An experimental investigation of the impact of alert frequency and relevance on alert dwell time

Aim: This study aimed to determine the impact of alert frequency and relevance on alert dwell time. Method: A 2 × 3 design was used where 127 university students completed 60 prescribing tasks and were presented with a variable frequency of computerized alerts (low, medium and high) with variable relevance (low and high). Participants were instructed to override an alert if it was not relevant to their prescription, and to cancel the order if the alert signalled an error in their order. Results: Participants presented with a small number of alerts spent more time attending to alert content than participants presented with a medium or high number of alerts (respectively median 15.6 s vs 10.8 vs 10.2 s). Alert relevance had no impact on alert dwell time. Alerts requiring an override response were 4.5 times more likely to be correctly actioned than alerts requiring the order to be cancelled. Discussion: Dwell time was influenced by alert frequency, with greater exposure to alerts associated with shorter dwell times. We hypothesize that this was because participants came to learn that spending time on alert information was unnecessary. We propose that when users experience no consequences or feedback from overriding alerts they quickly learn that this action is more efficient and so more rewarding than taking any other action.