{"stable_id":"41cba9f08878beb9","slug":"germany-praim-mammography","company":"German organised mammography screening programme (PRAIM implementation study; 12 sites)","workflow_name":"AI-supported double reading with normal triaging and a post-read safety net (Vara MG)","function_code":"healthcare_screening","pattern_codes":["autonomous_with_backstop","exception_based_operations"],"changed_assumption":"Two unaided human readers plus consensus are not the only safe model for a national screening programme.","evidence_strength":"verified","publication_tier":"showcase","freshness":"current","reviewed_at":"2026-08-23","updated_at":"2026-08-22","source_quality_summary":"2 peer reviewed; publication outcomes are verified.","caveat_summary":"Observational; radiologists chose the viewer. Authors used propensity-score overlap weighting because some readers preferentially used AI on normal-tagged exams. Not a reader-replacement deployment.","freshness_caveat":null,"workflow_summary":{"before":[{"actor":"Two radiologists","action":"Independently double-read four-view mammograms without AI predictions.","actor_type":"control"},{"actor":"Consensus conference","action":"Confirms or dismisses suspicion and issues recall.","actor_type":"control"}],"hinge":"Two unaided human readers plus consensus are not the only safe model for a national screening programme.","after":[{"actor":"Vara MG AI","action":"Tags confident-normal cases and fires a localization safety net when humans call suspicious cases normal.","actor_type":"ai"},{"actor":"Radiologist (AI viewer)","action":"Reads with normal tags visible and must accept or reject AI localization after a safety-net prompt.","actor_type":"human"}],"decision_mode":"bounded_autonomy","decision_marker":"AI acts within a human backstop"},"before":[{"order":1,"actor":"Two radiologists","action":"Independently double-read four-view mammograms without AI predictions.","handoff_to":"Consensus conference if either flags suspicion","control":"Binding national double-reading guideline"},{"order":2,"actor":"Consensus conference","action":"Confirms or dismisses suspicion and issues recall.","handoff_to":"Diagnostic assessment","control":"At least the two readers plus a head radiologist"}],"after":[{"order":1,"actor":"Vara MG AI","action":"Tags confident-normal cases and fires a localization safety net when humans call suspicious cases normal.","handoff_to":"Radiologist using the AI-supported viewer","control":"CE-marked device; live vendor monitoring; radiologist may ignore the AI viewer"},{"order":2,"actor":"Radiologist (AI viewer)","action":"Reads with normal tags visible and must accept or reject AI localization after a safety-net prompt.","handoff_to":"Consensus if suspicion remains","control":"Final recall remains human; 204 cancers in the AI group were diagnosed after accepted safety-net prompts"}],"decision_rights":"Radiologists choose the viewer per exam. The safety net can force reconsideration; it cannot recall a woman.","exception_path":"Readers can stay on the non-AI viewer. Examinations tagged normal by AI still produced 20 cancers in the AI group after human consensus, so the normal tag is not an autonomous discharge.","removed_work":["Some reread time on AI-normal exams (median 16 s versus 30 s unclassified in the AI group)","Not a removed second reader; that scenario is a post-hoc analysis, not the deployed protocol"],"outcomes":[{"metric":"Model-based breast cancer detection rate","baseline":"5.70 per 1,000 in the contemporaneous non-AI viewer group","result":"6.70 per 1,000 in the AI-supported group; +17.6% (95% CI +5.7% to +30.8%) after overlap weighting","period":"2021-07-01 to 2023-02-23","scale":"461,818 women analysed; 260,739 AI group; 201,079 control; 119 radiologists; 12 sites","attribution_caveat":"Observational; radiologists chose the viewer. Authors used propensity-score overlap weighting because some readers preferentially used AI on normal-tagged exams. Not a reader-replacement deployment.","evidence_label":"verified"}],"executive_lesson":"PRAIM is not Denmark. Germany kept two human readers and added a machine that can force a second look after a human-normal call. Detection rose at national implementation scale. Do not sell this as retired double reading; the paper’s 56.7% automation scenario is hypothetical.","anti_pattern":"Reporting the fictitious ‘do not read AI-normals’ scenario as the deployed workflow.","questions_for_leaders":["If we cannot drop a second reviewer, will we still force a machine interrupt after a human-normal call?","How will we stop readers from using the AI path only on easy exams?"],"collections":["regulated-autonomy"],"bundle_version":"1.0.0","bundle_fingerprint":"sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93","canonical_url":"https://brianletort.ai/transformations/germany-praim-mammography"}