Real-time AI alerts reduce patient deaths by 43%, Mount Sinai study finds

A study from Mount Sinai Hospital, published in Critical Care Medicine, indicates that AI can greatly improve patient outcomes by predicting and preventing health deterioration. The research found that hospitalized patients who received AI-generated alerts were 43% more likely to get escalated care and had significantly lower mortality rates. Traditionally, healthcare relied on manual methods like the MEWS, but this study shows that AI algorithms can prompt earlier intervention, potentially saving lives.

The study involved over 2,700 adult patients split into two groups: one receiving real-time AI alerts and another where alerts were generated but not delivered. The group with AI alerts saw increased likelihood of receiving circulation-support medications and a reduced risk of death within 30 days. Dr. David Reich, President of Mount Sinai Hospital, highlighted that real-time AI alerts act as precise and timely decision-making aids, ensuring appropriate care reaches patients at critical moments.

Although the study was halted early due to the COVID-19 pandemic, it led to the implementation of a simplified AI system in Mount Sinai’s stepdown units. Here, intensive care physicians review patients with the highest predicted risk daily, making treatment recommendations. The algorithm continuously improves its accuracy through reinforcement learning. This research, along with 15 other AI-based clinical tools at Mount Sinai, underscores the transformative potential of AI in healthcare, enabling earlier interventions and better decision-making to save lives.