Speech recognition for clinical documentation from 1990 to 2018: a systematic review

Area of Science:

• Medical Informatics
• Health Services Research

Background:

• Speech recognition (SR) technology is increasingly adopted for clinical documentation.
• Existing research on SR's impact is varied, with differing methodologies and outcomes.

Purpose of the Study:

• To systematically review and synthesize literature on SR technology in clinical documentation.
• To assess SR's effects on document accuracy, provider efficiency, and institutional costs.

Main Methods:

• A comprehensive literature search was conducted across 10 databases for articles published between 1990 and 2018.
• Included articles were analyzed for research topics, medical domains, SR systems, and key findings.

Main Results:

• 122 articles were reviewed, with significant focus on radiology and emergency medicine.
• Productivity and documentation time showed mixed results; turnaround time consistently decreased.
• Error rates varied widely (4.8%-71% for documents, 7.4%-38.7% for word error rates).
• Cost impacts were mixed, with more studies reporting decreases than increases.
• Nuance Communications was the most common SR vendor studied.

Conclusions:

• Research on SR for clinical documentation is heterogeneous, hindering definitive conclusions.
• SR's expanding use beyond radiology necessitates further research into its effectiveness in varied clinical environments.