Attributable and Preventable Impact Calculator

This tool supports research planning and does not replace statistical consultation.

Scope

This calculator translates a binary-outcome risk ratio into exposed-group, population, and counterfactual impact measures. It accepts cohort-style counts or a supplied risk ratio. It does not accept case-control data, odds ratios, hazard ratios, rate ratios, continuous or multilevel exposures, time-to-event outcomes, or regression coefficients for direct adjusted-PAF estimation.

Formula and inference

For increased risk, the attributable fraction among exposed equals 1 − 1/RR and the Levin population attributable fraction equals p(RR − 1) / [1 + p(RR − 1)]. For reduced risk, the prevented fraction among exposed equals 1 − RR and the population prevented fraction relative to no exposure equals p(1 − RR). A partial change from current exposure prevalence p to counterfactual prevalence p* uses the potential impact fraction (p − p*)(RR − 1) / [1 + p(RR − 1)]. Count-pathway risks use Wilson intervals, the risk difference uses Newcombe's hybrid-score interval, and the RR uses a log-Wald interval when estimable.

Conditional impact intervals

When RR confidence limits are available, impact limits are obtained by transforming those endpoints while holding exposure prevalence, reference risk, population size, and outcome count fixed. These conditional intervals do not include uncertainty in those inputs, model choice, bias, transportability, or causal assumptions.

Assumptions and limitations

• All inputs describe compatible populations, one binary outcome, and the same risk period.
• The supplied RR is applicable across the target-population exposure scenario and is a risk ratio—not an odds, hazard, prevalence, or rate ratio.
• Causal or preventable interpretation additionally requires adequate control of confounding and bias, consistency, positivity, a well-defined feasible intervention, and transportability.
• Crude count estimates may be confounded. Effect modification or covariate adjustment may require standardized model-based PAF methods.
• PAFs for overlapping causes generally cannot be added and need not sum to 100%.
• A large impact fraction does not establish intervention feasibility, clinical benefit, or cost-effectiveness.

Validation

The statistical methods introduced in version 1.0.0 were checked using an independently coded reference script, WHO categorical comparative-risk formulas, published definitions, and manually verified harmful, protective, null, partial-change, and boundary fixtures. Count-pathway Wilson, Newcombe, and log-risk-ratio calculations reuse the independently validated EPI01 methods. Automated tests cover formulas, identities, interval transformations, parsing, reporting, privacy, and responsive-interface contracts. Availability does not constitute regulatory validation.

Privacy

All calculations use aggregate values in browser memory. The app does not transmit, upload, or persist entered counts, estimates, labels, or results. Do not enter identifiable or confidential information.

References

1. Miettinen OS. Proportion of disease caused or prevented by a given exposure, trait or intervention. Am J Epidemiol. 1974;99:325–332.
2. Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health. 1998;88:15–19.
3. World Health Organization. Preventing disease through healthy environments. Technical annex on population attributable fractions. WHO; 2016.
4. Greenland S, Drescher K. Maximum likelihood estimation of the attributable fraction from logistic models. Biometrics. 1993;49:865–872.
5. Newcombe RG. Interval estimation for the difference between independent proportions. Stat Med. 1998;17:873–890.
6. Stevenson M, Sergeant E. epiR: tools for the analysis of epidemiological data; epi.2by2.