Coma-Seizure-Fever	Abnormal		Post-viral ADEM	No sequels	No
25	3–9 years	Female	AMS-Seizure-Fever			Viral	No sequels	No
26	3–9 years	Female	AMS-Coma-Seizure-Fever	Abnormal		Viral		No
29	< 1 year	Male	Status epilepticus	Abnormal	Abnormal	Post-viral ADEM	No sequels	No
31	< 1 year	Female	Seizure-Fever	Abnormal	Abnormal	Viral	No sequels	No
32	< 1 year	Male	AMS-Coma-Fever	Normal	Abnormal	Bacterial		Yes
33	< 1 year	Male	AMS-Seizure-Fever	Abnormal	Abnormal	Post-viral ADEM	No sequels	No
34	3–9 years	Male	Status epilepticus-Fever	Normal	Normal	Bacterial	No sequels	No
35	< 1 year	Male	AMS-Coma-Fever	Normal		Bacterial		Yes
36	3–9 years	Male	AMS-Seizure	Abnormal	Abnormal	Limbic Encephalitis	Neurosensory sequels	No
38	3–9 years	Male	AMS		Abnormal	Indeterminate origin	No sequels	No
39	3–9 years	Female	Status epilepticus-Fever	Normal	Abnormal	Viral	No sequels	No
40	< 1 year	Female	Seizure-Fever	Abnormal	Abnormal	Bacterial		Yes
41	3–9 years	Male	Seizure			Indeterminate origin	No sequels	No
42	1–2 year	Male	AMS-Coma-Status Epilepticus-Fever		Abnormal	Indeterminate origin		Yes
43	3–9 years	Female	Seizure-Fever	Abnormal	Abnormal	Bacterial		No
45	3–9 years	Male	AMS-Coma-Status Epilepticus-Fever		Abnormal	Viral		No
46	10–14 years	Male	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Indeterminate origin		Yes
48	1–2 year	Male	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Post-viral ADEM	Neurosensory sequels	No
49	3–9 years	Male	Status epilepticus-Fever	Normal		Bacterial		No
50	1–2 year	Male	Seizure-Fever	Abnormal	Abnormal	Viral		Yes
52	10–14 years	Female	AMS-Seizure-Fever	Abnormal	Abnormal	Viral		No
53	< 1 year	Male	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Viral		No
58	< 1 year	Female	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Viral	Motor sequels	No
59	1–2 year	Female	Coma-Status Epilepticus-Fever		Normal	Viral		No
61	3–9 years	Female	AMS-Seizure-Fever			Indeterminate origin		No
62	10–14 years	Male	Seizure-Fever			Indeterminate origin		No
63	< 1 year	Male	Status epilepticus-Fever	Abnormal	Abnormal	Post-vaccinal ADEM	No sequels	No
65	< 1 year	Male	Fever			Bacterial		Yes
67	1–2 year	Female	Status epilepticus-Fever	Abnormal	Abnormal	Viral	No sequels	No
68	< 1 year	Female	AMS-Coma-Seizure-Fever	Abnormal	Abnormal	Bacterial	Neurosensory sequels	No
69	< 1 year	Female	Seizure	Abnormal		Viral		No
70	10–14 years	Male	AMS-Coma-Fever	Normal		Bacterial		No
71	10–14 years	Male	AMS-Coma-Seizure	Abnormal	Abnormal	Post-viral ADEM	No sequels	No
72	1–2 year	Male	AMS-Coma-Status Epilepticus-Fever	Abnormal	Abnormal	Post-viral ADEM	No sequels	No

Follow-up MRI performed in only 3 patients; findings not representative of the cohort

ADEM Acute disseminated encephalomyelitis, AMS Altered mental state, EEG Electroencephalogram, MRI Magnetic resonance imaging

Analytical study

In the analytical assessment of factors associated with outcomes, demographic characteristics such as age under one year (p = 0.393) and sex (p = 0.264) were not significantly associated with neurological sequelae. Among clinical signs at presentation, univariate analysis identified a significant association between the presence of abnormal movements and neurological sequelae (OR 6.33, 95% CI: 1.48–27.11, p = 0.010) (Table 2). No significant associations were found between sequelae and any initial paraclinical findings—including hematological, inflammatory, cerebrospinal fluid, neuroimaging, or electroencephalographic parameters (Table 3). Likewise, the etiological classification of encephalitis (Table 4) and the length of ICU stay (p = 0.445, Fig. 3) showed no significant relationship with the development of sequelae. Notably, the need for supplemental oxygen via nasal cannula showed a significant univariate association with neurological sequelae (OR 4.50, 95% CI: 1.18–17.14, p = 0.027) (Table 5). Due to the small sample size and multiple comparisons, these findings should be interpreted as exploratory and hypothesis-generating rather than confirmatory.

Table 2.

Association between acute clinical manifestations and neurological sequelae in pediatric encephalitis

Encephalitic signs	No sequelae (n = 26)	Sequelae (n = 9)	*p*-value
Altered mental state	17 (68%)	8 (32%)	0,235
Coma	11 (61,1%)	7 (38,9%)	0,121
Seizure	23 (71,9%)	9 (28,1%)	0,553
Status epilepticus	9 (69,2%)	4 (30,8%)	0,698
Agitation	2 (66,7%)	1 (33,3%)	0,603
Deficit signs	0 (0%)	1 (100%)	0,257
Behavioral disorders	1 (100%)	0 (0%)	0,743
Abnormal movements	1 (20%)	4 (80%)	0,010
Functional impairment	1 (50%)	1 (50%)	0,454

Percentages are row percentages (e.g., of the 25 patients with altered mental status, 68% had no sequelae and 32% had sequelae). Fisher's exact test was used for all comparisons

GCS Glasgow Coma Scale

Table 3.

Association between initial paraclinical findings and neurological sequelae in pediatric encephalitis

Paraclinical data		No sequelae (n = 26)	Sequelae (n = 9)	*p*-value
Hyperleukocytosis (> 10,000/ml)	No	12 (70,6%)	5 (29,4%)	0,500
	Yes	13 (76,5%)	4 (23,5%)
Leukopenia (< 4,000/ml)	No	24 (72,7%)	9 (27,3%)	0,735
	Yes	1 (100%)	0 (0%)
Positive CRP (> 10 mg/l)	No	21 (80,8%)	5 (19,2%)	0,192
	Yes	5 (55,6%)	4 (44,4%)
Lumbar puncture appearance	Clear	20 (71,4%)	8 (28,6%)	0,550
	Hematic	3 (100%)	0 (0%)
Lumbar puncture culture	Negative	23 (74,2%)	8 (25,8%)	-1
	Positive	0 (0%)	0 (0%)
Brain magnetic resonance imaging	Normal	6 (100%)	0 (0%)	0,141
	Abnormal	15 (62,5%)	9 (37,5%)
Brain CT scan	Normal	16 (76,2%)	5 (23,8%)	0,417
	Abnormal	6 (60%)	4 (40%)
Electro-encephalogram	Normal	2 (100%)	0 (0%)	0,487
	Abnormal	17 (68%)	8 (32%)

Percentages represent row percentages for each finding (e.g., of the 17 patients without hyperleukocytosis, 70.6% had no sequelae). Fisher’s exact test was used for all statistical comparisons

CRP C-Reactive Protein, MRI Magnetic Resonance Imaging, CT Computed Tomography

¹*p*-value not applicable (no positive cultures in either outcome group)

Table 4.

Association between etiological classification and neurological sequelae in pediatric encephalitis

Etiology		No sequelae (n = 26)	Sequelae (n = 9)	*p*-value
Infectious encephalitis	No	16 (80%)	4 (20%)	0,451
	Yes	10 (66,7%)	5 (33,3%)
Acute disseminated encephalomyelitis	No	14 (63,6%)	8 (36,4%)	0,109
	Yes	12 (92,3%)	1 (7,7%)
Encephalitis of undetermined origin	No	22 (75,9%)	7 (24,1%)	0,635
	Yes	4 (66,7%)	2 (33,3%)

Percentages are row percentages (e.g., among the 20 patients without an infectious etiology, 80% had no sequelae). Fisher’s exact test was used for all statistical comparisons

Fig. 3.

Length of hospital stay and sequelae in pediatric encephalitis patients

Table 5.

Association between administered therapies and neurological sequelae in pediatric acute encephalitis

Therapeutic measures		No sequelae (n = 26)	Sequelae (n = 9)	*p*-value
Acyclovir	No	3 (75%)	1 (25%)	0,732
	Yes	23 (74,2%)	8 (25,8%)
Antibiotics	No	5 (71,4%)	2 (28,6%)	0,594
	Yes	21 (75%)	7 (25%)
Corticoids	No	14 (73,7%)	5 (26,3%)	0,619
	Yes	12 (75%)	4 (25%)
Plasmapheresis	No	26 (74,3%)	9 (25,7%)	-1
	Yes	0 (0%)	0 (0%)
Immunoglobulins	No	21 (72,4%)	8 (27,6%)	0,507
	Yes	5 (83,3%)	1 (16,7%)
Anticonvulsants	No	6 (100%)	0 (0%)	0,304
	Yes	20 (69%)	9 (31%)
Sedation	No	12 (80%)	3 (20%)	0,697
	Yes	13 (68,4%)	6 (31,6%)
Vasoactive drugs	No	26 (76,5%)	8 (23,5%)	0,257
	Yes	0 (0%)	1 (100%)
Nasal cannula oxygen therapy	No	24 (82,8%)	5 (17,2%)	0,027
	Yes	2 (33,3%)	4 (66,7%)
Intermittent mandatory ventilation	No	10 (76,9%)	3 (23,1%)	0,557
	Yes	16 (72,7%)	6 (27,3%)

Percentages are row percentages for each intervention (e.g., of the 4 patients who did not receive acyclovir, 75% had no sequelae). Fisher’s exact test was used for all statistical comparisons

IVIG Intravenous Immunoglobulins

Bold *p*-value indicates statistical significance (*p* < .05)

¹*p*-value not applicable; no patient in the cohort received plasmapheresis

In terms of mortality, on univariate analysis, infants under one year of age had a significantly higher mortality rate (OR 4.29, 95% CI: 1.33–13.85, p = 0.015) (Table 6). Conversely, the presence of seizures at admission was associated with lower mortality (OR 0.17, 95% CI: 0.05–0.56, p = 0.003) (Table 7). Neuroimaging findings also proved prognostic, as an abnormal brain CT scan showed a significant univariate association with mortality (OR 3.41, 95% CI: 1.13–10.29, p = 0.030) (Table 8). The etiological classification of encephalitis showed no significant association with mortality (Table 9), nor did the duration of ICU stay (Fig. 4). However, advanced resuscitation measures were associated with mortality: sedation (OR 7.20, 95% CI: 1.72–30.14, p = 0.008), vasoactive agents (OR 5.63, 95% CI: 1.21–26.24, p = 0.028), and invasive mechanical ventilation (OR 9.33, 95% CI: 2.03–42.86, p = 0.005) (Table 10). These advanced resuscitation measures reflect treatment responses to severe illness rather than independent baseline prognostic factors, and the wide confidence intervals reflect the imprecision inherent in small-sample univariate analyses.

Table 6.

Association between epidemiological characteristics and mortality in pediatric acute encephalitis

Epidemiological data		Death		*p*-value
		No (n = 46)	Yes (n = 12)
Age < 1 year	No	34 (89,5%)	4 (10,5%)	0,015
	Yes	12 (60%)	8 (40%)
Sex	Male	23 (71,9%)	9 (28,1%)	0,121
	Female	23 (88,5%)	3 (11,5%)

Percentages are row percentages (e.g., among the 38 patients aged ≥1 year, 89.5% survived). Fisher’s exact test was used for statistical comparisons

Bold *p*-value indicates statistical significance (*p* < .05)

Table 7.

Association between clinical manifestations and mortality in pediatric acute encephalitis

Encephalitic signs	Death		p
	No (n = 46)	Yes (n = 12)
Altered mental state	31 (77,5%)	9 (22,5%)	0,736
Coma	23 (74,2%)	8 (25,8%)	0,303
Seizure	42 (87,5%)	6 (12,5%)	0,003
Status epilepticus	16 (94,1%)	1 (5,9%)	0,088
Agitation	7 (100%)	0 (0%)	0,325
Deficit signs	1 (33,3%)	2 (66,7%)	0,106
Speech disorders	1 (50%)	1 (50%)	0,374
Gait disorders	1 (50%)	1 (50%)	0,374
Behavioral disorders	2 (66,7%)	1 (33,3%)	0,508
Abnormal movements	6 (100%)	0 (0%)	0,328
Functional impairment	2 (50%)	2 (50%)	0,186

Data presented as n (%). Percentages are row percentages (e.g., 77.5% of patients with altered mental status survived). Fisher’s exact test was used for all comparisons

Bold *p*-value indicates statistical significance (*p* < .05)

Table 8.

Association between initial paraclinical findings and mortality in pediatric acute encephalitis

Paraclinical data		Death		*p*-value
		No (n = 46)	Yes (n = 12)
Hyperleukocytosis (> 10,000/ml)	No	23 (85,2%)	4 (14,8%)	0,273
	Yes	22 (73,3%)	8 (26,7%)
Leukopenia (< 4,000/ml)	No	43 (78,2%)	12 (21,8%)	0,620
	Yes	2 (100%)	0 (0%)
Positive CRP (> 10 mg/l)	No	33 (84,6%)	6 (15,4%)	0,180
	Yes	13 (68,4%)	6 (31,6%)
Lumbar puncture appearance	Clear	34 (75,6%)	11 (24,4%)	0,562
	Hematic	4 (100%)	0 (0%)
Lumbar puncture culture	Negative	38 (79,2%)	10 (20,8%)	0,224
	Positive	0 (0%)	1 (100%)
Brain magnetic resonance imaging	Normal	8 (72,7%)	3 (27,3%)	0,664
	Abnormal	29 (82,9%)	6 (17,1%)
Brain CT scan	Normal	25 (92,6%)	2 (7,4%)	0,030
	Abnormal	15 (65,2%)	8 (34,8%)
Electro-encephalogram	Normal	3 (75%)	1 (25%)	0,563
	Abnormal	29 (82,9%)	6 (17,1%)

Data presented as n (%). Percentages are row percentages for each finding. Fisher’s exact test was used for all statistical comparisons

Bold *p*-value indicates statistical significance (*p* < .05)

Table 9.

Association between etiological classification and mortality in pediatric acute encephalitis

Etiology		Death		*p*-value
		No (n = 46)	Yes (n = 12)
Infectious encephalitis	No	22 (81,5%)	5 (18,5%)	0,703
	Yes	24 (77,4%)	7 (22,6%)
Acute disseminated encephalomyelitis	No	33 (76,7%)	10 (23,3%)	0,712
	Yes	13 (86,7%)	2 (13,3%)
Encephalitis of undetermined origin	No	38 (80,9%)	9 (19,1%)	0,681
	Yes	8 (72,7%)	3 (27,3%)

Data presented as n (%). Percentages are row percentages (e.g., 81.5% of patients without infectious encephalitis survived). Fisher’s exact test was used for all statistical comparison

Fig. 4.

Length of hospital stay and mortality in pediatric encephalitis patients

Table 10.

Association between administered therapies and mortality in pediatric acute encephalitis

Therapeutic measures		Death		*p*-value
		No (n = 46)	Yes (n = 12)
Acyclovir	No	6 (75%)	2 (25%)	0,665
	Yes	40 (80%)	10 (20%)
Antibiotics	No	7 (100%)	0 (0%)	0,325
	Yes	39 (76,5%)	12 (23,5%)
Corticoids	No	28 (87,5%)	4 (12,5%)	0,088
	Yes	18 (69,2%)	8 (30,8%)
Plasmapheresis	No	45 (78,9%)	12 (21,1%)	0,793
	Yes	1 (100%)	0 (0%)
Immunoglobulins	No	38 (77,6%)	11 (22,4%)	0,668
	Yes	8 (88,9%)	1 (11,1%)
Anticonvulsants	No	8 (61,5%)	5 (38,5%)	0,116
	Yes	38 (84,4%)	7 (15,6%)
Sedation	No	23 (95,8%)	1 (4,2%)	0,008
	Yes	22 (66,7%)	11 (33,3%)
Vasoactive drugs	No	43 (84,3%)	8 (15,7%)	0,028
	Yes	3 (42,9%)	4 (57,1%)
Nasal cannula oxygen therapy	No	38 (80,9%)	9 (19,1%)	0,681
	Yes	8 (72,7%)	3 (27,3%)
Intermittent mandatory ventilation	No	20 (100%)	0 (0%)	0,005
	Yes	26 (68,4%)	12 (31,6%)

Data presented as n (%). Percentages are row percentages for each intervention. Fisher’s exact test was used for all statistical comparisons

IVIG Intravenous Immunoglobulins

Bold *p*-values indicate statistical significance (*p* < .05)

Discussion

Epidemiological data

Acute encephalitis is more common in children, with annual incidences ranging from 4 to 12.6 per 100,000 in high-income nations [8, 10–15], and significantly higher burdens in endemic regions such as Southeast Asia [2, 5]. Rates vary geographically, from 5.9 per 100,000 in England [6] to 13.8 in Greece [16], with peaks consistently observed in infants [17]. Seasonality depends on pathogen type and climate: temperate regions often peak in cooler months [18–20], while summer peaks occur where arthropod-borne or enteroviral infections predominate [21]. In our cohort, 57.2% of cases occurred in spring and summer, consistent with patterns reported in regions where enteroviral and vector-borne infections are prevalent, despite the lack of systematic pathogen identification in all cases.

It is important to note that our study cannot provide true population-based incidence rates, as we lacked a defined catchment population denominator. The 3.4 cases per year reported represent the average annual case count diagnosed within our single tertiary PICU, not the incidence in the general pediatric population.

Incidence typically peaks in children aged 5–6 years [22–24], with an unexplained male predominance [25–30] also reflected in our cohort (55.2%). This finding aligns with previous pediatric series, although the underlying biological or behavioural explanations remain unclear.

Key risk factors include rural residence, due to increased zoonotic and vector exposure [31], and low vaccination coverage. In Tunisia, the absence of national surveillance likely leads to underestimation of local incidence.

Clinical profile

Prodromal symptoms often precede pediatric acute encephalitis. Common features include febrile illness (50–75%), gastrointestinal symptoms, and respiratory manifestations [25, 32]. Preceding viral infection occurs in 55–86% of juvenile ADEM cases, while vaccination precedes 4–18% [30, 33]; in anti-NMDAR encephalitis, 70% report prodromal gastrointestinal or respiratory symptoms [34]. In our cohort, 22.4% had recent flu-like symptoms, with a mean interval from prodrome to neurological presentation of 6.4 days (SD 6.23) [32]. The lower frequency observed may reflect underreporting in medical records rather than true epidemiological variation, given the retrospective design. The classic triad of fever, altered consciousness, and seizures was common [26]. Altered mental status occurred in 69% of patients, consistent with pediatric series [35]. Seizures were present in 82.8%, exceeding the typical reported range of 14–81% [9, 21, 25, 36–38], and are frequently seen in Japanese encephalitis (7–67%), herpes simplex encephalitis (33–50%), and ADEM (13–35%) [39, 40]. The high seizure frequency likely reflects PICU-based recruitment, which selects for more severe neurological presentations.

Behavioral disturbances (5.2% in our cohort) are prominent in anti-NMDAR encephalitis [34], while irritability occurs in 7–44% of acute disseminated encephalomyelitis (ADEM) cases [41]. Abnormal movements such as chorea and dystonia, observed in 10.3% of our patients, are consistent with autoimmune etiologies [42] and previously reported Tunisian data [43]. Speech disorders (3.4%), including mutism and echolalia, were noted; aphasia is more common in children [44] and occurs in over 50% of pediatric anti-NMDAR cases [45].

Focal neurological deficits were present in 5.2% of our patients [46], lower than the 33–36% reported in other studies [44, 47]. Fever was frequent (82.8%), within the 64–90% range described in the literature [16, 21, 26, 28, 35]. Meningeal signs (17.2%) matched reported frequencies of 22–31% [42, 45]. Cranial nerve involvement occurred in 6.9% of cases, compared with 18–39% in ADEM [41], most often manifesting as oculomotor palsy [48].

Overall, our clinical profile confirms the heterogeneity of pediatric encephalitis presentations, with a predominance of severe, seizure-dominant forms likely influenced by tertiary-care referral bias.

Paraclinical study

Paraclinical diagnosis relies on CSF analysis, neuroimaging, and EEG, with timing and repetition critical for detecting evolving abnormalities. Standard labs may show hyperleukocytosis (51.7%) or leukopenia (3.4%) and elevated CRP (32.7%)—similar to previous reports [47]. Lumbar puncture with CSF analysis is essential and may be repeated if symptoms persist [15]. In viral encephalitis, CSF typically shows moderate lymphocytic-predominant pleocytosis (25.8% in our series), though neutrophilic predominance can occur early [20]. White blood cell counts are usually < 100 cells/µL [41, 49]; elevated RBCs (> 500/µL)—suggestive of necrotizing processes such as HSE [34, 50]—occurred in 6.9% of our cases. Protein levels are typically normal or mildly elevated (< 1 g/L) [51]; our median CSF protein was 0.38 g/L. Age-specific reference values for CSF protein and pleocytosis in infants have been established [52–54]. Hypoglycorrhachia, rare in viral encephalitis but characteristic of bacterial, fungal, or tuberculous etiologies [55], was noted in 5.2% of patients.

Overall, CSF findings were consistent with prior literature, although microbiological confirmation was limited due to the retrospective study design.

Microbiological workup includes serology (ELISA for arboviruses) and PCR. CSF HSV PCR, the diagnostic standard for HSE [56], was positive in 10.3% of our cohort. PCR for Mycoplasma pneumoniae, respiratory viruses, and enteroviruses from extra-CSF samples can aid diagnosis. Blood cultures remain adjunctive [1]. Autoimmune markers—such as anti-NMDAR antibodies in CSF, which correlate with disease activity and may signal underlying tumors [57]—are increasingly recognized. Anti-MOG antibodies are often detected in acute ADEM but may wane [58]; other myelin-targeting antibodies exist [59]. Biopsy was not performed in our series.

The diagnostic yield of etiological testing was influenced by non-systematic testing and resource availability.

Neuroimaging is central: CT, performed in 50 patients, was abnormal in 23 (46.0%). MRI, performed in 46 patients, was abnormal in 35 (76.1%), confirming its superior sensitivity [60]. The following imaging findings (Figs. 5, 6, 7 and 8) are provided as illustrative examples from individual patients and are not aggregate cohort data. In HSE, diffusion-weighted sequences show bilateral temporal lobe involvement [61] (Fig. 5). ADEM typically exhibits multiple T2/FLAIR hyperintense subcortical white matter lesions, often asymmetric and non-enhancing [9] (Fig. 6). In AIE, hippocampal FLAIR abnormalities may be suggestive [34, 50] (Fig. 7).

Fig. 5.

Imaging findings in a 7-month-old infant with confirmed Herpes Simplex Encephalitis (HSE) (illustrative case from the cohort; not aggregate data): a Initial Non-contrast CT: Demonstrates asymmetric cortical involvement with prominent hypodensity in the left temporal lobe and subtle, mild hypodensity in the right temporal lobe. A focal area of hyperattenuation in the left frontal operculum (arrow) is consistent with a petechial hemorrhagic focus, a characteristic finding in HSE. b Axial FLAIR MRI: Acquired at presentation, showing bilateral, asymmetric hyperintense signal within the temporal lobes (left > right), indicating cytotoxic edema and inflammatory changes. c T1-weighted Post-Contrast MRI: Reveals diffuse leptomeningeal enhancement over the cortical surfaces, reflecting active meningeal inflammation. d Delayed Contrast-Enhanced CT: Follow-up imaging illustrates the progression of the disease, showing extensive encephalomalacic changes (tissue loss) and profound volume loss in the bilateral temporal and frontal lobes, signifying the chronic sequelae of the infection

Fig. 6.

Axial MRI in FLAIR sequences in a child with acute disseminated encephalomyelitis revealing multiple patchy, poorly demarcated hyperintense lesions primarily localized within the subcortical white matter. These lesions are characteristically large, asymmetric, and "fluffy" in appearance, often involving the deep gray matter such as the thalami and basal ganglia, which is a key feature distinguishing ADEM from multiple sclerosis. (illustrative case from the cohort; not aggregate data)

Fig. 7.

Coronal FLAIR MRI typically revealing, in case of autoimmune encephalitis, asymmetric, hyperintense signal and subtle swelling within the mesial temporal lobe, most notably the hippocampus (red arrow) and amygdala. These high-signal intensities represent localized edema and inflammatory infiltration of the limbic system, which can be triggered by either autoimmune or paraneoplastic processes. (illustrative case from the cohort; not aggregate data)

Fig. 8.

The initial EEG demonstrates a significant hemispheric asymmetry, characterized by voltage attenuation and a loss of fast-frequency background activity over the left hemisphere (black brackets). These findings indicate focal cortical dysfunction and neuronal injury, correlating with the localized inflammation and edema seen on neuroimaging. In a pediatric context, such a marked reduction in amplitude and frequency is a key indicator of the primary site of neurological insult. (illustrative case from the cohort; not aggregate data)

EEG, performed in 39 patients (67.2% of the cohort), was abnormal in 35 (89.7% of those tested). The most frequent pattern was background slowing, present in 70% of abnormal EEGs, followed by periodic lateralized epileptiform discharges (PLEDs) in 10.6%. PLEDs are highly suggestive of necrotizing encephalitis, such as herpes simplex virus (HSV) infection, a finding consistent with the 11.7% HSV rate in our cohort.

Overall, ancillary investigations were consistent with classical encephalitis patterns; however, their diagnostic utility was limited by incomplete testing and the retrospective design of the study.

Etiologies

In our pediatric cohort, an etiology was identified in 81% of cases, which is at the higher end of the reported range of 27.5–79% [5]. This notably high etiological yield may reflect the use of a broad, though non-standardized, diagnostic approach. This included cerebrospinal fluid PCR testing (particularly for HSV and, when indicated, enteroviruses), serological testing, and autoantibody screening. Additionally, the high prevalence of vaccine-preventable encephalitides in our region during the study period likely contributed to the yield.

Caution regarding etiological yield. While our 81% etiological yield appears higher than many published series (27.5–79%), this figure requires careful interpretation. First, 12 of 13 bacterial cases (92.3%) lacked microbiological confirmation and were classified as “possible” based on clinical and radiological findings alone. Some of these cases may represent viral encephalitis with coincidental bacterial colonization, or secondary encephalopathy due to systemic infection without direct CNS invasion. Second, testing was non-standardized over the 17-year study period, with diagnostic investigations performed based on clinician judgment and resource availability rather than a fixed protocol. Third, the distinction between primary encephalitis and encephalopathy secondary to systemic infection (e.g., sepsis-associated encephalopathy) remains challenging in retrospective studies, and some “undetermined etiology” cases (n = 10) may represent secondary encephalopathies rather than true encephalitis. Therefore, our 81% yield should be viewed as an upper estimate, and readers should focus on the confirmed (certain) diagnoses (11 cases, 19.0% of the total cohort) as the most reliable findings.

The incidence of confirmed viral encephalitis in our pediatric cohort was 31%, consistent with ranges reported in other series (23–53%). The frequency and distribution of viral etiologies show significant geographical variation, influenced by climate, vector-borne disease patterns, and vaccination programs. While herpes simplex virus (HSV) is a predominant cause in the United States, Europe, and Australia [8, 62, 63], arboviruses such as Japanese encephalitis virus are leading causes in Asia [64], and enteroviruses are frequently reported in India, Greece, and China [65].

HSE is the most common necrotizing infectious encephalitis worldwide, accounting for 2–19% of cases in most series [8]. In our cohort, HSE was identified in 11.7% of patients. Although HSV PCR on CSF is the diagnostic standard, a negative early result does not exclude HSE, necessitating integration of clinical, EEG, and neuroimaging data, with consideration of repeat testing if suspicion remains high [66]. Enterovirus encephalitis, the second most common viral cause, often presents with gastrointestinal or exanthematous symptoms; characteristic MRI findings include T2/FLAIR hyperintensities in the brainstem and spinal cord [67], and diagnosis may require PCR testing of CSF, stool, and throat samples due to potential false-negative CSF results [17]. Varicella encephalitis is rare in immunocompetent children (< 5%) and typically manifests as cerebellitis, with MRI showing suggestive ischemic lesions [68]; diagnosis relies on serology and intrathecal antibody synthesis, though PCR and IgM can be negative. Rubella encephalitis, a rare complication (≈ 1/6,000 cases), saw a temporary increase in our region during an outbreak prior to widespread vaccination, consistent with observations that vaccination reduces its incidence [16]. Measles encephalitis, the most severe complication of measles, has declined significantly with vaccination; it includes acute post-eruptive, subacute inclusion, and subacute sclerosing panencephalitis forms, the latter being fatal and characterized by progressive neurological decline. COVID-19-associated encephalitis, recognised during the SARS-CoV-2 pandemic, poses diagnostic challenges in children due to often low or undetectable viral loads and antibodies in CSF [69]. Rarely, COVID-19 has been linked to acute necrotizing encephalopathy, particularly in pediatric patients with genetic predispositions such as RANBP2 mutations. Rabies encephalitis, almost invariably fatal, classically presents with hydrophobia and aerophobia; survival is exceptionally rare and its mechanisms remain uncertain [70, 71].

Bacterial encephalitis includes Mycoplasma pneumoniae, a leading etiology in Europe and North America [72, 73], diagnosed via combined serology and PCR; tuberculous meningoencephalitis, a severe form with high mortality and morbidity, often complicated by hydrocephalus [48, 74–76]; rickettsial encephalitis, suggested by systemic symptoms and skin findings [77]; and typhoid encephalitis, a rare complication in children in endemic regions, diagnosed primarily via culture [78].

Immune-mediated encephalitis includes ADEM, a demyelinating condition typically triggered by infection or vaccination, diagnosed using established pediatric criteria [9]; we recorded 13 post-viral and 2 post-vaccinal ADEM cases. AIE, such as anti-NMDAR encephalitis, is defined by specific neuronal antibodies and may be paraneoplastic, particularly associated with ovarian teratomas in older patients [57, 79]; seronegative cases are diagnosed using clinical criteria [80]. We observed one case of limbic encephalitis. Despite advances, a substantial proportion of pediatric encephalitis cases remain of undetermined etiology, ranging from 37% to 67% in the literature [5, 11, 17, 26, 81]; our proportion was lower, at 19%.

Treatment

In our cohort, empirical antimicrobial therapy was near-universal: acyclovir in 86.2% of patients and antibiotics in 87.9%, most commonly cefotaxime. These practices align with international guidelines for suspected viral and bacterial encephalitis [44, 82].

Supportive critical care was required in a substantial proportion of our severe cases: 65.5% received invasive mechanical ventilation, 56.9% sedation, and 12.1% vasoactive drugs. The need for advanced resuscitation measures was strongly associated with mortality (p < 0.001), highlighting the severity of illness in this cohort.

For immune-mediated encephalitis (27.6% of our cohort), first-line immunomodulatory therapy consisted of high-dose corticosteroids (44.8%) and intravenous immunoglobulins (15.5%). These rates are consistent with published pediatric protocols for ADEM and anti-NMDAR encephalitis [41, 83].

Long-term rehabilitation and seizure management are essential for survivors, though post-encephalitic epilepsy protocols remain individualized [28, 84].

Clinical course

The clinical course of pediatric acute encephalitis ranges from full recovery to severe complications and lasting neurological sequelae, with outcomes heavily influenced by initial disease severity and complications. The length of hospital stay varies, with reported averages of 16–25 days in the literature [85, 86]; in our cohort, the median ICU stay was 6 days. Short-term outcomes can be favorable, with clinical improvement often noted within days of initiating treatment and complete recovery within weeks in many cases [87]. Prognosis is generally good in pediatric AIE and ADEM with early, aggressive immunotherapy, with up to 85% of ADEM cases recovering favorably [41, 87]. However, severe forms defined by refractory seizures, major cognitive impairment, or the need for intensive care carry a worse prognosis [88]. Refractory AIE is defined by a lack of response to second-line immunotherapy after 1–3 months [88]. Mortality in our series was 20.7%, higher than the typical range of 5–13% reported in pediatric studies [11, 52, 53, 89]. This elevated rate may reflect the severity of cases managed in a tertiary paediatric intensive care unit (PICU), where a high proportion of patients presented with coma and status epilepticus at admission. Mortality varied by aetiology: hepatic encephalopathy and rabies were associated with high fatality rates, whereas mortality in immune-mediated encephalitis was below 5% [2].

Long-term sequelae affect approximately half of pediatric survivors, including neurological deficits, cognitive impairment, neuropsychiatric disorders, and epilepsy, posing a substantial global health challenge [11, 17, 18, 21, 52, 54]. In our cohort, among the 46 survivors, 9 (19.6%) developed neurological sequelae. Acute necrotizing encephalopathy (ANE), a rare but severe complication of viral infections such as influenza and COVID-19, is associated with particularly high mortality and morbidity. Christou et al. recently described a pediatric case of COVID-19-related ANE with a RANBP2 mutation, underscoring the role of genetic susceptibility in severe viral encephalitis.

Relapses occur in 12–25% of severe AIE cases, warranting prolonged immunosuppression and surveillance [34, 57, 90], while ADEM, though typically monophasic, can recur, particularly if corticosteroid therapy is abbreviated [42]. HSE relapses, though less common, have also been documented, usually within the first month [47]. Residual deficits are diverse: motor deficits were noted in 3.4% of our patients and have been significantly associated with poor outcomes in other studies [47, 64, 81]. Cognitive impairment and delayed psychomotor development affect 20–30% and up to 35% of survivors, respectively [26]. Behavioral disorders and chronic headaches are also reported, though headache was less frequent in our series (4%) than in some literature (20%) [47]. Outcomes are often assessed using tools like the modified Rankin Scale (mRS), adapted for children, to grade disability [88].

Before interpreting the following associations, several statistical limitations must be emphasized. Given the small sample size (n = 58) and the performance of multiple univariate comparisons across multiple tables (including Tables 2–10), there is an increased risk of type I errors (false-positive findings). The significant p-values reported should be interpreted as exploratory and hypothesis-generating, not as definitive evidence of independent associations. Furthermore, we have not adjusted for multiple comparisons (e.g., using Bonferroni correction), which would have rendered all associations non-significant given the small sample. Readers are cautioned against overinterpreting individual p-values.

Predicting long-term outcomes in pediatric patients remains challenging, as acute clinical presentation often correlates poorly with subsequent sequelae [91]. In our univariate analysis, age under one year was not significantly associated with neurological sequelae (p = 0.393). However, due to the limited sample size, we could not adjust for potential confounders such as illness severity or etiology; therefore, these findings should be considered hypothesis-generating rather than confirmatory of independent predictors. Although seizures during the acute phase are known to increase the risk of subsequent epilepsy [28], we observed no statistically significant association in our cohort. Similarly, a low Glasgow Coma Scale (GCS) score—a validated prognostic indicator linked to poorer recovery [91]—did not correlate with sequelae. While focal neurological deficits at presentation have been identified as factors associated with persistent impairment in other studies [26], only abnormal movements showed a statistically significant association with sequelae in our univariate analysis. Furthermore, no significant association between initial GCS and mortality was identified, despite its well-established prognostic value in the literature. Paraclinical findings such as neuroimaging abnormalities, particularly focal MRI hyperintensities, have been linked to worse outcomes in other series; however, we found no association between MRI or CT abnormalities and sequelae in our cohort.

Likewise, although specific EEG patterns (e.g., focal slowing or multifocal spikes) may indicate a higher risk of sequelae, our findings align with studies reporting no significant correlation. Biological markers like leukopenia have been predictive in some research [47], but in our analysis, neither leukocytosis, elevated CRP, nor infectious etiology were associated with sequelae. The intensity of supportive care, including ICU admission, is often a marker of severity and has been correlated with poorer outcomes [28, 57]; however, while we found no association between ICU admission itself and sequelae, the need for supplemental oxygen via nasal cannula was significantly associated with the development of neurological sequelae, highlighting it as a potential marker of disease severity and hypoxic injury impacting long-term recovery.

Given the exploratory nature of all univariate associations reported in this study, readers should not derive clinical action items from individual p-values. The associations presented here are intended to generate hypotheses for future research, not to guide patient management.

Factors associated with mortality

Numerous factors are associated with mortality in pediatric acute encephalitis. Young age, particularly under one year, is a well-established risk factor, with studies indicating infants have up to a fivefold higher risk of death compared to older children [46]; in our cohort, age under one year showed a significant univariate association with mortality (p = 0.015). Clinical signs also hold prognostic value: seizures during the acute phase have been linked to mortality in other studies, and in our series, on univariate analysis, the presence of seizures was significantly associated with death (p = 0.003). Additionally, a low GCS score (< 8) has been identified as a risk factor for fatal outcomes in both infectious and immune-mediated encephalitis [25].

Paraclinical findings contribute to risk stratification. Neuroimaging abnormalities, such as necrotic lesions on MRI [47] or an abnormal brain CT scan—as observed in our study—are associated with increased mortality. In contrast, EEG abnormalities did not show a significant association with mortality in our analysis, consistent with some literature. Certain biological markers, including leukopenia and hyperleukocytosis, have been identified as risk factors in viral encephalitis, though no such associations were found in our series.

Etiology influences prognosis, with infectious encephalitis identified as a mortality risk factor in some studies [28]; however, we found no significant association between etiological classification and death. The intensity of supportive care reflects, rather than causes, disease severity. In our cohort, patients who died were more likely to have received advanced resuscitation measures (sedation, vasoactive agents, mechanical ventilation). However, these interventions were administered in response to clinical deterioration (e.g., respiratory failure, shock, refractory status epilepticus) and should not be interpreted as independent risk factors for mortality. Rather, they are markers of underlying illness severity. The strong univariate associations observed (p < 0.001 for mechanical ventilation) likely reflect the critical condition of non-survivors rather than a causal effect of the interventions themselves. While the length of ICU stay has been correlated with death in other studies, early and specific interventions can alter outcomes, as evidenced by the reduction in HSE mortality from approximately 70% to 20% with timely acyclovir treatment [92].

Prevention

Preventive strategies are crucial in reducing the burden of acute encephalitis. The epidemiological landscape of viral encephalitis has shifted with widespread vaccination, making HSV a leading cause in children, while vaccine-preventable forms such as mumps, measles, rubella, Japanese encephalitis, and rabies have declined [93]. Effective immunization remains the cornerstone of prevention for these specific etiologies. For arboviral encephalitides, including Japanese encephalitis and West Nile virus (WNV) encephalitis, vector control and environmental sanitation are essential to prevent outbreaks. Given the uniformly fatal course of rabies encephalitis, prevention in endemic regions is paramount and involves animal vaccination and timely post-exposure prophylaxis for humans, which is only effective if administered during the incubation period [70]. In Tunisia, rabies prophylaxis is provided free of charge as part of the national rabies control program.

Clinical implications

The following clinical implications should be considered hypothesis-generating and require validation in larger prospective studies before being incorporated into clinical guidelines. They are derived from exploratory univariate associations and should not be interpreted as practice-changing based on this dataset alone.

The findings of this 17-year cohort study have several practical implications for clinicians managing children with suspected acute encephalitis, particularly in resource-limited settings:

First, the high etiological yield (81%) observed in our cohort, although requiring confirmation in prospective studies with standardized testing protocols, raises the hypothesis that investment in basic neurodiagnostic capacity (including CSF PCR, serology, and autoantibody testing) could improve diagnostic certainty in similar resource-limited settings. However, this finding may reflect testing biases and the inclusion of ‘possible’ bacterial cases without microbiological confirmation.

Second, the univariate association between abnormal movements and neurological sequelae (p = 0.010) is exploratory and hypothesis-generating only. If confirmed in larger prospective cohorts with adjustment for confounders, this finding would suggest that children presenting with chorea, dystonia, or tremor may warrant closer follow-up and earlier consideration of immunomodulatory therapy. At present, no clinical recommendation can be made based on this single univariate finding.

Third, the univariate association between age under one year and higher mortality (p = 0.015), if confirmed in larger studies, would suggest that infants may represent a high-risk subgroup warranting consideration for aggressive triage to tertiary centers with neurocritical care capabilities. However, given the small sample size and lack of adjustment for confounders, this finding should not be interpreted as a definitive indication for triage protocols. The association between invasive mechanical ventilation and mortality reflects disease severity rather than a direct effect of ventilation.

Fourth, the counterintuitive univariate association between seizures at presentation and lower mortality (p = 0.003) should not be interpreted as protective or as having any clinical implication. This finding is highly likely to be due to unmeasured confounding (e.g., earlier care-seeking, different etiologies, or survivor bias) and requires further study in much larger cohorts. Clinicians should continue to treat seizures aggressively regardless of this observation, which should be considered a statistical artifact until confirmed.

Finally, the observed neurological sequelae rate among survivors with follow-up (19.6%, with a plausible range of 19.6–43.5% due to loss to follow-up) underscores the potential need for structured long-term follow-up, including neurodevelopmental assessment, rehabilitation services, and monitoring for post-encephalitic epilepsy. However, given the incomplete follow-up (19% lost) and the lack of validated outcome scales, this estimate should be interpreted with caution and confirmed in prospective studies with complete follow-up.

Strengths, limitations and future directions

This 17-year retrospective cohort study has several strengths, including a long observation period, the use of standardized international diagnostic criteria, and a comprehensive multidisciplinary assessment of clinical, laboratory, imaging, and outcome data within a single tertiary PICU in southeast Tunisia, offering insights from an underrepresented region.

However, several limitations must be acknowledged. First, the retrospective, single-centre design introduces inherent biases and limits generalizability to other populations and healthcare settings.

Second, and most importantly regarding our analytical approach, all statistical comparisons were univariate. Multivariate logistic regression was not performed due to the small sample size (n = 58) and low number of outcome events (12 deaths, 9 with neurological sequelae), which would have risked model overfitting and produced unreliable estimates. Consequently, the associations we report — identified by p-values < 0.05 — are unadjusted for potential confounders such as age, disease severity at presentation, etiological category, comorbidities, and timing of treatment initiation. We have therefore deliberately avoided the term “predictor” throughout this manuscript, instead describing our findings as “univariate associations” or “factors associated on univariate analysis.” Readers should interpret these findings as exploratory and hypothesis-generating, requiring confirmation in larger, prospective cohorts that can adequately adjust for confounding. Consequently, no clinical recommendation or practice change should be derived from the univariate associations reported in this study. These findings serve only to generate hypotheses for future prospective research.

Third, a substantial proportion of patients (11/58, 19.0%) were lost to follow-up, which materially affects the precision of our sequelae estimates. The true rate of neurological sequelae among survivors could range from 19.6% (assuming no sequelae in lost patients) to 43.5% (assuming all lost patients had sequelae). Moreover, outcome assessment was based on predefined clinical criteria rather than validated standardized scales (PCPC, GOS), limiting comparability with other studies and potentially introducing assessment bias.

Fourth, despite extensive diagnostic investigations, unknown etiologies remained common, and distinguishing primary encephalitis from encephalopathy secondary to systemic infections (e.g., sepsis-associated encephalopathy) was challenging in the absence of standardized testing protocols. The lack of systematic diagnostic investigations in all patients may have led to misclassification of some “unknown etiology” cases, potentially affecting the accuracy of etiological distribution and outcome analysis.

Finally, given the retrospective and observational design, the observed associations should not be interpreted as causal.

Fifth, our reported ‘3.4 cases per year’ represents the average annual number of encephalitis cases diagnosed in a single tertiary PICU, not a true population-based incidence rate. Without a defined catchment population denominator, we cannot calculate incidence in the epidemiologic sense (cases per 100,000 person-years). Readers should interpret these figures as descriptive case frequencies within our PICU cohort, not as population incidence estimates.

Sixth, our statistical approach has several important limitations beyond the absence of multivariate regression. First, the performance of multiple univariate comparisons across numerous variables (Tables 2–10) increases the risk of type I errors (false-positive findings). We did not adjust for multiple comparisons (e.g., Bonferroni correction), as such adjustment would have rendered all associations non-significant given the small sample size. Therefore, all reported p-values should be interpreted as exploratory. Second, we did not calculate effect sizes (odds ratios or risk ratios) for non-significant associations, and for significant associations, the wide confidence intervals (e.g., OR for abnormal movements: 1.48–27.11) reflect the imprecision inherent in small-sample univariate analyses. Third, several variables showing significant associations with mortality (sedation, vasoactive use, mechanical ventilation) are treatment responses to severe illness rather than baseline prognostic factors; these associations should not be misinterpreted as causal or as independent predictors.

Seventh, case ascertainment and etiological classification are vulnerable to misclassification. Over the 17-year study period, diagnostic testing was non-standardized and performed according to clinical indication and resource availability rather than a fixed protocol. Among the 13 cases classified as bacterial encephalitis, only 1 had microbiological confirmation (CSF culture); the remaining 12 were diagnosed based on clinical, CSF, and radiological findings (“possible” certainty). Some of these cases may represent viral encephalitis or secondary encephalopathy due to systemic infection. Furthermore, distinguishing primary encephalitis from encephalopathy secondary to systemic infection (e.g., sepsis-associated encephalopathy) was challenging, and some “undetermined etiology” cases may reflect secondary processes rather than true encephalitis. The 81% etiological yield should therefore be interpreted as an upper estimate, with the 19.0% certain diagnoses (11/58) representing the most reliable findings.

Future research should prioritize prospective, multicenter collaborations to enhance generalizability and statistical power. Incorporating advanced diagnostic techniques, such as metagenomic sequencing, will help reduce uncertainty regarding the underlying causes of pediatric encephalitis. Additionally, structured long-term neurodevelopmental follow-up is essential to better understand and improve lifelong outcomes for survivors.

Conclusion

Acute encephalitis in children is a severe neurological emergency with significant morbidity and mortality, requiring prompt recognition in any child presenting with encephalopathy and signs of cerebral dysfunction. In this 17-year retrospective cohort study from a tertiary pediatric intensive care unit in Tunisia, acute encephalitis accounted for 2.6% of all PICU admissions, with an average of 3.4 cases diagnosed per year in this tertiary PICU setting. The predominant clinical features were seizures (82.8%), altered consciousness (69%), and fever (82.8%). Etiological investigations identified a cause in 81% of cases, most commonly infectious. Despite aggressive empirical antimicrobial therapy and intensive supportive care, mortality was 20.7% (12/58), and among the 46 survivors, 9 (19.6%) experienced neurological sequelae (though this figure may be an underestimate due to 19% loss to follow-up; the true rate could be as high as 43.5%), while infant age under one year, admission seizures, abnormal neuroimaging, and requirement for advanced resuscitation were significantly associated with mortality. In exploratory univariate analysis (with the inherent limitations of small sample size, multiple comparisons, and lack of adjustment for confounders), several associations reached nominal statistical significance. However, none of these findings should be interpreted as clinical recommendations or practice-changing inferences based on this dataset alone. They represent hypothesis-generating observations that require confirmation in larger, prospective, appropriately adjusted studies before any clinical application can be considered. Readers are cautioned against using these univariate associations to guide clinical decision-making. These findings highlight the importance of early recognition, comprehensive diagnostic evaluation—including advanced microbiologic and autoimmune testing—and vigilant neurocritical care to optimise outcomes. Neuroimaging remains essential for prognostication and may guide surgical intervention in cases complicated by refractory intracranial hypertension or hydrocephalus. Preventive strategies, particularly expanded immunisation against neurotropic pathogens, are crucial to reducing the global burden of paediatric encephalitis.

Abbreviations

ADEM

Acute Disseminated Encephalomyelitis

AEI

Acute Encephalitis of Immune origin

AEP

Auditory Evoked Potentials

AIE

Autoimmune Encephalitis

anti-GAD

Anti-Glutamic Acid Decarboxylase

anti-MOG

Anti-Myelin Oligodendrocyte Glycoprotein

anti-NMDAR

Anti–N-Methyl-D-Aspartate Receptor

anti-VGKC

Anti-Voltage-Gated Potassium Channel

BCG

Bacillus Calmette–Guérin

CMV

Cytomegalovirus

CRP

C-reactive Protein

CSF

Cerebrospinal Fluid

CT

Computed Tomography

EBV

Epstein–Barr Virus

EEG

Electroencephalogram

ELISA

Enzyme-Linked Immunosorbent Assay

GCS

Glasgow Coma Scale

GOS

Glasgow Outcome Scale

HSE

Herpes Simplex Encephalitis

HSV

Herpes Simplex Virus

ICU

Intensive Care Unit

IME

Immune-Mediated Encephalitis

MRI

Magnetic Resonance Imaging

mRS

modified Rankin Scale

PCPC

Pediatric Cerebral Performance Category

PCR

Polymerase Chain Reaction

PICU

Pediatric Intensive Care Unit

PLEDs

Periodic Lateralized Epileptiform Discharges

RBC

Red Blood Cells

SD

Standard Deviation

VEP

Visual Evoked Potentials

WBC

White Blood Cell

WNV

West Nile Virus

Authors’ contributions

Concept: A.E Design: A.E, M.B Resources: A.E Data Collection and/or Processing: F.K Analysis and/or Interpretation: H.B Literature Search: F.K Writing Manuscript: A.E Supervision: F.S , A.E Critical Review: A.E.

Funding

The authors received no specific funding for this work.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The requirement for ethics approval was waived by the Institutional Ethics Committee of Hédi Chaker University Hospital, Sfax, Tunisia, because this study was a retrospective chart review involving no patient intervention. The requirement for informed consent was also waived by the Institutional Ethics Committee of Hédi Chaker University Hospital, Sfax, Tunisia, due to the retrospective nature of the study. No reference number was issued. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.