How To Calculate Etiologic Fraction

Use this professional calculator to estimate etiologic fraction among exposed and population attributable fraction from either relative risk data or a full 2×2 table.

Expert Guide: How to Calculate Etiologic Fraction Correctly

Etiologic fraction is one of the most useful concepts in epidemiology because it turns abstract risk estimates into a practical statement about preventable disease burden. When clinicians, public health teams, and researchers ask, “How much disease is due to this exposure?”, they are usually asking for an etiologic fraction. If the metric is computed correctly and interpreted in context, it helps prioritize prevention policy, clinical counseling, and resource allocation.

In plain language, the etiologic fraction asks what share of disease cases could be avoided if an exposure were removed, assuming the exposure causes the disease and the model assumptions are satisfied. Two forms are used most often:

• Etiologic fraction among exposed (also called attributable fraction in the exposed): proportion of cases among exposed people that are attributable to exposure.
• Population attributable fraction (PAF): proportion of all cases in the total population attributable to exposure.

Core formulas you should know

If you have a valid relative risk (RR), then:

1. Etiologic fraction among exposed: EF_exposed = (RR – 1) / RR
2. Population attributable fraction: PAF = Pe × (RR – 1) / [1 + Pe × (RR – 1)] where Pe is population exposure prevalence

If you have cohort count data in a 2×2 table, you can derive RR directly:

• Incidence in exposed I_e = a / (a + b)
• Incidence in unexposed I_u = c / (c + d)
• RR = I_e / I_u
• EF_exposed = (I_e – I_u) / I_e
• PAF = (I_t – I_u) / I_t, where I_t is total incidence = (a + c) / (a + b + c + d)

Why etiologic fraction matters in real decisions

Risk ratios alone do not tell you how many total cases could be avoided in a community. An exposure can have a high RR but low prevalence, producing modest population burden. Another exposure can have lower RR but very high prevalence, producing large preventable burden. That distinction is exactly why public health planning relies on PAF as well as RR.

For example, smoking has a very high RR for lung cancer, and the attributable burden is large despite declining prevalence. In contrast, a milder risk factor with broad exposure in the population can still produce substantial case counts because so many people are exposed.

Step by step calculation workflow

1. Define exposure and disease precisely, including time window.
2. Use the most valid effect estimate available, ideally an adjusted RR from a high quality cohort or meta analysis.
3. Estimate exposure prevalence for the same target population.
4. Calculate EF among exposed and PAF using consistent units.
5. Interpret under assumptions: causal relationship, adequate confounding control, and realistic removal of exposure.

Important: Etiologic fraction is not always the same as directly preventable fraction in real programs. Real world interventions may only partially reduce exposure, and some damage may be irreversible.

Using odds ratio in place of relative risk

Many studies report odds ratios rather than risk ratios, especially case control designs. OR can overstate RR when outcomes are common. If disease is rare, OR is often close to RR and can be used as an approximation. If not rare, convert OR to RR with baseline risk in unexposed:

RR ≈ OR / [(1 – P0) + (P0 × OR)]

Where P0 is risk among unexposed. This calculator supports both options so your etiologic fraction estimate is more defensible.

Comparison table: Real U.S. statistics that influence etiologic burden

Risk factor	Statistic	Why it matters for etiologic fraction	Source
Current cigarette smoking in U.S. adults	11.6% (2022)	Lower prevalence than prior decades, but smoking still has strong disease associations, so PAF remains meaningful for several outcomes.	CDC
Adult obesity prevalence in the U.S.	41.9% (NHANES 2017 to March 2020)	High prevalence can generate substantial PAF even when RR is moderate, making obesity related burden large at population level.	CDC
Adults with hypertension in the U.S.	48.1%	Very common exposure states often dominate preventable burden calculations in cardiovascular outcomes.	CDC

Comparison table: Examples of attributable burden estimates

Exposure-outcome pair	Reported attributable proportion	Interpretation for EF thinking	Source
HPV and cervical cancer	About 90% of cervical cancers are caused by HPV	This is conceptually similar to a very high etiologic fraction for that exposure-outcome relationship.	CDC
Cigarette smoking and lung cancer deaths	About 80% to 90% of lung cancer deaths linked to smoking	Shows how a strong effect size can drive high attributable burden despite reduced smoking prevalence.	CDC

Worked example with RR and prevalence

Suppose a cohort reports RR = 2.5 for an exposure-disease pair and exposure prevalence in your target population is 25%.

• EF among exposed = (2.5 – 1) / 2.5 = 0.60, or 60%
• PAF = 0.25 × 1.5 / (1 + 0.25 × 1.5) = 0.2727, or 27.27%

Interpretation: roughly 60% of cases among exposed individuals could be attributed to exposure, and about 27% of all cases in the population could be attributable to that exposure, assuming causal validity and accurate measurements.

Worked example with 2×2 table data

Imagine a cohort with:

• a = 80 exposed with disease
• b = 920 exposed without disease
• c = 30 unexposed with disease
• d = 970 unexposed without disease

Then:

• I_e = 80 / 1000 = 0.08
• I_u = 30 / 1000 = 0.03
• RR = 2.67
• EF among exposed = (0.08 – 0.03) / 0.08 = 0.625 or 62.5%
• Total incidence I_t = 110 / 2000 = 0.055
• PAF = (0.055 – 0.03) / 0.055 = 0.4545 or 45.45%

Common mistakes that bias etiologic fraction

• Using OR as RR without checking outcome rarity. This can exaggerate EF and PAF.
• Mixing populations. RR from one group and prevalence from a different group can produce misleading PAF.
• Ignoring confounding. Unadjusted RR can inflate attributed burden.
• Treating association as causation. EF has causal interpretation only with strong design and assumptions.
• Not reporting uncertainty. Confidence intervals around RR should propagate to EF and PAF uncertainty ranges.

How to report results professionally

For publication or technical reporting, include: exposure definition, outcome definition, effect measure source, adjustment set, prevalence data source, formula used, and assumptions. A clear reporting sentence can look like this: “Using adjusted RR 2.5 and exposure prevalence 25%, we estimated EF among exposed at 60% and population attributable fraction at 27.3%.”

If possible, add sensitivity analyses. For example, re-run with lower and upper RR confidence limits, or with plausible exposure prevalence bounds. Decision makers trust results more when they can see how robust estimates are to modeling assumptions.

Authoritative references

• CDC Tobacco Fast Facts (.gov)
• CDC Adult Obesity Facts (.gov)
• CDC HPV and Cancer (.gov)

Bottom line: calculating etiologic fraction is straightforward mathematically, but high quality interpretation depends on study design, causal reasoning, and valid population inputs. Use the calculator above to perform the arithmetic quickly, then apply careful epidemiologic judgment before using estimates for policy or clinical decisions.