CDARS is an electronic platform of the HA in which patients' clinical diagnoses, drug prescription details, laboratory investigation results, medical interventions, and length of hospital stay can be retrieved in an anonymous manner (15). Public hospitals under the HA manage 28,929 in-patient beds, which account for 85% of all in-patient admissions of HKSAR (16). We systematically collected the influenza hospitalisation, neurological symptoms and outcome of paediatric patients (<18 years) admitted to the paediatric departments of the 13 public hospitals in Hong Kong between January 2014 and December 2018.

Influenza cases were defined by a positive result confirming influenza A or influenza B by either influenza immunofluorescence assay or reverse transcription real time polymerase chain reaction of nasal pharyngeal aspirates. The proportion of influenza hospitalisations with respect to overall paediatric admission during the study period was also calculated.

The IANCs included in this study:

Influenza patients with any of the above conditions with an onset of neurological symptoms within 7 days of infection were defined as having IANCs. We then identified these patients via retrieval of the respective diagnostic codes from CDARS (Supplementary Table 1).

Data including the age, sex, type of influenza infection, length of stay, intensive care admissions, diagnoses, neurological complications, use of mechanical ventilation, presence of coinfection with bacteria or other viruses, uses of antiviral therapy, presence of pre-existing neurological or neurodevelopmental disorder and survival outcome of each patient was systematically collected from CDARS (Supplementary Table 2). The percentage of neurological complications and the annual incidence rate of influenza, as well as of IANCs and IAEs among different influenza subtypes and age groups (0–<2, 2–<6, 6–<12, and 12–<18 years) were calculated using age-specific population data provided by the Hong Kong Census and Statistics Department (Supplementary Table 3). The incidence rate of hospitalisation was weighted by 0.85 to reflect the proportion of patients served by public hospitals in Hong Kong (Supplementary Table 4).

To examine the clinical burden of IANCs, we compared the in-patient length of stay, need for PICU admission, use of mechanical ventilation, and the mortality rate, between influenza patients with neurological complications and those without. The vaccination coverage rate of seasonal influenza during the study period was extracted from the Centre for Health Protection. This study was approved by the Institutional Review Board of the University of Hong Kong (UW19-308).

Data analysis was performed using the Statistical Package for Social Sciences (IBS SPSS Statistics 25 Inc., Somers, NY). Length of stay of patients with IANCs and influenza patients without neurological complications were compared using the unpaired t-test. PICU admission, mechanical ventilator use, and death rates were compared using Chi-square tests as appropriate. A 95% confidence interval was calculated for the incidence rate. All analyses were two-tailed. The level of significance was set at p ≤ 0.05.

A total of 28,016 admissions due to seasonal influenza A or B amongst patients aged <18 years of age were identified from CDARS within the 5-year study period (2014–2018), accounting for 5.7% (28,016/489,955) of total paediatric admissions. Amongst these patients, 67.3% (18,841) had influenza A and 32.7% (9,175) had influenza B. Just over half of these patients (53.9%; 15,100) were male. The median age for hospitalisation due to influenza was 3.3 years for influenza A and 5.0 years for influenza B (Table 1).

Two thousand four hundred eighty-three patients with IANCs were identified, giving an overall IANC incidence rate of 8.9% (Table 1). 85.5% of patients with IANCs occurred in children <6 years of age. The median age of patients with IANCs was 3.0 years. Two-thirds of the patients with IANCs were male (64.9%) (Table 1). The most common IANC was febrile seizures, which accounted for 80.6% of all IANC cases, followed by myositis (11.4%) and seizures with fever (5.4%). 2.6% (64/2,483) of patients had IAE. One patient with influenza A had GBS (Table 1).

The mean annual incidence of IANCs among children <18 years of age was 57 per 100,000 population (Supplementary Table 3). Between the different age groups, the mean annual incidence was highest in children aged <2 years (149 per 100,000 population) and lowest in children aged 12– <18 years (2 per 100,000 population). The overall mean incidence of IANCs was significantly higher in influenza A at 42 per 100,000 children compared to 15 per 100,000 children in influenza B (Supplementary Table 3). The mean annual incidence of IAE for all children (<18 years old) was 14.8 per 1,000,000 population and was highest in children aged <2 years (22.4 per 1,000,000 population).

Although the influenza hospitalisation rate increased from 46 to 83 per 10,000 children (<18 years old) from 2014 to 2018 (Figure 1A), annual IANC incidence amongst paediatric in-patients with influenza A or B decreased from 11.5% in 2014 to 6.7% in 2018 (Table 1). This decline in the annual IANC percentage was observed in all paediatric age groups (Figure 1B).

Benign febrile seizures were the most common neurological complication in both influenza A and influenza B (Table 1). Overall, patients with influenza A had a higher percentage of febrile seizures compared to influenza B (8.4 vs. 4.6%, p < 0.001). The percentages of seizure with fever and IAE were similar between influenza A and influenza B patients (0.5 vs. 0.5%; 0.2 vs. 0.2%, p > 0.05). On the other hand, patients with influenza B showed a higher percentage of myositis than influenza A (2.0 vs. 0.6%, p < 0.001) (Supplementary Figure 1A). In the younger age groups (0– <2 years old & 2– <6 years old), the incidence of IANCs in influenza A patients was higher than that of influenza B (10.0 vs. 6.7%: 12.2 vs. 9.4%, p < 0.001). In the older age groups, however, influenza B patients showed a higher incidence of IANCs compared to influenza A (6.1 vs. 4.3% in 6– <12 years old, p < 0.01; 1.6 vs. 2.1 % in 12– <18 years old, p > 0.05) (Supplementary Figure 1B).

Paediatric patients with IANCs had significant longer hospital stays (p < 0.001), higher percentages of mechanical ventilation use (p < 0.05) and PICU admissions (p < 0.001), and higher mortality rates (p < 0.001) compared to those without neurological complications (Table 2).

Overall, IAE had the greatest clinical burden among IANC group. Patients with IAE had the highest median length of stay (6 days) compared to the other IANCs (2 days). IAE also accounted for 75% (9/12) of mechanical ventilation use and 63% (41/65) of PICU admission within the IANC group. Amongst those with IANCs, IAE was the sole cause of all seven reported mortalities. The mortality rate of IAE was 10.9% (7/64), and all deaths occurred amongst those <7 years old.

Other reasons of PICU admission for IANC patients included respiratory distress/respiratory failure (29%; 19/65), status epilepticus (SE) either from the benign febrile seizures or seizures with fever groups (6%; 4/65) and GBS (1/65).

Increase use of antiviral therapy (oseltamivir) was observed in all paediatric age group (Figure 1C). The use of antiviral therapy (osteltamivir) in patients with neurological complications was also found higher than those without neurological complications (36.9 vs. 28.3%, p < 0.001) (Table 2). Eighty percent (51/64) of the IAE patients were treated with antiviral therapy.

Only 2.46% (61/2,483) of paediatric influenza patients with neurological complications carried a pre-existing neurological or neurodevelopmental disorder. These include epilepsy, autism, and developmental delays.

We observed a spectrum of IANCs in the paediatric populations. The IANC that most commonly led to hospitalisation was febrile seizures (80.6% of IANCs). Only 1.1% of the febrile seizure group required PICU admission, and there was no associated mortality. As explained by Prerna et al. (20), febrile seizures are a major complication in influenza infections, however most patients recover well.

The proportion of febrile seizure cases admitted in our cohort was much higher compared to that of western countries. From a study in United States, only 37.5% (27/72) of patients with IANC admitted were due to febrile seizure (1). Additionally, another study from Australia showed that only 26% (14/54) of IANC are due to febrile seizure (14). This may be due to the difference in the clinical practise managing febrile seizure in our locality compared to western countries. According to the American Academy of Pediatrics, Royal College of Physicians, and the British Paediatric Association, simple febrile seizures that last <15 min and do not recur within 24 h usually do not require further evaluation in hospital (21, 22). In our study, the higher percentage of IANC admissions related to benign febrile seizures could reflect the parental anxiety, which could be attributed to the excessive media reports on the few annual influenza-related deaths (23). Moreover, when the children developed febrile seizures at home, many parents did not consult a family doctor and instead directly attended the emergency department. The lack of primary care physicians or family doctors in Hong Kong leads to the difficulty to set up the gatekeeping roles from primary care providers (24, 25). This may explain the higher percentage of IANC admissions related to febrile seizure in Hong Kong. Currently the Hong Kong College of Family Physicians is working closely with the government to increase the opportunity of family doctors' training. Further promotion of family medicine careers, as well as more public education on the benign nature and management of febrile seizures, would help to utilise medical resources more effectively and lower the healthcare burden of IANC-related febrile seizure admissions.

The second most common IANC leading to hospital admissions in our cohort was myositis (11.4%). We found that myositis was more prevalent in influenza B than influenza A infection, which confirms previous literature findings (26–28). We also found a single case of GBS after an influenza A infection. GBS is a rare neurologic complication and it is thus difficult to estimate the incidence rate (29). Similar to Glaser et al. (30), we could only find one case from this population-based cohort (n = 2,069).

Only a small percentage of patients with IANCs in our cohort carried pre-existing neurological diseases or neurodevelopmental disorders (2.5%), unlike other IANC studies in Italy and Taiwan (31, 32) (33 and 40%, respectively). The difference could be due to the high admission rate of benign febrile seizure in our studied population. We also recorded a low percentage (1.1%) of co-infection with other virus and bacteria in IANC cases. The pathogens included S.pyogenes, H.influenzae and Rotavirus which have been reported in previous studies (33, 34). No IAE patient in our cohort was found to have a co-infection.

The patients with IANCs had longer in-patient length of stays, higher rates of antiviral therapy use and mechanical ventilator use, and higher PICU admissions and influenza-related deaths when compared to other influenza in-patients without neurological complications. Our data supports previous findings that influenza in-patients with neurological complications have statistically significantly higher ICU admissions compared to influenza patients without neurological complications (18). Patients with IAE had the longest in-patient hospital stay, highest rates of antiviral therapy use, PICU admissions and mechanical ventilator use, and was the only cause of mortality, making it the most severe IANC. All seven cases of IANC-related deaths occurred within 13 days of influenza admission. All deceased patients were below 7 years old (Range: 1.4–6.9 years old). The high mortality rate of 10.9% (7/64) in our patients with IAE echoes the poor clinical outcomes found in previous studies in Japan and Australia, with mortality rates of 7.8% (22/283) and 15% (2/13), respectively (13, 14). Apart from IAE, other IANC cases had good recovery with no mortality despite PICU admissions.

Between 2014 and 2018, we observed an increased admission rate of paediatric patients with influenza A or B infections. This could be related to the increased availability of influenza diagnostic tests in both private family doctor clinics and public hospitals, as well as increased public awareness of the neurological risks of influenza, especially following local mass media coverage of the death of a young child due to acute necrotising encephalopathy (ANE) associated with influenza infection in 2016. As such, it is possible that more parents preferred early admission and closer observation for their child following an influenza diagnosis especially when presented with convulsion, thus explaining the increased rate of admissions (35).

Although the admission rate increased, there was a decline in the percentage of IANCs amongst paediatric in-patients. This coincided with the promotion of universal influenza vaccinations in Hong Kong around the same period. In 2014/15, the government introduced the Childhood Influenza Vaccination Subsidy Scheme to subsidise seasonal influenza vaccination (SIV) in children aged 6 months to <6 years and expanded to children aged between 6 and 12 years in 2016/17 (36). This scheme boosted the vaccination coverage rate of children aged 6 months to 12 years from 17.4% in 2015/16 to 45.8% in 2018/19 (37, 38) (Supplementary Figure 2). Seasonal influenza vaccinations (SIV) are associated with lower symptom severity, decreased ICU admissions, and reduced mortality rates in infected individuals (39–41). According to a US-based paediatric study (39), full influenza vaccinations could reduce three-quarters of PICU admissions. Another study from Flannery et al. also echoes the effectiveness of influenza vaccines by showing that influenza vaccine had 65% efficacy in reducing paediatric influenza-associated deaths. Therefore, the increase in the SIV rate could be one of the reasons for the decreased incidence of IANCs during the study period. Another possibility is the increased use of antiviral therapy (oseltamivir) in all age groups in our cohort over the years during the study period. Beside symptomatic treatment to maintain hydration, control fever and nasal congestion, oseltamivir would be recommended to patients with confirmed influenza A or B infection if the presentation was within 2 days of onset of symptoms or when severe influenza-related complications developed. All doses were given twice daily for 5 days. Possible side effects are explained to the parents. The antiviral drug would only be started with parent's consent. It is known that antiviral therapy usage is associated with the reduction of complications (42, 43).