Clinical and prognostic features among children with acute encephalitis syndrome in Nepal; a retrospective study

The classification of AES was based on the World Health Organization's (WHO) definition [1]. The results of the JE testing, undertaken as part of the JE surveillance programme, were related to the identified cases. The AES cases, based on the results of their microbiological and serological tests, were classified as AES of suspected viral aetiology (Confirmed JE, Non-JE and JE Status unknown) and AES of non-viral aetiology (AES-bacterial or parasitic aetiology). The clinical features within each class were examined. Cases definitions were as follows:

Outcome at discharge was classified as good or bad. Good outcome was defined as being alive with no impairment of consciousness or neurological sequelae. Bad outcome was defined by death or neurological sequelae at discharge. Patients that self-discharged or were referred to another hospital prior to discharge were excluded from outcome analysis.

Neurological sequelae were defined by the presence of one or more of the following at discharge; impaired consciousness, weakness (monoparesis hemiparesis, quadriparesis), focal or generalized abnormal limb tone (hypertonia, hypotonia), focal or generalized abnormal limb reflexes (hypereflexia, hyporeflexia), diagnosis of new onset or recurrent seizures or new or recurrent extra pyramidal movement disorders [5].

JE virus exposure was tested by MAC-ELISA (IgM antibody capture-Enzyme Linked Immunosorbent Assay). The ELISA plates were supplied by the Armed Forces Research Institute of Medical Sciences (AFRIMS), Bangkok, Thailand [12]. ELISA measurements were undertaken following the protocol supplied by AFRIMS. Diluted patient sera (1:100) or CSF (1:10) were added to the plates. Absorbance of experimental, positive and negative control samples were measured in duplicate in 96-well plates using a micro-titre plate reader (HumaReader Single Plus, Human GmbH, Wiesbaden, Germany). Single experimental patient samples with a mean absorbance ≥ 40 units (following subtraction of the absorbance for the negative controls) were considered positive. External quality assurance was undertaken by AFRIMS [12].

The acquired data for all AES patients was initially validated, coded and entered in SPSS Statistics software version 17.0 (IBM-SPSS, New York) for analysis. Differences between clinical groups were compared using Student independent samples t-tests (for Normally distributed data), Mann-Whitney U-tests (for non-Normally distributed data) and Fisher Exact tests (for categorical data/proportions). All clinical feature variables (with the exception of all duration and numbers of episodes variables apart from duration of fever) were entered into a forward stepwise multiple logistic regression model to identify variables independently associated with bad outcome or JE positive status; multiple imputation methods (10 iterations) were used to overcome problems of missing observations and collinearity statistics were examined to ensure independence of the predictor variables. The median age and numbers of AES cases were mapped using ArcGIS version 9.3 (Esri Ltd., California) software to show their distribution in the different districts across Nepal in order to identify any spatial patterns in infection dynamics. Statistical significance was set at the conventional 5% level for all analyses.

A total of 225 children with acute encephalitis syndrome (AES) were admitted to the two hospitals between January 1^st, 2006 and January 1^st, 2008. Forty (18%) of these patients did not have any lumbar puncture (LP) results available in their notes. These patients were analysed separately.

Of the remaining 185 AES patients, 40 (22%) were diagnosed with either bacterial (n = 39) or Plasmodium falciparum infection (n = 1). Eight of these patients also had elevated anti-JE virus titres (≥ 40 units) on serum IgM testing during acute illness. Since bacterial co-infection can change clinical features and influence patient outcome, these patients were analysed separately (Figure 1).

One hundred and forty five AES patients (145/185;78%) were considered to have AES of suspected viral aetiology. Forty-two of these patients (23%) were confirmed as JE. Among the other 103 patients, 44 tested negative to JE using samples collected after the ninth day of their illness and therefore were classified as Non-JE. For the remaining 59 patients JE status was unknown (Figure 1).

The majority of JE positive patients, 41/50 (82%), were identified by serum testing positive for anti-JEV IgM antibodies, with 31/41 (79%) identified by a single serum sample. The other serum samples tested positive in paired samples. The remaining 9 patients were diagnosed by testing positive for anti-JEV IgM antibodies in the CSF.

The majority of the non-viral AES patients were diagnosed with suspected bacterial meningitis based on a raised cell count with a polymorph predominance and raised protein in the CSF. The commonest identified cause of infection was gram stain positive bacteria in CSF.

Patients presented to Kanti Children's (n = 208) or Patan hospital (n = 17) from a wide range of outlying districts from Kathmandu including the hill and mountain districts. There were no marked differences in geographic distribution for the number of cases or age at presentation among the different sub-groups of suspected viral AES patients (Figure 2). Route of presentation to hospital was documented in 175 AES patients. Self-referral was the commonest route of presentation, reported by 96/175 (55%) of AES patients.

Outcome at discharge was recorded for 183/185 patients where LP results were available and 38/40 patients without LP results. Among the patients without LP results, seventeen (45%) had a bad outcome; 10 (26%) died and a further 7 (18%) had neurological sequelae at discharge.

Among the AES patients with non-viral infection, 38/40 (95%) had an outcome at discharge recorded. Five (13%) had a bad outcome; 2 died and a further 3 had neurological sequelae at discharge. All 8 patients with bacterial and JE co-infection exhibited complete recovery at discharge. The one patient with P.falciparum infection recovered completely.

Among the AES patients with suspected viral infection, 45/145 (31%) had a bad outcome; 8 died and 37 had neurological sequelae at discharge. Among the sub-set of confirmed JE patients 20/42 (48%) had a bad outcome; 4 died and 16 had neurological sequelae (Table 1).

A significantly higher proportion of patients without LP results died compared to those where LP results were available (10/38 [26%] versus 10/183 [5%]; P < 0.001). Similarly, the proportion of patients with bad outcome were higher among patients without LP results (17/38 [45%] versus 51/183 [28%]; P = 0.053). Exclusion of the patients without LP results may have influenced outcome within the AES groups. To help address this issue, an additional analysis was undertaken, whereby patients without LP results were classified into AES categories based on their hospital discharge diagnosis supported by their JE serological results. The patients contributed to all AES groups. Each AES category exhibited a higher rate of bad outcome. There were no significant changes in the proportion of patients with bad outcome between groups (Table 1).

A significantly higher proportion of patients with AES of suspected viral aetiology had a bad outcome compared to AES patients with a non-viral infection (45/145 [31%] versus 5/38 [13%]; P = 0.039). A significantly higher proportion of JE patients exhibited a bad outcome compared to AES patients of unknown viral aetiology (20/42 [48%] versus 25/103 [24%]; P = 0.01). A similar trend was observed when JE patients were compared to Non-JE patients, with a higher proportion of JE patients exhibiting a bad outcome (20/42 [48%] versus 12/44 [27%]); P = 0.07; Table 1).

Multiple parameters were associated with bad outcome at discharge for both AES cases of suspected viral aetiology (n = 145) and confirmed JE (n = 42) (Table 2). In both groups, bad outcome was associated with a longer duration of fever prior to admission, a lower Glasgow coma score, a focal neurological deficit, older patient age and higher weight at admission.

To identify whether these features were independently associated with bad outcome, these variables were entered into a forward stepwise logistic regression model separately for each patient group. For the larger group, all AES of suspected viral aetiology, fever duration (P = 0.002), GCS (P < 0.001), and focal neurological deficit (P = 0.001) were retained in the model for outcome. In JE, fever duration (p = 0.007) and GCS (p = 0.020) were again independently associated with bad outcome (but not focal neurological deficit) together with age (P = 0.011).

To assess whether length of time from onset of illness to hospital admission was a prognostic marker for bad outcome, we used length of reported fever prior to admission as a proxy marker. We selected a threshold of 7 days of fever prior to admission (Figure 3). We found fever duration over 7 days was associated with a bad outcome. Fever duration was reported in 142/145 (98%) of patients with AES of suspected viral aetiology. Thirty-two patients reported fever duration of more than 7 days prior to admission. Of these, 19 had bad outcome and 13 had good outcome. The remaining 110 patients reported a fever duration under 7 days, of which 23 had bad outcome and 87 had good outcome (19/32 [59%] versus 23/110 [20%]; P < 0.001). This indicates that AES patients who present with a prolonged fever prior to hospital admission have an increased risk of having a bad outcome at discharge (Relative Risk = 2.8 [95% confidence interval: 1.6 - 4.4]).

Among the JE patients, a reported fever duration of more than 7 days prior to admission was again linked to a significantly higher rate of bad outcome compared to those who presented with a shorter duration of fever (9/11 [82%] versus 11/31[36%]; P = 0.013). Among patients with AES of non-viral aetiology, prolonged duration of fever prior to admission was also linked to a higher rate of bad outcome (2/6 [33%] versus 3/22 [14%]; P = 0.28).

Of the treatments given to AES patients of suspected viral aetiology, mannitol and phenytoin were more frequently prescribed among those with a bad outcome compared to those with a good outcome; 27/61 (44%) versus 19/119 (16%), P < 0.001 for mannitol; 23/61 (38%) versus 12/119 (10%), P < 0.001 for phenytoin, in bad and good outcome patients respectively.

Confirmed JE cases were compared to AES cases of unknown viral aetiology (Non-JE and JE status unknown). As the latter group also contained patients that did not have a confirmed negative test for JE, a further analysis was undertaken comparing JE against Non-JE cases.

The rates of many clinical features and laboratory parameters were similar on admission between confirmed JE, confirmed Non-JE and JE-status unknown (Tables 3 and 4). Importantly, the rates of many recorded neurological features were comparable between the patient groups. There were no significant differences in prevalence of an altered sensorium at admission, depth or duration of coma, frequency, duration or type of seizures between patient groups. However, there was a significantly higher prevalence of patients presenting with focal neurological deficits at admission among confirmed JE patients compared to AES cases of unknown viral aetiology (13/40 versus 11/103; P = 0.005; Relative Risk = 3.0 [95% confidence interval: 1.4-6.7]). Among AES patients of suspected viral aetiology, presence of a focal neurological deficit at admission had a positive predictive value of 32% (sensitivity 54%) for JE. Absence of a focal neurological deficit at admission had a negative predictive value of 89% (specificity 77%) for the patient not having JE.

JE patients presented with a lower body weight (15.25 versus 20 Kg; P = 0.031), and a higher respiratory rate (30 versus 28 breaths per minute; P = 0.003) compared to confirmed non-JE patients. With the caveat that serum urea, creatinine and electrolytes were measured in only 25 AES patients of suspected viral aetiology, there was a trend for higher serum urea (15 versus 8.9 mmol/L; P = 0.27), sodium (137 versus 133 mmol/L; P = 0.09) and potassium (4.3 versus 3.9 mmol/L; P = 0.21) among JE compared to Non-JE patients (Table 4[13]). These parameters again showed a non-significant trend for higher median values among JE patients compared to AES patients of unknown viral aetiology (Table 4).

Both weight and respiratory rate, which were significantly different among JE patients compared to Non-JE patients, are in part dependent on patient age. To further dissect the influence of these parameters, age, weight and respiratory rate were entered into a stepwise logistic regression model. Only raised respiratory rate (P = 0.011) was retained.

Admission patterns also differed between JE and Non-JE patients. JE patients demonstrated a clear peak in hospital admissions rates in the months immediately following the rainy seasons each year (August and September in 2006-2007). Non-JE patients didn't demonstrate any clear seasonal variation in hospital admission rates (Figure 4).