Calculation of Accuracy Rates for Clinical Coding

Following the 1996-1997 SMR01 assessment, Medical Records/Information Managers suggested that the accuracy rates for clinical coding produced in SMR01 assessments did not take account of the fact that information available to coders is often limited.  In the 2000-02 SMR01 assessment, in collaboration with hospital staff we identified as clearly as possible the information used by coders at the time of coding.  By doing this, two accuracy rates could be provided for each clinical data item:

• the standard accuracy rate, based on all the relevant information available in the case-note
• the coder accuracy rate, based on the information in the document(s) coders had available at the time of coding when this documentation is subsequently available to DQA.

Basis of Accuracy Calculations
From the 2000-02 SMR01 assessment the calculation of accuracy rates has given coders credit for the number of decisions they have to make when coding a record.

Standard accuracy rates are calculated using:

• the number of times a code was recorded correctly for Main Condition, Other Conditions, Main Operation/Procedure and Other Operations/Procedures
• the number of times coders had correctly made the decision to leave the next clinical code field blank i.e. to stop coding

For example, a discharge summary states that the patient "fractured his right distal radius falling from a cliff ".  There are no co-morbidities to be recorded.  For this record, DQA would include in the accuracy calculations the first three diagnostic positions, including the correctly blank DG3:

	Diagnosis	ICD10 Code
DG1	Fractured distal radius	S52.50
DG2	Fell off cliff	W15.8
DG3		Coder correctly left blank
DG4		-
DG5		-
DG6		-

 

As the coder has no decision to take regarding positions DG4-6 these are not included in the accuracy calculations.

Coder Accuracy Rates
These are also calculated using the method above, but differ from the standard accuracy rates in that clinical coding errors are included in the accuracy calculations only when:

• the coders' primary coding document was available to DQA and
• DQA found that this document contained sufficient information to enable the coder to code the clinical data item correctly

In other words coder accuracy rates attempt to exclude errors in clinical coding when those errors are due to inadequate information reaching the coders.  The intention behind coder accuracy rates is to highlight capabilities of clinical coding staff and to identify any areas where information issues affects the coders' ability to deliver clinical coding of high quality.  For example, if the coder accuracy rate is noticeably higher than the standard accuracy rate for a particular clinical data item, this could indicate that the quality of the information available to the coders is affecting the quality of the SMR01 with regard to that item.

Minimum Target Rate for Clinical Coding Accuracy
The use of a minimum target accuracy rate of 90% for clinical coding at 3-digit level suggested by the Definitions and Quality Issues Group, the service-led group meeting at ISD, before the 1996-1997 assessment.  It was based on the average 3-digit level accuracy rate achieved by over 30 hospitals each taking part in four SMR01 assessments between 1989 and 1994.  There are no minimum target accuracy rates for clinical coding at 4-digit level or for non-clinical data items.