Summary
Speech recognition technology (SRT), the automated conversion of voice into text, is one solution that promises to combat burnout by saving clinicians’ time and improving documentation efficiency.
Electronic health records (EHRs) have improved the provision of healthcare in many ways, such as, legibility of and access to information, clinical decision support, and increased reporting capabilities. Yet EHRs have also created stress for physicians and contributed to physician burnout.
According to Mayo Clinic researcher Liselotte Dyrbye, MD, 50% of the workforce have substantial symptoms of burnout. She claims: “The electronic health record is certainly a component. Physicians spend two hours interacting with the EHR for every one hour they spend with patients.”
Speech recognition technology (SRT), the automated conversion of voice into text, is one solution that promises to combat burnout by saving clinicians’ time and improving documentation efficiency. SRT also helps fulfill the increasing requirements for near real-time data and lowers costs by increasing productivity.1 Used since the 1980s to improve clinical documentation, SRT has recently gained traction. A 2018 survey showed that 90% of hospitals planned to increase their SRT use.
There are two types of SRT: back-end speech recognition (BESR) and front-end speech recognition (FESR). With BESR, a speech recognition engine captures the physician’s dictation and converts it to text. A healthcare documentation specialist then edits this text and returns it to the physician for review.
For the physician, the BESR method does not differ from traditional transcription, as it employs dictation as usual. It improves productivity on the back end, however, as healthcare documentation specialists edit the text produced by the speech recognition engine rather than transcribe dictation. In order to achieve the maximum benefit from BESR, recognition accuracy must be very high. Otherwise, it may take longer for the documentation specialist to edit the document than to transcribe it.
Conversely, healthcare documentation specialists do not edit FESR. Instead, a monitor displays the physician’s dictation in real time. The physician manages the whole process by reviewing and editing the text before authenticating it. This makes FESR the more efficient of the two methods. When interfaced with the EHR, it enables the physician to use triggers and alerts that improve productivity and safety.1
Although SRT provides many benefits, the technology also carries some risks, including:
Another study involved the emergency department of a tertiary academic teaching hospital. A random sample of 100 notes dictated by attending emergency physicians revealed 128 total errors (1.3 errors per note). Seventy-one percent of the notes contained errors, and 15% of these contained one or more critical errors. Annunciation errors ranked the highest at 53.9% (n=69).
A third study by Zhou et al. involved a random sample of 217 dictated notes from 144 unique physicians who used BESR. The results showed an error rate of 7.4%. The error rate decreased to 0.4% after healthcare documentation specialist review and to 0.3% after physician signature. Discharge summaries had higher mean error rates than other types of notes (8.9% versus 6.6%). Deletion errors were the most common at 34.7%, followed by insertion errors at 27%.
The third study illustrates how important is for physicians to review notes thoroughly prior to authentication. Examples of clinically significant errors from that study follow:
It is important to note that healthcare documentation companies are using disclaimers to call attention to SRT problems. One such disclaimer was reported to The Joint Commission: “Portions of the record may have been created with voice recognition software. Occasional wrong-word or sound-a-like substitutions may have occurred due to the inherent limitations of voice recognition software. Read the chart carefully and recognize, using context, where substitutions have occurred.” While some substitutions may be clinically insignificant, others may be significant and can potentially cause serious harm if misinterpreted. Dictation created by SRT should be held to the same standards of accuracy as medical transcription.
Consider the following to minimize SRT-associated risk:
References:
Speech Recognition in the Electronic Health Record (2013 update). AHIMA. Updated 2013. http://library.ahima.org/doc?oid=300181. Published September 2013. Accessed June 16, 2021.
Copyrighted. No legal or medical advice intended. This post includes general risk management guidelines. Such materials are for informational purposes only and may not reflect the most current legal or medical developments. These informational materials are not intended, and must not be taken, as legal or medical advice on any particular set of facts or circumstances.
According to Mayo Clinic researcher Liselotte Dyrbye, MD, 50% of the workforce have substantial symptoms of burnout. She claims: “The electronic health record is certainly a component. Physicians spend two hours interacting with the EHR for every one hour they spend with patients.”
Speech recognition technology (SRT), the automated conversion of voice into text, is one solution that promises to combat burnout by saving clinicians’ time and improving documentation efficiency. SRT also helps fulfill the increasing requirements for near real-time data and lowers costs by increasing productivity.1 Used since the 1980s to improve clinical documentation, SRT has recently gained traction. A 2018 survey showed that 90% of hospitals planned to increase their SRT use.
There are two types of SRT: back-end speech recognition (BESR) and front-end speech recognition (FESR). With BESR, a speech recognition engine captures the physician’s dictation and converts it to text. A healthcare documentation specialist then edits this text and returns it to the physician for review.
For the physician, the BESR method does not differ from traditional transcription, as it employs dictation as usual. It improves productivity on the back end, however, as healthcare documentation specialists edit the text produced by the speech recognition engine rather than transcribe dictation. In order to achieve the maximum benefit from BESR, recognition accuracy must be very high. Otherwise, it may take longer for the documentation specialist to edit the document than to transcribe it.
Conversely, healthcare documentation specialists do not edit FESR. Instead, a monitor displays the physician’s dictation in real time. The physician manages the whole process by reviewing and editing the text before authenticating it. This makes FESR the more efficient of the two methods. When interfaced with the EHR, it enables the physician to use triggers and alerts that improve productivity and safety.1
Although SRT provides many benefits, the technology also carries some risks, including:
- Incorrect use and expectation of capabilities.
- Insufficient standards for style, grammar, and readability.
- Unclear roles and standards for healthcare documentation specialists.
- Untimely review and authentication of notes created by BESR.
- Inadequate or superficial review of notes by physicians prior to authentication.
- Minimal monitoring of documentation processes.
- Minimal regulatory oversight.
Another study involved the emergency department of a tertiary academic teaching hospital. A random sample of 100 notes dictated by attending emergency physicians revealed 128 total errors (1.3 errors per note). Seventy-one percent of the notes contained errors, and 15% of these contained one or more critical errors. Annunciation errors ranked the highest at 53.9% (n=69).
A third study by Zhou et al. involved a random sample of 217 dictated notes from 144 unique physicians who used BESR. The results showed an error rate of 7.4%. The error rate decreased to 0.4% after healthcare documentation specialist review and to 0.3% after physician signature. Discharge summaries had higher mean error rates than other types of notes (8.9% versus 6.6%). Deletion errors were the most common at 34.7%, followed by insertion errors at 27%.
The third study illustrates how important is for physicians to review notes thoroughly prior to authentication. Examples of clinically significant errors from that study follow:
| SR Technology | Intended |
|---|---|
| Verapamil 4 mg po daily | Rapamune 4 mg po daily |
| Adequate evaluation to exclude neoplasia | Inadequate evaluation to exclude neoplasia |
| Allergies: yellow dye | Allergies: furosemide, gabapentin, oxcarbazepine, yellow dye |
| The patient did agree with this recommendation | The patient disagreed with this recommendation |
| There is no distal biliary obstruction observed | There is distal biliary obstruction observed |
Consider the following to minimize SRT-associated risk:
- Train practitioners on appropriate SRT use and ensure that the accuracy of speech recognition is validated for each user prior to using the software. Practitioners should attain a pre-established recognition rate percentage prior to using SRT.1
- Educate practitioners on SRT-associated errors and on strategies to reduce them.
- Establish practical workflow policies and procedures for BESR and/or FESR.
- Ensure that all stakeholders know and understand the policies and procedures.
- Encourage and ensure thorough and timely review of all notes.
- Develop clinical quality assurance and auditing programs for SRT.
- Provide feedback on audit results to all involved in the process.
References:
Speech Recognition in the Electronic Health Record (2013 update). AHIMA. Updated 2013. http://library.ahima.org/doc?oid=300181. Published September 2013. Accessed June 16, 2021.
Copyrighted. No legal or medical advice intended. This post includes general risk management guidelines. Such materials are for informational purposes only and may not reflect the most current legal or medical developments. These informational materials are not intended, and must not be taken, as legal or medical advice on any particular set of facts or circumstances.