Automation of Document Type in the Document Identification Queue

Overview

One of the challenges in healthcare documentation is to get the correct Document Type (work type) on each dictation as it enters the platform.  When that does not happen, the transcriptionist must cope with the extra effort and delay associated with changing the document type.  First Draft Dictation Recognition system (speech recognition) also struggles to provide the best recognition when the document type is not known.  First Draft is an optional piece of the InfraWare 360 platform (KB15). The Document ID Queue (KB424) was designed to allow an administrator to listen to a few seconds of dictation to identify the work type BEFORE being processed by First Draft.  The automation process can automatically select the Document Type.  This process is even aware of the margin of error for each Author, so managers can specify how often to use the predicted document type. 

Use of this feature has the potential to provide more First Draft opportunity, improve the quality of First Draft output, and decrease the number of FD Discards (KB573).  

Requirements

• InfraWare platform user with the Administrator or System Administrator Role assigned (Administrator Role is required to access the First Draft tab).
• Models for Automation of Document Type are created only when a high percentage of jobs have Document Type changes.

Location

Management of the Document Identification Queue and Automation of Document Type in the Document Identification Queue can be done in two places:

• Author Settings:  Administration tab ⇒ Users ⇒ Click name to select ⇒ Click the Author Settings link in the Facility Roles section ⇒ Scroll down to the Document Identification Queue section.  This setting is done at the Author level, so if an Author is set up for multiple Facilities Document Identification Queue and/or Automation of Document Type in the Document Identification Queue for this Author on one Facility will be set for all Facilities. 
  
• FD Author Details:  First Draft tab ⇒ Author Details page ⇒ Click name to select ⇒ Scroll down to the Document Identification Queue section.
  
  

Options

• Automatically Predict Document Type:  Selecting this option will enable Automation of Document Type in the Document Identification Queue for this Author.  
• Allow Auto Doc ID Queue to change Facility:  Selecting this option will allow the automation process to change the Facility as well as the Document Type. 
• Confidence Threshold:  The Confidence Threshold is the margin for error in choosing the document type.  This is determined on a per-Author basis and for each threshold value and includes an estimate (in parentheses) of total volume for this Author which meets each Accuracy percentage.  Choose from the drop-down menu.
  
• For the ___% of jobs below ___ Accuracy:  When setting determines the process for jobs which have a Confidence Threshold below what is set above. 
  • Use Normal Document Identification Queue Behavior:  This is the default.  If this option is set for an Author the Document Identification Queue settings will be observed for the the timeout threshold and whether or not a job which times out will go to First Draft.  See KB425 for details. 
    
  • Use Best Guess and Route to First Draft if enabled:  If this option is selected the process would choose the best guess for Document Type and route the job to the next workflow step.  If the Author-Document Type-Dictation Source is enabled for First Draft the job would be processed by First Draft.  If not enabled the job would go to Transcription.  Basically, with this setting selected the feature will come in at the last possible second and assign the job to the Document Type best guess. 

Notes

• If no Dictation Source is enabled for the Document Identification Queue the Automation option will be grayed out and cannot be enabled. 
  • If the Dictation Source is enabled for the Document Identification Queue and the option for Automation is still grayed out there is no Document Type Automation Model available.
    
• This feature is not dependent on use of First Draft back-end Speech Recognition Services.  The algorithm processes data about the job and determines the best classifiers that are best for this Author (combination, one, etc.).  Some use speech recognition output and some do not, so an Acoustic Model is not required for use of this feature. 
• At least 20 jobs are needed in order to gather metrics.  
• The confidence score will be evaluated on a per-job basis for jobs which enter the platform once the feature is enabled for an Author.
• The threshold percentages for Accuracy of volume will vary by Author and may change over time.