Voice recognition: The key to hospital dominance
By Tim K. Zinn, August 2010
Although still three to five years away, digitized voice recognition and input may hold the key to a comprehensive electronic patient clinical record.
Editor's Note: This is the 12th and final installment in our year-long 30th anniversary "Pioneers in Healthcare IT" celebration, featuring articles from past issues of Health Management Technology, formerly called Computers in Healthcare. This article appeared in the December 1991 issue. At the time, Tim Zinn was president of Zinn Enterprises Ltd., a Chicago-based hospital information system consulting organization. A graduate of Harvard, Zinn was a nationally known healthcare consultant and futurist specializing in healthcare trends.
Digitized voice recognition has far-reaching ramifications for healthcare information systems (HIS). It holds great promise in the hospital setting, where most professionals are working with their eyes and hands continually. This approach will allow doctors and nurses to enter information about their tasks while they are actually working, rather than relying on their memories to record accurate facts at a later time.
Documenting while doing is especially advantageous for those charged with making detailed and accurate records that are both timely and critical in nature, a situation physicians and nurses encounter regularly during the course of patient care. Further, information can subsequently be sorted into multiple data formats with only one handwritten copy of doctors' orders. Clearly, this form of data entry will not only be more convenient, but will capture close to 100 percent of clinically necessary information, as opposed to the 40 percent that current HIS automation now captures.
Primary challenge for the year 2000
The primary computer challenge facing healthcare administrators today is to contain and control the enormous flow of clinical information, disseminating this data in an orderly manner.
Currently, hospitals maintain three forms of information for every patient: written records, such as charts, doctors' orders and lab results (hardcopy data); X-ray, MRI and CT scans (images); and dictation notes (voice).
State-of-the-art computer systems will translate all of this information into a single format that can easily be transported within the hospital, made available to clinicians (both on and off campus) and transmitted to payers, such as insurance companies, self-insured businesses and governmental agencies. If this concept sounds familiar, it is, because we have just defined the concept of the electronic patient clinical record.
At Zinn Enterprises, we are continually conducting surveys of hospital senior executives to determine their perspectives on computing technology. When asked to rank emerging technologies relative to their importance for the future success of their institutions, CEOs ranked the electronic patient clinical record extremely high, while digitized voice was rated a distant eighth. This indicates that even the most perceptive and perhaps most experienced CEOs have failed to grasp the importance of digitized voice recognition as a critical key to the success of the electronic patient clinical record.
Status of voice recognition today
Now in their fourth generation, digitized voice systems have come a long way from the initial vocabularies that were limited to a few dozen words. Current systems are capable of deciphering up to 30,000 words, thanks to the larger 386 and 486 machines now available. System designers are also realizing that each system only needs to hold the vocabulary necessary to produce certain records and transactions. This will allow healthcare user interface for data entry and dictation in radiology and pathology, results reporting, inquiry, encounter records and the emergency room.
Early systems were speaker dependent and required considerable training to trace each individual's speech patterns. Newer systems are speaker independent and require little or no training time. One major concern is the concept of continuous speech. Current systems have a 1/8- to 1/4-second pause between words, allowing 40 to 50 words-per-minute dictation. Again, proper training increases accuracy.
Besides technological limitations, one must also consider the human factor. Digitized voice recognition works best when physicians and nurses are trained to give reports in standard formats. The restriction sometimes presents problems for trained clinicians.
Today, hospitals are struggling to maintain their role as the central focus of healthcare diagnostics and services. If hospitals are to continue in this capacity, they must embrace the development of electronic clinical patient records using voice recognition.
- Developing networks: As vendors develop the workstation concept to solidify internal data and linkages to ensure uninterrupted data flow, hospitals must plan for the future. We are facilitating strategic and action plans for our most sophisticated hospital clients to develop networks to move large volumes of data efficiently and effectively throughout the hospital and its clinics and labs.
- Capitalize on medical staff: Our clients are also capitalizing on an increasingly computer-literate and demanding medical staff who are finding it advantageous to collaborate with HIS administration to implement systems that supply them with the records and results they require in a timely manner.
- Information flow: Finally, we are working with our clients to facilitate the flow of information from care provider to payer (third-party payers, the government and benefits managers), a concept that will ultimately benefit the hospital through improved receivables.
One technological development we cannot ignore is the development of ISDN, a broadband, high-speed transmission capability to send voice, text and images over the same line. Although few hospitals in the United States currently have this capability, ISDN will be essential for the hospital to transmit large volumes of data outside the hospital in the long run. These high-speed lines will carry information many times faster than regular telephone lines and facilitate records transactions.
In summary, serious application of voice recognition will not truly come for another three to five years, but these technologies are definitely on the horizon.