Speech Recognition: Decreasing Documentation Costs Without Impacting Physician Practice Patterns

1
Speech Recognition Decreasing Documentation
Costs Without Impacting Physician Practice
Patterns
AZ HIMSS June 3, 2005
Robert ShollinVersatile Information Products
2
Objective

• Implement Speech Technology
• Improve Transcription Productivity
• Save Time and Money
• Dont Impact Physicians Negatively

3
Agenda

• What is the Need?
• Speech Recognition Primer
• Why Give It To Transcription?
• Why is the New Technology Different?
• Case Study Examined
• Questions/Open Discussion

4
What Is The Need?
5
Transcription Trauma

• Transcription Need is Increasing
• Domestic Labor Is Shrinking
• Off-Shore Labor Force Is Too Small

6
Supply and Demand

• Five Year Transcription Outsourcing Forecast

9 Billion Transcription Outsourcing
7
Supply and Demand

• Five Year Transcription Outsourcing Forecast

Domestic Labor Force 5 Billion
8
Supply and Demand

• Five Year Transcription Outsourcing Forecast

Off-Shore Labor Force 1 Billion
9
Supply and Demand

• Five Year Transcription Outsourcing Forecast

Void 3 Billion
10
Need is the mother of invention

• Transcription Need is Increasing
• Domestic Labor Is Shrinking
• Off-Shore Labor Force Is Too Small
• TECHNOLOGY MUST EVOLVE

11
Speech Recognition Solution
12
What is Speech Recognition?

• Automatic Conversion of Spoken
• Language into Text

Recognition
Speech
Text
Knowledge Sources
13
Traditional Dictation Work Flow
Dictation
Transcription
14
Transparent to the Physician
Front End or Interactive
RESISTANCE
15
Applications of Speech
Back End
16
Work Flow Diagram
Step 1 Capture Dictation
17
Work Flow Diagram
Step 2 Submit Recording
18
Work Flow Diagram
Step 3 Recognition Is Performed
19
Work Flow Diagram
Step 4 Submit For Correction
20
Work Flow Diagram
Step 5 Adaptation Is Performed
21
Why Give it to Transcription?

• Dictators Dont Do Anything
• Transcriptionists Are Trained Document
  Specialists
• Less Expensive To Implement
• Who would you rather pay to proofread?

22
Measurable Outcomes
1,200
1,200
1,200
1,200
1,200
1,200
23
Measurable Outcomes
1,200
1,200
1,200
1,200
1,200
1,200
1,200
1,200
1,200
24
Positive Results

• Reduce Turn Around Time
• Reduce Cost of Documentation
• Eliminate Costly Outsourcing
• Increase Transcription Bandwidth

25
Why is the new technology different?
26
U.S. Speech Recognition Market

• Past Successes Failureswhy?
• Poor Recognition
• Long Training Times
• All or Nothing Usage no middle ground
• Many looked at it, some bought it, few used it
• Force Change on the Physician
• Weve spent years injecting assistants into the
  physicians day to free them from everything but
  patient care
• Reimbursements decreasing, salaries declining

27
Convince Me

• Failure, failure, failure
• So why is it any different today?

28
Whats Changed?

• No Enrollment
• Highly Accurate
• Intelligent Speech Interpretation
• Transparent to the Physician
• Seamless and Productive Editing by a
  Transcriptionist

29
No Enrollment

• Evaluation Mode
• Learning physician recognition at no cost to you
• Both the Transcriptionist and the SR system
  transcribes the report
• Result compared
• Accuracy Percentage derived
• Advanced algorithms analyze
• Physician Acoustic Reference File automatically
  updated for future dictations (80)

Whats changed?
30
Highly Accurate

• Technology Improvements (Dual Pass Tri-grams)
• Automatic Acoustic Adaptation
• Recognition increases automatically

Whats changed?
31
Intelligent Speech Interpretation

• Focus on what was intended, not just what was
  said
• Automatic Formatting
• Automatic Punctuation
• Physicians may say they dictate punctuation,
  but the fact is that theyve never had to worry
  about it before now.
• Automatic Filtering
• Remove non-word utterances the ummms, ahhhs,
  ahems, coughs, etc.

Whats changed?
32
Intelligent Speech Interpretation(focus on
what was meant, not just what was said)

• Automatic punctuation
• Repeated phrases
• Abbreviations
• Formatting dates numbers
• Enumerated lists (e.g., 1, 2, next)
• Imbedded commands (e.g., bold, all caps)
• Unnecessary words (E.g., The following are the
  medications.. instead of Medications)

• Headings
• Carriage Returns
• Noise AND Silence

The ability to speak freely
Whats changed?
33
Seamless Editing

• Punctuation edits are eliminated
• Ummms, Ahhhs, silence removed
• Voice Cursor/Mouse Cursor Synchronized
• Recognition improves from corrections made
• Confidence level indicators provided
• Separate playback and editing cursors
• Use of traditional transcription productivity
  tools
• Performance has to be measured for both typed and
  speech recognized reports

Whats changed?
34
Whats Changed? Recap

• No Enrollment
• Highly Accurate
• Intelligent Speech Interpretation
• Transparent to the Physician
• Seamless and Productive Editing by a
  Transcriptionist

35
Case Study Examined
36
Case Study

• Medical Center in Detroit
• 3,000 Physicians
• 36 Transcriptionists
• In-Source and Out-Source
• Objective
• Eliminate Outsourcing Dependency

37
Case Study

• Measurements
• Track Lines Per Hour
• Without Speech Recognition
• With Speech Recognition
• Compare Gains
• Enroll More Physicians

38
Outcome Report Lines/Hr
39
(No Transcript)
40
Variables For The Equation

• Physician Statistics
• Some Physicians Will Not Work
• Transcription Statistics
• Some Transcriptionists Will Not Adapt
• Report Type Statistics
• Some Are Better Than Others

41
Presentation Extra The Dos and Dont of what
to look for
AZ HIMSS June 3, 2005
Robert Shollin
42
Do Look for

• An Evaluation Learning Model the ability to
  qualify and improve recognition with no physician
  or transcriptionist involvement (Critical for
  Initial Implementation)
• Automatic Adaptation (Show Me!)
• Acoustic (Physicians Voice Habit) Adaptation
• Context (Dictionary) Adaptation
• Background Capable Adaptation still occurs when
  a transcriptionist (not just the physician) does
  the correction
• Recognition Accuracy Improves progress
  reporting available
• Comprehensive Document Workflow the ability to
  electronically route (output) documents
  throughout your facility based on rules-based
  criteria

43
Do Look for

• Automatic Report Formatting Keyword Layout
• Automatic Punctuation
• Hesitation and non-word utterance handling
• Incentive Programmable performance incentives
  (hourly performance ratings) automatically
  calculated whether typed manually or
  corrected.
• Highly Efficient Correction
• Confidence Level Indicators
• Ability to turn audio AND/OR visual indicators on
  or off
• Ability to edit and listen at the same time (Dual
  Cursor one for audio footpedal, one for
  editing keyboard/mouse)

44
The Donts

• Dont assume foreign accent physicians will not
  workyou may be very surprised.
• Dont select a system without seeing all of the
  key features discussed, demonstrated, and
  understood.

45
The Donts

• Integration Dont buy a new dictation system
  without assuring the ability to record at 64 or
  128 kbps (speech recognition grade voice capture)
• Dont buy a transcription system without
  confirming and detailing the parameters and cost
  for adding speech recognition to it.

46
Demonstration

• Captured A Dictation
• Speech Recognition Was Performed
• Submitted For Correction

47
Thank You

• Robert Shollin
• General Manager
• Versatile Information Products, Inc
• 800.794.4044 Bosho_at_Earthlink