Getting data into an electronic health record is hard. Most clinicians do not want to type complex structured notes. We've tried macros and templates, which have helped some. However, dictation is the clear winner among clinicians for entry of free text.
The challenges with dictation are turn around time, expense, and lack of structure. BIDMC's solution to this problem as been server-based voice recognition from eScription.
Here's how it works:
* A clinician dictates into a phone or handheld
* The voice file is sent to the voice recognition server where we store the voice profiles of 3000 clinicians
* In near real time, the voice files are processed into text and inserted into the electronic health record as an unsigned note
* Correctionists (we no longer use transcriptionists) review the notes for accuracy. We achieve over 90% accuracy across all speakers and all note types
* Clinicians sign their corrected notes.
By using server-based voice recognition, we have reduced our transcriptions costs more then 50%, reduced turn around time to less than an hour, and used the technology to increase the structure of our free text notes. How?
When clinicians dictate operating room notes, history and physicals, outpatient notes, or radiology reports, they tend to dictate in their own preferred order, not via a universal template. Some clinicians may dictate chief complaint, history of present illness, physical exam, review of systems, assessment/plan. Others have completely different approaches. However, using voice recognition, we can recognize a key phrase like "physical exam" and automatically place it in a template. We've been able to increase the structure of our notes by 30% using voice recognition of subject headings.
Our total cost of implementation was about $500,000 and our savings over just the last year was over $1.5 million.
Happy clinicians, more structured notes, better turn around time and $5 million in savings since we implemented the technology. Now that's cool.