Information Technology Reference
In-Depth Information
and process and how best to visualize clinical data to support providers and patients
in achieving the best results.
Collecting the Data:
Efficiently collecting clinical data from providers and prop-
erly coding it is a very difficult problem. The human body is extremely complex. It
consists of many parts that are involved in many processes. It is subject to many
disorders and malfunctions. As a result, SNOMED CT, the
subset
of SNOMED for
electronic health records, has some 311,000 concepts connected by some 1,360,000
links or relationships. How can a busy physician with limited time be expected to
accurately navigate something this complex for
every clinical concept
in
each
patient's chart? Obviously, they cannot. This problem was essentially not resolvable
in real clinical practice until recently, in large part because computers weren't pow-
erful enough to help. As a result most electronic charting was done either by input-
ting free text via typing or dictation or by using templates - essentially paper check
off forms transferred into a computer. Today these are still, by far, the most common
methods for collecting the subjective and objective clinical data obtained from patient
interviews and physical examinations. The majority of commercial electronic health
record systems use these techniques but there are innovative approaches that take
advantage of the power of modern, yet still relatively inexpensive, computers.
M*Modal
began in 1998 when a group of three PhD students (Michael Finke,
Detlef Koll and Juergen Fritsch) from Carnegie Mellon University's highly ranked
School of Computer Science spun voice recognition technology they had been
working on into the commercial space by forming Interactive Systems. In 2001 the
company was renamed M*Modal and the focus was narrowed to healthcare. It was
acquired by MedQuist in 2012 and the entire company took the M*Modal name.
It now offers a suite of web services under the name M*Modal Fluency. Specific
examples include Fluency Direct (to speech-enable EHRs), Fluency for Imaging
(radiology-focused reporting solution), Fluency Mobile (iPad & iPhone based clini-
cal documentation), Fluency for Transcription (traditional back-end clinical docu-
mentation service based on human editing), and Fluency for Coding (billing coding
workflow with predictive code generation from narrative text).
The technology had for years been very widely used by medical transcription
services around the globe. The benefit to the services was that most of the transcrip-
tion was done automatically so expensive human labor was only needed to correct
what the computer got wrong (it tells the humans what it's not certain of) or cannot
transcribe at all (it also points those cases out).
Every instance of use goes back to the company's servers which observe the cor-
rections humans make and use that knowledge to further train the system which,
over time, gets better even with various speaking styles, languages and dialects.
It can now recognize text sufficiently well to put it into the proper section of a note
so that, for example, the chief complaint go into that part of the note while the
physical exam, assessment and plan go into their respective parts.
If that weren't remarkable enough, in many instances clinical concepts are
located in the text and are coded into structured medical terminology such as
SNOMED CT and into various coding systems for medications and into ICD, CPT
