Opportunities for machine learning to improve surgical ward safety

Area of Science:

• Medical Safety
• Surgical Care
• Clinical Informatics

Background:

• Delayed recognition of patient decompensation and failure-to-rescue on surgical wards cause significant preventable harm.
• Current surveillance methods are often inadequate, leading to under-triage and delayed interventions for high-risk surgical patients.

Purpose of the Study:

• To critically evaluate evidence on improving surgical ward safety.
• To identify opportunities for earlier recognition of patient instability and improved rescue strategies.

Main Methods:

• Systematic review of 58 articles from major databases (Cochrane Library, EMBASE, PubMed).
• Analysis of existing evidence on risk assessment, patient monitoring, and decision-making in surgical wards.

Main Results:

• Patient arrest survival rates are low (15-20%), with subtle signs of instability often preceding critical events.
• Existing risk assessments are coarse, leading to under-triage and reliance on manual, error-prone surveillance methods.
• Technological advancements offer potential for more efficient and accurate risk assessment and earlier detection of complications.

Conclusions:

• Leveraging electronic health records, continuous monitoring, and AI (deep learning, reinforcement learning) can improve risk assessment.
• These technologies can facilitate earlier recognition of instability and better clinical decisions for reversible conditions.
• Enhanced technological integration is key to improving surgical ward safety and patient outcomes.