Budget-Impact Analysis of AI-Supported Fracture Detection: A Multi-center Study in Moravian Silesian Hospitals

Musculoskeletal (MSK) fractures are a major emergency department (ED) finding, yet diagnosing them on plain X-ray is error-prone under high workload. Available radiology AI software can flag fractures with high accuracy, but their economic value remains largely unknown. To quantify the one-year budget-impact and return on investment (ROI) of deploying the AI software in five Moravian–Silesian regional hospitals, we developed a decision tree comparing: (1) standard care–radiologists evaluate 339 828 annual MSK X-rays (sensitivity 0.824; specificity 0.957); and (2) AI-assisted care–AI pre-reads all X-rays (Se \(=\) 0.921; Sp \(=\) 0.897; cost €1/scan) prior to radiologist confirmation. The prevalence of first-diagnosis fractures was set at 6% (n \(=\) 20 390). Downstream costs included ED revisits (€150 per false negative [FN]), 1-day admissions for 25% of FNs (€194), and outpatient referrals for 10% of false positives (FPs) (€20). Radiologist time saved (1 min/read at €27/h) was valued; litigation costs were excluded. AI assistance reduced FNs by 55% (from 3 587 to 1 614) and increased FPs by 140% (13 735–32 891). Downstream costs fell from €740 677 (radiologist-only) to €385 646 (AI-assisted), netting €355 031 savings. Radiologist time saved (4 803 h) added €129 681 in value. After subtracting AI fees (€339 828, using €1/scan), the net benefit was €143 116 region-wide (€28 623 per hospital). The benefit–cost ratio was 1.42 (ROI 42%); at €0.75 and €0.50 per scan, ROI grew to 90 and 184%, respectively. These findings support scaling the AI deployment across additional hospitals, with continuous monitoring of diagnostic accuracy and budget impact in routine practice.