Skip to content Skip to navigation


September 1, 2007
by Andrew Ury
| Reprints
Practices can speed productivity and reduce costs by adding speech recognition to their EHRs

As physicians, we are constantly bombarded with information on how technology can help us improve care and reduce costs. The problem is, just managing the day-to-day demands of running a successful practice often leaves us too strapped for time to fully implement the technology we already have, much less evaluate new releases.

The potential benefits of using speech recognition, particularly in conjunction with an electronic medical record (EHR) system, are well worth the short amount of time you'll need to spend evaluating, deploying and learning the software.

Dramatic improvements in speech recognition technology have increased its accuracy rate and ease-of-use. As a result, demand has been growing steadily among physicians in search of tools to improve both work-flow processes and quality of care.

Two reasons for the growing popularity of speech recognition among physicians are its affordability and the ease with which it can be deployed within daily practice. As a result, even those among us who are less comfortable with advanced technology can tap into the benefits of speech recognition, including:

  • Reduced costs, particularly through the elimination of transcription expenses that can cost up to $17,000 per physician per year.

  • Faster per-dictation turnaround time, which can take on average up to 48 hours with manual transcription.

  • Increased accuracy, particularly with the latest versions that are 97 to 99 percent accurate.

  • Fast electronic capture of information on complex cases beyond the scope of traditional EHR templates.

Best of all, speech recognition is flexible enough to fit each physician's needs and preferences. Some prefer to use it to supplement the data they gather through EHR templates, such as expanding progress notes to accommodate multiple conditions or to capture personal information that may be impacting a patient's health. Others prefer to use speech recognition as their principal means of data collection — particularly if they are not comfortable typing or clicking through screens with a mouse.

Dan Engeberg, M.D., one of six physicians with Hanford Family Practice Associates in Hanford, Calif., is an example of someone who has realized benefits from using speech recognition in conjunction with the group's EHR.

Engeberg says he is "keyboard challenged" and dislikes the repetitive clicking and scrolling required to move through templates. He says it's also frustrating to make the mental switch from patient encounter to templates. So a few years after Hanford Family Practice implemented an integrated practice management and EHR system, he decided to try speech recognition after seeing how it worked for another physician.

After selecting a product that integrated with the EHR system in place at Hanford, Engeberg taught himself the basics. Eventually, he decided he could use some specialized training.

Today, Engeberg is using speech recognition for logging into his system and progress notes, letters and e-mails. In addition to voice commands to perform EHR tasks such as looking up patients, opening charts and accessing provider schedules, he has programmed the speech recognition software to:

  • Record free-form text in the progress note.

  • Use voice commands to insert clinical information into the note from other parts of the chart, such as problem list, medication list and past medical history, and export information from the current progress note to automatically populate other areas of the chart.

  • Navigate through progress note templates and expand macros that contain commonly used clinical phrases.

  • Navigate from one folder to another, such as moving from the progress note to the medications folder.

The primary consideration for or against speech recognition is whether or not you and your EHR are both stylistically suited to its use.

In terms of the EHR, speech recognition integrates best into systems that are progress note-centric. In other words, if your EHR does not require wading through numerous secondary screens to reach the progress note, it is well-suited for speech recognition. However, if your EHR uses many secondary screens, the benefits of speech recognition are somewhat muted. Although speech recognition can be used for commands to move through the various screens, using a mouse or pen is a faster method for doing so.

A note-centric system can also use speech-entered information to populate other parts of the patient's record, such as problem and medication lists, once that information has been processed and saved through the progress notes. This allows you to retrieve discrete elements of data even if you used speech recognition technology to enter it.

In terms of your own personal style, the main issues are how comfortable you are with the concept of speech recognition and how tolerant you are of errors.

Today's speech recognition software has achieved what I call the Holy Grail of functionality — it has a large vocabulary, it's speaker-independent and it allows for continuous speech. As a result, the training that once took an hour can now be accomplished in about 10 minutes.

Yet speech recognition isn't perfect. There is still an error rate of 2 to 3 percent in terms of accuracy, and it takes time and practice to become fully comfortable with the process. How tolerant you are of those errors and the time it takes to train — and be trained — on the software will obviously impact your overall satisfaction. However, if you're comfortable dictating, are using a note-centric EHR and you can talk faster than you can type, speech recognition is worth exploring.

Andrew Ury, M.D. (