Accuracy is not mortality: corrected impact framework for psychiatric and genomic decision support

AI-assisted methodological correction and replacement impact framework, not a clinical efficacy study. This note identifies and corrects unsupported attribution of an external diagnostic-accuracy result to a proprietary engine, transport of attempt-population fractions to suicide deaths, unvalidated positive intervention effects, misuse of Monte Carlo intervals as confidence of clinical benefit, and arithmetic errors in legacy author manuscripts. The programme-specific causal effect of Athena or DNAGenomics on diagnosis, care or suicide mortality remains unknown. Null benefit, harms, pathway uptake, treatment access, overlapping mechanisms and uncertainty must be tested, not assumed away. This is a targeted correction of impact claims and does not validate all remaining genomic, psychometric or theoretical content of the legacy manuscripts. No third-party health records or private operational instructions are included. No medication or clinical deployment recommendation is made.