Artificial Intelligence Counterfactual Fairness Audit of Community Clinic Anaemia Screening Under Unmeasured Haemoglobinometer Calibration Bias: A Healthcare Sex-Equity Guarantee Evaluation

Counterfactual fairness audits of clinical screening algorithms typically assume that the measured covariates are error-free, yet community-clinic haemoglobinometers drift in calibration in ways that are unrecorded and that may differ systematically by patient sex. The framework separates the counterfactual fairness of the screening rule from the identifiability of that fairness under mismeasured haemoglobin, and it derives a bound on the sex-specific screening error that survives arbitrary calibration drift within a declared range. A replication-based evaluation of the ['Intercept', 'exposure', 'group', 'severity'] estimator was conducted: 600 data sets of N = 500 were drawn from a declared data-generating process (intercept 12.0, exposure 0.42, group -3.1, severity 1.35, error SD 4.2), and the behaviour of the ordinary least squares estimator under the stated process was characterised. Across the 600 replications the estimated bias was -0.00237 for the intercept and -0.00014 for the exposure coefficient, with RMSE 0.88241 and 0.01556 respectively, and the empirical coverage of the nominal 95% confidence intervals was 1.0 for both parameters. The framework proposes that sex-equity guarantees in anaemia screening should be stated as bounds that are robust to calibration drift rather than as point equalities, and it identifies the conditions under which a screening rule that is counterfactually fair on true haemoglobin remains fair on measured haemoglobin.