Luigi Lepanto MD, Presenter: Research Grant, Bracco Group Speaker, Johnson & Johnson

Imane Hammana PhD, Abstract Co-Author: Nothing to Disclose

My Sandra Ly, Abstract Co-Author: Nothing to Disclose

Christian Bellemare MSc, Abstract Co-Author: Nothing to Disclose

Charles Poder PhD, Abstract Co-Author: Nothing to Disclose

To conduct a systematic review of studies describing the impact of speech recognition systems (SRS) on report error rates and productivity, as well as the literature describing the factors influencing user acceptance.

The following electronic databases were searched: Embase, INAHTA, CINALH, OVID Healthstar, Cochrane Database, Medline, Pubmed.  Published health technology assessment reports were also searched. Comparative studies reporting error rates, turnaround times, and radiologist productivity were selected. Quality assessment was performed using GRADE. Descriptive studies were selected if they described factors affecting user acceptance.

A report produced by a Health Technology Assessment Agency in 2008 was identified, as well as 24 studies that met the selection criteria. There were 14 studies comparing the use of SRS to human transcription (HT). Following quality assessment, 3 prospective and 3 retrospective studies were retained. There was a large amount of heterogeneity in the studies. Findings reveal the underperformance of SRS compared to HT. The average performance level of SRS was 93.9% (84% to 98.5%) compared with an average of more than 99.2% for HT (97 to 99.97%). Error rates increased with report complexity. Turnaround time increased by 15% to 190%, depending on user experience and report complexity. Studies on the perceived benefits identified technical and organizational factors influencing adoption of SRS. Gradual and selective implementation, with adequate support, is more likely to lead to successful adoption.

The negative impact of SRS on error rates and productivity is still significant. Appropriate implementation strategies are essential to favor adoption.

Implementing speech recognition systems remains a challenge.  Maximum benefit is likely only if implementation is gradual and selective.

Lepanto, L, Hammana, I, Ly, M, Bellemare, C, Poder, C, The Impact of Speech Recognition Systems on Error Rates and Productivity: A Systematic Review.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12030477.html