Case Fatality Rate

The case fatality rate (CFR)—more accurately termed the case fatality risk or case fatality ratio in biostatistics—is a core severity metric in epidemiology and public health that measures the proportion of individuals diagnosed with a specified clinical disease who die from that disease within a defined follow-up window. Expressed as a percentage (0% to 100%) or probability (0.0 to 1.0), the CFR evaluates the lethality of a clinically recognized illness under prevailing medical care and host demographics. Strictly speaking, despite the historical word "rate" in its name, the standard CFR is a dimensionless cumulative incidence proportion (where every numerator death is drawn directly from the denominator of diagnosed cases) rather than an instantaneous hazard rate with units of inverse time (person-time⁻¹). Because its denominator is restricted to ascertained clinical or laboratory-confirmed cases, the CFR is distinct from both the infection fatality rate (IFR)—which includes all asymptomatic and undiagnosed infections in the denominator—and the cause-specific mortality rate, which divides deaths by the entire mid-year population at risk.

1. The Naive (Crude) Cumulative Case Fatality Ratio

In a completed outbreak or a closed clinical cohort where all C diagnosed cases have reached a final clinical resolution—either recovery (R) or death (D) attributable to the disease (C = D + R)—the exact case fatality ratio is:

CFR = ( D / C ) × 100%

However, when an epidemic is actively expanding in real time (at calendar time t), many recently diagnosed cases remain hospitalized or acutely ill with an unknown final outcome. Computing the naive (crude) real-time estimator:

CFR_naive(t) = [ D(t) / C(t) ] × 100%

where D(t) is cumulative deaths reported by day t and C(t) is cumulative cases diagnosed by day t, introduces two opposing epidemiological distortions that can skew early outbreak severity assessments by an order of magnitude.

2. Right-Censoring Lag (Downward Bias) vs. Ascertainment Bias (Upward Bias)

During an emerging infectious disease outbreak (such as the 2003 SARS-CoV-1 epidemic, the 2009 H1N1 influenza pandemic, or the early 2020 SARS-CoV-2 pandemic), two competing biases operate simultaneously on CFR_naive(t):

• Right-Censoring and Onset-to-Death Delay (Downward Bias During Exponential Growth): Fatal outcomes do not occur instantaneously upon diagnosis; there is a biological lag distribution f(τ) (typically 10 to 24 days for viral pneumonia) between symptom onset/confirmation and death. During the exponential growth phase of an epidemic (when the effective reproduction number R_e > 1 and daily cases double every few days), the cumulative case denominator C(t) includes a large surge of newly diagnosed patients from the past 1 to 2 weeks who will eventually die, whereas the numerator D(t) reflects deaths emerging from the much smaller cohort of cases diagnosed 2 to 3 weeks earlier. Consequently, CFR_naive(t) systematically underestimates the true CFR during exponential epidemic growth.
• Surveillance Ascertainment Bias / Severity Selection (Upward Bias): Conversely, when diagnostic laboratory capacity (such as RT-qPCR testing) is scarce or restricted to patients presenting to hospital emergency departments and intensive care units with severe respiratory distress or hemorrhagic shock, mild, subclinical, and ambulatory symptomatic cases go untested and excluded from C(t). Because severe hospitalized cases have a much higher probability of dying than mild outpatient cases, restricted testing systematically overestimates the true clinical CFR relative to all symptomatic cases.

3. Delay-Adjusted, Closed-Outcome, and Kaplan-Meier Competing-Risk Estimators

To eliminate right-censoring bias during an active outbreak, biostatisticians led by Donnelly, Ghani, Jewell, and Nishiura developed three rigorous estimators:

1. Closed-Cohort Outcome Ratio: Restricting analysis to resolved cases with known outcomes:

CFR_closed(t) = D(t) / [ D(t) + R(t) ] *(Note: If average time from hospitalization to death is shorter than average time to full hospital discharge, CFR_closed(t) can transiently overestimate CFR until discharge lags equilibrate.)*

1. Delay-Distribution Convolution Adjustment (Nishiura et al., 2009): If c(s) is the daily incidence of confirmed cases on day s and F(t − s) is the cumulative distribution function of the time interval from case confirmation to death, the expected effective denominator of cases who have had sufficient time to experience death by day t is ∑_s=0^t c(s) F(t − s), yielding the delay-adjusted estimator:

CFR_delay-adj(t) = D(t) / [ ∑_s=0^t c(s) · F(t − s) ]

1. Non-Parametric Kaplan-Meier and Aalen-Johansen Competing-Risk Survival Estimators: When individual patient line-list data are available with exact dates of diagnosis, death, recovery/discharge, or right-censoring, estimating the cumulative incidence of death I_D(τ) vs. recovery I_R(τ) via competing-risk survival analysis provides an unbiased real-time trajectory of the CFR.

Epidemiological surveillance strictly distinguishes four population metrics that share overlapping numerators or denominators:

Notice the fundamental mathematical relationship linking these metrics across a closed epidemic:

• Because clinically confirmed cases (C) are a subset of all biologically infected individuals (I_total), the CFR is always greater than or equal to the IFR (CFR ≥ IFR), with equality holding only for diseases where 100% of infections produce unmistakable clinical symptoms and are diagnosed (such as human rabies or tetanus).
• Furthermore, the population Cause-Specific Mortality Proportion equals the product of the Infection Attack Rate (IAR = I_total / N) and the Infection Fatality Rate (IFR = D / I_total):

Population Mortality Risk = (I_total / N) × (D / I_total) = IAR × IFR

Across infectious diseases, the CFR spans more than five orders of magnitude—from < 0.01% in endemic rhinovirus infections to virtually 100% in untreated encephalitic rabies or prion diseases—and varies substantially depending on timely access to essential medicines, critical care, oral rehydration therapy, and vaccination.

Because the CFR is a conditional probability rather than an immutable physical constant of a microbe, it varies strongly across four biological and structural axes:

1. Chronological Age and Immunosenescence: Almost all acute infectious diseases exhibit a J-shaped, U-shaped, or log-linear age gradient in CFR. In infant mortality (such as pertussis, respiratory syncytial virus, or diarrheal disease), an immature neonatal immune system and small airway caliber elevate CFR in the first 6 months of life. Conversely, in SARS-CoV-1, SARS-CoV-2, and pneumococcal pneumonia, thymic involution, reduced naive T-cell diversity, impaired Type I interferon signaling, and baseline cardiometabolic comorbidities drive an exponential rise in CFR past age 60. Consequently, comparing crude CFRs between two countries with different demographic age pyramids requires direct age-standardization across identical 5- or 10-year age strata (otherwise, an older population like Italy or Japan exhibits a substantially higher crude COVID-19 CFR than a younger population even when age-specific lethality is identical—an instance of Simpson's Paradox).
2. Inoculum Dose and Route of Exposure: As demonstrated across experimental and observational cohorts, the initial infectious dose (viral or bacterial inoculum) and anatomical portal of entry modify CFR. In anthrax (Bacillus anthracis), cutaneous inoculation carries an untreated CFR of ~20% (< 1% treated), gastrointestinal ingestion carries a CFR of 25% to 60%, and inhalation of endospores into mediastinal lymph nodes carries an untreated CFR of 85% to 95%.
3. Virulence Evolution and Pre-Existing Adaptive Immunity: Prior infection or vaccination decouples infection incidence from fatal severity by establishing memory B lymphocytes and T lymphocytes that restrict viral replication to the upper respiratory tract and prevent viremic dissemination, reducing the post-vaccination breakthrough CFR by 80% to 95%.
4. Hospital Surge Capacity and Healthcare Infrastructure: During epidemic peaks that exceed intensive care unit bed, mechanical ventilator, or medical oxygen capacity, the effective CFR rises sharply—a phenomenon termed surge-induced excess case fatality—underscoring why public health interventions that "flatten the epidemic curve" reduce total deaths not only by preventing infections, but also by keeping concurrent hospital demand within the threshold required to deliver standard-of-care universal health coverage.

1. Ghani, A. C., Donnelly, C. A., Cox, D. R., Griffin, J. T., Fraser, C., Lam, T. H., Ho, L. M., Chan, W. S., Anderson, R. M., Hedley, A. J., & Leung, G. M. (2005). "Methods for estimating the case fatality ratio for a novel, emerging infectious disease." American Journal of Epidemiology, 162(5), 479–486. https://doi.org/10.1093/aje/kwi230
2. Jewell, N. P., Lei, X., Ghani, A. C., Donnelly, C. A., Leung, G. M., Ho, L. M., Cowling, B. J., & Hedley, A. J. (2007). "Non-parametric estimation of the case fatality ratio with competing risks data: an application to Severe Acute Respiratory Syndrome (SARS)." Statistics in Medicine, 26(9), 1982–1998. https://doi.org/10.1002/sim.2691
3. Nishiura, H., Klinkenberg, D., Roberts, M., & Heesterbeek, J. A. P. (2009). "Early epidemiological assessment of the virulence of emerging infectious diseases: a case study of an influenza pandemic." PLOS ONE, 4(8), e6852. https://doi.org/10.1371/journal.pone.0006852
4. Lipsitch, M., Donnelly, C. A., Fraser, C., Blake, I. M., Cori, A., Dorigatti, I., Ferguson, N. M., Garske, T., Mills, H. L., Riley, S., Van Kerkhove, M. D., & Hernán, M. A. (2015). "Potential biases in estimating absolute and relative case-fatality risks during outbreaks." PLOS Neglected Tropical Diseases, 9(7), e0003846. https://doi.org/10.1371/journal.pntd.0003846
5. Kelly, H., & Cowling, B. J. (2013). "Case fatality: rate, ratio, or risk?" Epidemiology, 24(4), 622–623. https://doi.org/10.1097/EDE.0b013e318296c2b6