Systemically biased AI models can reduce clinicians' diagnostic accuracy, and explanations alone may not mitigate this harm, according to a multicenter study involving over 450 clinicians.