Antivirals for childhood influenza are prescribed inconsistently, largely because the randomised evidence is about symptom duration rather than about the outcome anyone cares about. This multicentre case-control study addressed the outcome directly.
It included all paediatric outpatients with laboratory-confirmed influenza at participating Taiwanese hospitals between 2020 and 2023 — 1,492 children, mean age 7.1 years. Cases were the 354 children subsequently hospitalised or who died because of influenza; the 1,138 controls were matched by date of initial visit and age stratum. Early antiviral therapy meant treatment started within 48 hours of symptom onset, compared with delayed or no treatment.
Early treatment was associated with an 81% lower risk of hospitalisation (adjusted odds ratio 0.19, 95% CI 0.14 to 0.27). The effect was the same in children aged 5 or under (0.19, 0.12 to 0.31) and older children (0.15, 0.09 to 0.25), and held across sensitivity analyses by antiviral agent.
The practical implication is about the clock, not the drug. A parent who brings a febrile child on day one is in the window; one who waits for the fever to declare itself is not. That makes the intervention partly a communication problem — telling families with a child at risk what to watch for and when to come, before the season starts.
In Indian practice the limiting step is usually testing. Where laboratory confirmation takes longer than the 48-hour window, the decision has to be clinical, made on syndrome and local influenza activity rather than on a result.
- Start antivirals within 48 hours of symptom onset in a child with influenza who has any risk factor for severe disease
- Do not wait for laboratory confirmation if the result will arrive outside the treatment window — decide on syndrome and local activity
- Tell families of at-risk children before the season what symptoms should prompt same-day attendance
- The benefit was equal in under-fives and older children, so age alone should not decide who is treated
- This is observational: children treated early may differ systematically from those not treated
The statistics, in plain English
This is a case-control study, so it shows association, not proof of causation, and confounding by indication runs both ways here: sicker children may be treated faster, which would understate benefit, while better-resourced families may both attend early and avoid admission, which would overstate it. The authors addressed this with E-values, which ask how strong an unmeasured confounder would have to be to explain away the result. Their E-values of 9.5 to 19.5 mean an unmeasured factor would need to roughly ten-fold the odds of both early treatment and non-hospitalisation to account for the finding — implausibly strong, which is why this observational result deserves more weight than most.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for paediatrics, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free