I remember going to see our family doctor many years ago; I must have been about 14 years old. The visit was probably for something like a sore throat. When he completed the physical exam he wrote a paper prescription and handed it to me; he then reached for a black index-card box about the size of a shoe box. In it were two rows of neatly stacked index cards, one for each of his patients. He had the index cards of all the members of a nuclear family neatly attached to one another with a paper clip. I always fondly think of the symbolism of that connection. He must have written something like the date, “pharyngitis” and “Pen VK 500 for 10 days.” The visit, and its documentation, were then over.

Many years later as a practicing primary care internist, a member of the University of Rochester’s Medical Center’s Primary Care Network, I follow many of the same rituals. Today, however, I have the patient’s electronic medical record (EMR) open in front of me on a computer screen; I electronically prescribe a medication, and receive alerts electronically of any drug interactions or allergies. I provide the patient with a printed information sheet regarding their condition. I then step out of the room into my office, open a note in the EMR, put my microphone headset on, and dictate the narrative sections of the note using speech recognition software. Indeed much has changed.

EMR Implementation at URMC

In 2005, URMC selected an EMR as its ambulatory medical record and began implementing it to approximately 500 physician users and close to 1,500 general users. A survey recently published in the New England Journal of Medicine (NEJM) reported that only 4 percent of close to 3,000 respondents had EMRs with full functionality. URMC, being a large academic medical center, was typical of this group. With the implementation of the EMR there have been significant gains in safety, efficiency and cost reduction at URMC. There also has been very strong physician adoption along with increased patient and physician satisfaction.

Documenting the Clinical Encounter

Documentation of the clinical encounter has undergone a gradual evolution. Recording the history, physical exam, assessment and plan were traditionally undertaken to allow physicians to recall these facts at a later date. Allowing for seamless continuity of care, the note was also a means of communication among different physicians caring for the same patient. Having this information allowed physicians and members of the healthcare team to care for the patient with knowledge of previous evaluations and treatments. The medical note also serves as a legal document, describing the course of care provided to the patient.

In the mid 1990s, the Health Care Financing Administration (HCFA) introduced the first version of the “Documentation Guidelines for Evaluation and Management (E&M) Services,” which dictate the documentation standards that need to be met for physicians to justify the level of compensation for a particular service, thus turning the medical note into a key billing tool as well. In latter years, E&M guidelines are among the driving forces shaping the content and form of medical documentation, sadly sometimes even more than patient care needs and considerations.

One of the core functions of an EMR is to provide task-specific tools for documentation of the medical encounter; this is usually provided by a Note module. Many of the notes produced within these modules rely heavily on text templates or templates using discrete codified data. Unfortunately because of the characteristics of these tools, the notes produced tend to have a uniform “look and feel” with very little variation among different documenting clinicians and patients. These “homogenized” notes are no longer as helpful or effective in the clinical process. A recently published NEJM perspective titled “Off the Record — Avoiding the Pitfalls of Going Electronic” states that patients’ narratives, and clinical and personal stories, have become lost in a sea of templated, “canned,” repeated chunks of text and verbiage.

To balance these considerations, we have developed “best practices” that support a hybrid form of documentation. They combine the best of both worlds — the “narrative” and the “templated” — allowing for variations that stem from disparate specialties’ characteristics, and those specialties’ research agendas and needs.

For example, our pediatric division is heavily involved in translational research and has structured most of its note forms as templated, mineable documents. However, our primary care network has developed note forms that combine narrative sections that describe the patient’s history, and the assessment and plan formulated by the physician, but also allow liberal use of research-accessible templates for all other sections of the note. It is in this “hybrid model” of documentation that speech recognition has thrived at URMC.

Coming of Age

Prior to URMC’s 2005 EMR implementation, there were individual physicians who experimented with an early version of the speech recognition software. Those brave early adopters enthusiastically would start using the software, but would invariably become frustrated by its inaccuracy, which led a few months later to abandoning the software. The pattern repeated often and bills for traditional transcription service again increased.

Combining these two technologies has provided physicians with tools that capture the patient’s story in narrative form, within a highly structured “mineable” framework.

The full-scale implementation of the EMR at URMC, and the introduction of version 9 of Nuance’s Dragon NaturallySpeaking Medical software, presented a unique opportunity to re-evaluate and combine these two tools, this time with great success. The voice recognition software had come of age with remarkable accuracy, and a complete set of medical vocabularies that collectively span many specialties.

Prior to the EMR implementation, URMC’s primary care network relied on the medical center’s transcription service vendor for most of its transcription needs. A few practices had relationships with small, independent vendors adhering to varying standards in terms of turn around times and accuracy. The one constant was the high cost of these services. It is estimated that the primary care network alone (just more than 100 providers) was spending well over $1 million a year on transcription, not calculating the cost of managing the flow of transcription and staff involved in printing and filing these transcriptions into the paper charts.

By the middle of 2007, a critical mass of primary care offices had implemented the new software and they were reporting great results combining the EMR’s Note module with the speech recognition software. Physicians were also reporting significant flexibility creating personal efficiency enhancing “macros” (short cuts). Interestingly, this was a “grass roots” IT paradigm shift, driven by early adopting physicians rather than the institutional IT strategic plan.

Responding to these shifts late in 2007, the primary care network leadership chose to fully adopt the combined documentation model and discontinue using external transcription services. For a majority of physicians this was a relatively easy transition. However, a small group of physicians struggled with the decision to fully convert to speech recognition for dictation within the EMR. They were concerned that using voice recognition software shifts the onus of proof reading and correction to the physician from the professional transcriptionist, and that adding any tasks to already overburdened physicians could result in errors of transcription, reducing the accuracy and validity of the documentation.

A handful of physicians also were overwhelmed by the shift from paper to the EMR and felt they could not undertake another “electronic adventure.” This group was provided with individual training and work sessions by physician colleagues who helped them to work through hurdles in the use of the software.

Interestingly, it became clear early in this training that the main barriers to adoption had more to do with deficits in basic computer skills, rather than challenges inherent to the more sophisticated software programs.

Success

In February 2008, URMC completed the transition of the primary care network to the combined documentation utilizing voice recognition and the EMR Note module. No line item for transcription costs is included in our primary care 2009 budget.

ROI not withstanding, combining these two technologies has provided physicians with tools that capture the patient’s story in narrative form, within a highly structured “mineable” framework. Each note has the potential of capturing what is unique to patients at a specific moment. It also enables physicians to articulate the thought process behind their diagnostic and therapeutic decisions. Those responsible for ensuring the accuracy of the charge process also find it easier to audit notes that have a narrative component.

The increase in EMR adoption nationally may eventually render the current E&M framework obsolete; some even theorize that it will be the narrative description of the clinical process by which level of service is measured in the future. Those clinicians using a combination of EMR and speech recognition will be optimally positioned to meet that new day.

September 2008