Patient Information Summarization in Clinical Settings: Scoping Review

Area of Science:

• Medical Informatics
• Clinical Data Management
• Artificial Intelligence in Healthcare

Background:

• Clinical environments face information overload, necessitating effective clinical data summarization.
• Existing solutions for clinical summarization lack a comprehensive research overview.

Purpose of the Study:

• To identify and analyze state-of-the-art clinical summarization solutions.
• To evaluate the capabilities and properties of various summarization techniques.
• To provide a framework for comparing clinical summarization methods.

Main Methods:

• Conducted a scoping review of 128 articles published between 2005 and 2022.
• Utilized PubMed and Web of Science, extending searches via citation chaining.
• Analyzed summarization methods using the "collect-synthesize-communicate" framework, assessing conservation of temporality, uncertainty, and medical pertinence.

Main Results:

• Input data primarily included structured data (46.1%), text (41.4%), or both (10.2%).
• Summarization approaches varied, with concept/relation extraction (30 papers) and visual design (59 papers) being prominent.
• Temporality (57.8%) and uncertainty (88.3%) information were often not conserved; however, some studies addressed these aspects.
• Most studies (89.8%) reported evaluations, including human participant studies (n=31) and intrinsic evaluations (n=24).
• Only 8.6% of reports described systems deployed in clinical settings.

Conclusions:

• The literature presents numerous clinical summarization methods but few comparative analyses.
• Current summarization solutions often partially address the "collect-synthesize-communicate" framework.
• The conservation and utilization of temporal, uncertainty, and pertinent medical information vary significantly, with many solutions remaining preliminary.