01whole.pdf (2.25 MB)
Download file

Evaluating the use of speech recognition for electronic health record documentation

Download (2.25 MB)
thesis
posted on 28.03.2022, 11:15 authored by Tobias Hodgson
The near-universal adoption of electronic health records (EHRs) has made electronic clinical documentation an essential reporting method for most practicing clinicians. Identifying the fastest and safest way for clinicians to enter data into electronic medical records (EMRs) is a matter of urgent importance. Speech recognition (SR) is an increasingly popular input modality for EHR based clinical documentation, yet there is little research evaluating its use. Previous empirical research has focused on SR as a replacement for dictation and transcription, rather than on its effect on safety, efficiency and usability as an input modality for EHRs. In order to address this knowledge gap, an experiment was performed and then replicated (E1 andE2) with Emergency Department (ED) doctors to compare the impact of input modality used i.e.keyboard and mouse (KBM) or SR, on clinical documentation. In a controlled environment, doctors were asked to complete tasks of varying complexity within a commercial EHR (E1 8, E2 4). Safety (number and type of errors during documentation), efficiency (the time taken to complete tasks), and usability (doctors' ease of use perception) of the EHR configured with each input modality were measured. The experiments found no safety or efficiency gains when ED doctors utilised SR for EHR based clinical documentation. For complex tasks, SR assisted documentation took significantly longer to complete when compared to KBM (E1: 18.40%, P=0.009, CI: 9.61-47.73; E2: 16.94%, P=0.01, CI: 11.86-48.26). Overall tasks completed with SR resulted in more non-typographical errors when compared to KBM (E1: KBM 32, SR 138; P<0.01, CI: -1.87--1.16; E2: KBM 26, SR 137, P<0.01, CI: -2.01--1.17). Potential patient harm errors were significantly greater with SR for simple tasks. SR had a negative impact on the usability score (SUS score: KBM 67 vs. SR 61, P=0.045, CI: 0.14-12.00). Overall, SR was perceived to require more training and support than KBM.

History

Table of Contents

1. Introduction -- 2. Literature review : risks and benefits of speech recognition for clinical documentation -- 3. Experimental study : efficiency and safety of speech recognition for documentation in the EHR -- 4. Replication study : evaluating the use of speech recognition within a commercial EHR system -- 5. System usability study : evaluating the usability of speech recognition to create clinical documentation using a commercial EHR -- 6. Discussion -- Appendices.

Notes

Includes bibliographic references Thesis by publication. "Centre for Health Informatics, Australian Institute of Health Innovation, Faculty of Medicine and Health Sciences, Macquarie University" -- title page.

Awarding Institution

Macquarie University

Degree Type

Thesis PhD

Degree

PhD, Macquarie University, Faculty of Medicine and Health Sciences, Australian Institute of Health Innovation

Department, Centre or School

Australian Institute of Health Innovation

Year of Award

2019

Principal Supervisor

Enrico Coiera

Additional Supervisor 1

Farah Magrabi

Rights

Copyright Tobias Hodgson 2019. Copyright disclaimer: http://mq.edu.au/library/copyright

Language

English

Extent

1 online resource (ix, 136 pages) diagrams, graphs, tables

Former Identifiers

mq:71574 http://hdl.handle.net/1959.14/1275780