Flu A vs Flu B: What the Difference Actually Means

The winter a Vienna ward counted both viruses

In the 2017/2018 influenza season, clinicians at a Vienna hospital swabbed every patient admitted to its flu isolation ward. Of 396 who tested positive, 96 had influenza A and 300 had influenza B. That ratio is unusual — most seasons run the other way — but it handed the team something rare: two viruses compared side by side, in the same building, in the same weeks, under the same protocols.

The differences they found were real, and small. Patients with influenza B were older, with a median age of 77 against 67.5 for influenza A. The A patients were more often smokers, 27.7% against 16.8%, and more often had chronic lung disease, 39.6% against 26.3%. They also arrived hotter, 38.6°C against 38.3°C, and their C-reactive protein — a marker of inflammation — ran at 41 mg/L against 23 mg/L.

The B group had its own signature. Chronic kidney disease was present in 37% of them against 18.8% of the A patients, and vomiting on admission was roughly twice as common.

None of that lets a doctor tell the two apart at the door. The study's own conclusion said so plainly: the symptoms overlap, and only a polymerase chain reaction (PCR) test settles which virus a patient has. Influenza A was the more severe disease in that ward that winter, with in-hospital mortality of 8.3% against 4.7%, but 22 people died in total and the authors were careful not to dress a single season up as a general law.

Two letters, two different viruses

Influenza A and influenza B are the two types that drive seasonal epidemics in people. They look similar under a microscope and cause a similar illness, which is why the distinction gets waved away as academic. It isn't, but the reason has less to do with how sick you feel than with where the virus comes from and how it changes.

Influenza A circulates in birds, pigs and other animals as well as humans. That reservoir is why A can shuffle its genes with animal strains and produce something the human immune system has never met — the 2009 H1N1 pandemic came out of that process. The A viruses in seasonal circulation now are A(H1N1)pdm09 and A(H3N2), and they are the reason flu vaccines get reformulated almost every year.

Influenza B has no animal reservoir worth worrying about. It circulates in people, splits into lineages, and drifts. Two lineages, B/Victoria and B/Yamagata, co-circulated for decades in proportions that were hard to predict, which is why quadrivalent vaccines with two B components were introduced in 2013.

The split in case numbers is roughly three to one. According to the European Medicines Agency, influenza A accounts for about 75% of seasonal epidemic cases and influenza B about 25%. In a bad B year, that quarter is enough to fill a hospital.

Photo by Logan Voss on Unsplash

The evidence on severity doesn't run one way

Start with the largest recent dataset. A Centers for Disease Control and Prevention (CDC) study published in The Lancet Microbe tracked 104,969 flu hospitalisations across nine seasons in the United States. More than one in six of those patients had a severe outcome. ICU admission occurred in 16.7%, and 6.5% needed invasive mechanical ventilation or extracorporeal membrane oxygenation. Death occurred in 3%.

Compared with A(H3N2), the strain most people think of as the nasty one, patients hospitalised with A(H1N1)pdm09 were 42% more likely to be admitted to intensive care, 79% more likely to need ventilation or ECMO, and 25% more likely to die. Patients with influenza B were 6% more likely to reach ICU, 14% more likely to need ventilation or ECMO, and 18% more likely to die. The CDC's own summary of the findings is blunt: the letter on the test does not tell you how sick a patient will get.

That lines up with earlier work. A CDC analysis of eight seasons from 2005/06 to 2012/13 counted 21,186 influenza A and 3,579 influenza B hospitalisations among adults and found intensive care use, length of stay and death essentially comparable; 3% of the A patients and 3.3% of the B patients died in hospital. At the other end of the severity scale, a French primary-care surveillance study of 14,423 confirmed cases over ten seasons found the illness produced by A and B was indistinguishable to the sentinel doctors who saw them, even though B skewed towards school-age children.

Children are where B earns its reputation. A study across twelve Canadian paediatric hospitals from 2004 to 2013 found mortality of 1.1% in hospitalised children with influenza B against 0.4% with influenza A, and otherwise-healthy children aged ten and over with B had close to six times the odds of ICU admission. The Vienna cohort — single centre, one season, published in Wiener Klinische Wochenschrift — points the other way. Put the two together and the honest reading is that severity tracks the subtype and the patient far more than the letter.

The 2025/26 season, and a lineage that went missing

The most recent Northern Hemisphere winter showed how much the two viruses can diverge within a single season. In the United States, influenza A, mostly A(H3N2) of the newer subclade K, peaked in late December 2025. Influenza B kept climbing and peaked in March 2026. By the week ending 7 March 2026, influenza B made up 72.8% of positive clinical laboratory tests even though it accounted for around a fifth of the season's total detections.

Preliminary CDC estimates put that season at roughly 32 million illnesses, 390,000 hospital admissions and 24,000 deaths — moderately severe by the agency's framework, and the third-highest cumulative hospitalisation rate since 2010/11. Around 191 paediatric deaths were reported, roughly 85% of them in children who were not fully vaccinated.

Meanwhile one of the two influenza B lineages has not been confirmed in circulation since March 2020. B/Yamagata appears to have been pushed out during the pandemic, and the world's vaccine regulators responded. Every US flu vaccine has been trivalent since the 2024/25 season, covering one A(H1N1), one A(H3N2) and one B/Victoria component, as the CDC explains on its trivalent vaccine page. In February 2026, the World Health Organization recommended its 2026/2027 Northern Hemisphere formulation with all three components replaced: A/Missouri/11/2025 (H1N1)pdm09, A/Darwin/1454/2025 (H3N2) and B/Tokyo/EIS13-175/2025 for egg-based vaccines, with slightly different strains for cell-culture and recombinant products. The WHO's announcement keeps the door open on Yamagata returning; surveillance continues because a virus that stops being detected is not the same as a virus that is gone.

Photo by Logan Voss on Unsplash

What the letter changes at the bedside

Less than people assume. Antiviral treatment does not hinge on type. Oseltamivir, sold as Tamiflu, is approved for treating uncomplicated influenza from two weeks of age and for prevention from one year. Baloxavir, sold as Xofluza, is a single oral dose approved in the United States for people aged five and over.

There is one genuine B-specific wrinkle. The Merck Manual notes that baloxavir may be more effective than oseltamivir against influenza B, and laboratory work has long shown that B viruses have a higher baseline half-maximal inhibitory concentration for oseltamivir than A viruses do. Whether that translates into a difference patients notice is unresolved; the large hospital dataset found no gap in length of stay or mortality.

The practical consequence is that testing matters more than typing. A PCR test confirms influenza and identifies the type, usually within hours. Rapid antigen tests are faster and cheaper but miss cases, and a negative result in a high-risk patient should not end the conversation. Guidance from the CDC and the American Academy of Pediatrics is consistent on the timing: start treatment within 48 hours of symptom onset for anyone at high risk, and don't wait for a lab result if the illness is severe or the patient is in a priority group.

The short version

Influenza B is not the mild one. Influenza A is not automatically the dangerous one. The subtype and the person in front of you decide that, and the difference shows up in the data as a set of small shifts rather than a clean line.

If you're deciding whether to get tested or treated, the type on the result will not change what happens in the first 48 hours. The timing will.