The power of big data algorithms to accurately predict whether a patient will take their own life is close to zero and will not replace the need for doctors’ risk assessments any time soon, experts say.

The comments come on the back of new US research that aimed to determine the effectiveness of current suicide prediction models at identifying those at risk of taking their own lives.

The systematic review of 64 different models that had been used on 14 million people, including Australians, showed their predictive validity in identifying those who will suicide was “extremely low” at 0.01 for most.

The review also showed their use resulted in a high number of people being classified as at high risk of suicide who did not go on to take their own lives, and failed to detect a considerable number who did.