"New virus has 30% fatality rate." You've seen this headline before. It appeared during SARS, during MERS, during the early weeks of COVID-19, and it'll appear again during the next outbreak. Every single time, the number was misleading. Not because anyone was lying, but because the statistic being reported (the case fatality rate) is structurally biased toward overestimation during the early phase of any outbreak.
Knowing the difference between CFR and IFR is one of the most practical things you can learn about reading outbreak data.
What is CFR?
Case fatality rate (CFR) equals the number of deaths divided by the number of confirmed cases, expressed as a percentage. If 100 people are confirmed infected and 10 die, the CFR is 10%. Simple math, but dangerously incomplete.
The word "confirmed" is doing enormous work in that formula. Confirmed cases are people who were sick enough to seek medical care, got tested, and received a positive result. During most outbreaks, testing is limited to the sickest patients. Mild cases stay home. Asymptomatic infections are never counted. The denominator is artificially small, which makes the death rate look artificially high.
CFR accurately measures the death rate among known cases. But it's consistently mistaken for the death rate among all infections, and those are very different numbers.
What is IFR?
Infection fatality rate (IFR) equals the number of deaths divided by the total number of infections, including undetected and asymptomatic cases. IFR represents the true lethality of a pathogen across all infected people, not just the ones who showed up at a hospital.
Calculating IFR requires knowing the total number of infections, which is impossible to determine in real time. Epidemiologists estimate it later through seroprevalence studies, testing random population samples for antibodies. These studies take months.
IFR is almost always lower than CFR. Often dramatically lower. The gap between the two tells you how much undetected transmission is happening.
How wrong was early COVID-19 CFR?
Very wrong. On March 3, 2020, WHO Director-General Tedros Adhanom Ghebreyesus stated that the global CFR for COVID-19 was 3.4%. Headlines worldwide ran with it. A 3.4% fatality rate for a pathogen spreading this fast implied tens of millions of deaths.
At that point, testing was almost exclusively limited to hospitalized patients. If you had a mild cough and a fever in February 2020, you weren't getting tested in most countries. Only the severely ill showed up in the denominator. Of course the death rate looked high.
Later seroprevalence studies revealed the truth. The actual IFR was approximately 0.5-1.0% across the general population. For people under 40, IFR was below 0.1%. For people over 70, it exceeded 5%. The 3.4% CFR was roughly 3-7 times higher than reality.
Was COVID-19 still serious at a 0.5-1.0% IFR? Absolutely. Applied to hundreds of millions of infections, that rate produced millions of deaths. But the early 3.4% figure triggered policy responses calibrated to a disease roughly three times deadlier than what was actually circulating.
When does CFR match IFR?
CFR closely approximates IFR when a disease makes almost everyone visibly and severely sick, leaving few undetected cases. Ebola is the clearest example.
Ebola's CFR has historically ranged from 25% to 90% depending on the strain and outbreak. IFR for Ebola is nearly identical because the disease is brutal and obvious. You don't walk around with a mild case of Ebola. Nearly every infected person develops severe symptoms and enters the confirmed case count.
Diseases with a high proportion of asymptomatic or mild infections show the biggest CFR-IFR gaps. Polio infects many people but causes paralysis in fewer than 1%. If you only counted paralytic cases, polio's apparent fatality rate would look very different from its true population-level mortality.
How should you read fatality numbers during an outbreak?
When you see a fatality rate reported during the first weeks of a new outbreak, assume the CFR overstates the true risk by 2-10x until seroprevalence data is available. Track the direction of the number over time; a declining CFR usually means testing is expanding and milder cases are being captured.
Four things matter more than the raw fatality percentage:
Trend direction. Is the CFR falling as testing expands, or holding steady? A falling CFR suggests many mild cases are being found. A stable or rising CFR in the context of expanded testing is a more concerning signal.
Healthcare system capacity. The same pathogen can produce a 5% CFR in a country with well-equipped hospitals and a 20% CFR where healthcare has collapsed. During the 2014 West Africa Ebola outbreak, CFR varied from roughly 40% in areas with international treatment centers to over 70% in communities with no medical infrastructure.
Age distribution of cases. COVID-19's IFR varied by more than 1,000x across age groups: roughly 0.002% for children under 10 versus over 10% for adults over 80. An outbreak concentrated among elderly care home residents will produce a very different CFR than one spreading primarily among young adults.
Case definition changes. As an outbreak progresses, the criteria for what counts as a "case" often shift. Early in COVID-19, a case required a positive PCR test. Later, clinical diagnosis without testing was accepted in many countries. Expanding the case definition lowers the apparent CFR, even if nothing about the disease itself has changed.
What does PandemicAlarm show?
PandemicAlarm reports CFR when case and death counts are available from official sources, because that's the data we have in real time. We don't fabricate IFR estimates. Our severity scoring accounts for the known biases in CFR by weighting other factors — transmission mode, geographic spread, healthcare infrastructure, and pathogen novelty — alongside raw mortality numbers.
When a new outbreak shows a 30% CFR based on 50 confirmed cases, that number appears on the dashboard. But the severity score reflects the full picture, not just the mortality figure. A novel respiratory pathogen with a 2% CFR and rapid geographic spread will score higher than a known pathogen with a 30% CFR confined to a single healthcare facility, because pandemic potential depends on transmissibility at least as much as lethality.
How do you avoid panicking over headlines?
Read beyond the number. When a headline says "deadly new virus kills 1 in 3," ask yourself: How many total cases? If fewer than 100, the CFR is statistically unreliable. Who is being tested? If only hospitalized patients, the denominator is biased. What type of pathogen? Respiratory viruses almost always have large numbers of undetected mild cases.
Early outbreak numbers are blurry photographs. They'll get sharper over time. Your job is to monitor the focus as it improves, adjusting your assessment as better data arrives. If CFR remains high even as testing expands and case counts grow, the pathogen is genuinely dangerous. If CFR drops steadily with expanded surveillance, the initial alarm was likely overstated.
Patience with data is a preparedness skill.