Advancing Excellence in Health Care www.ahrq.

gov
Agency for Healthcare Research and Quality
Health IT
Improved Accuracy of
could improve coding accuracy through better
documentation and coding-decision support to
Coding
providers, translating into enhanced revenue.
A rural, family practice implementing a commercial
electronic medical record (EMR) with an electronic
Monitoring the use of current procedural
billing system developed an average practice case mix
terminology (CPT) codes can help organizations
based on evaluation and management codes (E/M),
determine whether health information technology
with higher weights reflecting increased
(health IT) improves coding accuracy and
reimbursement.
4
They found that their average case
completeness by providing decision support for
mix increased by 10 percent over the 2-year,
documentation activities. This can subsequently lead
postimplementation period, from 1.34 to 1.47
to enhanced revenue because visits could be coded
(p<0.01), suggesting that patient visits were
with higher complexity level-of-service billing codes.
undercoded pre-EMR implementation. This change
Measure Category: Financial Impact had a financial impact: they reported that average
monthly revenue increased 11 percent in the first
Quality Domain: Efficiency
year and 20 percent in the second year.
Current Findings in the Literature: Billing in
In another study, researchers evaluated the impact of
health care organizations is based primarily on CPT
an EMR at a large, multispecialty ambulatory
codes. These codes reflect the level of service for the
practice.
5
The EMR provided templates to help
patient visit and are based on multiple factors,
better document the level of care by recommending
including the severity of the problem, the complexity
the appropriate evaluation and management code
of the medical decision, and the duration of the
based on the documentation This increased the
examination. However, coding guidelines are
accuracy of coding by reducing the incidence of
complicated and substantial disagreement often
1
under- and overcoding. The study compared the
exists over assigning codes. Studies have found that
percentage of patient visits that were coded 99213
coding is accurate approximately half of the time,
(problems of low-to-moderate severity) and 99214
while the rest of cases are under or overcoded.
2,3
(problems of moderate-to-high severity) and found
One explanation for inaccurate and incomplete
an 11 percent overall increase in the use of 99214
coding could be the complexity of the coding system
codes that would previously have been coded 99213.
and a poor understanding on the part of clinicians
They estimated that this decrease in downcoding
regarding its use. In general, the more highly
would produce an average billable gain of $26 per
compensated encounters require the most rigorous
patient during the study period, for increased
documentation. Therefore, providers may be overly
revenue of $103,059.
conservative to protect themselves from fraud and
Source of Data for the Measure: Electronic
downcode the encounter, choosing the lower
Medical Record Data, Financial or Billing Data from
compensated reimbursement code. This
Practice Management Systems.
downcoding translates to lost revenue. Health IT
Methodology for Measurement
Relative Cost: Low, as most organizations already
track this billing information electronically, although
may be higher if a billing expert is needed to review
Study Design: Pre- and post-health IT imple-
coding of encounters.
mentation
Potential Risks: Evaluators should be aware of a
possible Hawthorne effect. If providers know that
Study Design: Define baseline and intervention
their billing is being monitored for a study, they will
time periods (e.g., number of months). tend to be more careful with their coding. Billing
and reimbursement regulations are constantly
Evaluation:
changing and new billing rules go into effect often.
Examine the change in the percent-
Evaluators will want to consider how these changes
age of categories of CPT codes. Assign costs to
in billing rules might impact their measurement.
different code levels and extrapolate. Compare the
costs pre to postimplementation. References
1. K.2, M#, #-&54 L, L.46/; M#. A((85&(; 3+ C!$
Analysis Considerations *9&08&7.32 &2) 1&2&,*1*27 (3).2, '; +&1.0;
Several issues should be addressed before proceeding
4-;6.(.&26. J A1 B3&5) F&1 !5&(7 2001 M&;-
with an analysis plan:
J82;14(3):184-92.
1. Evaluators may want to have experts in coding
2. K./&23 GE, G33):.2 MA, #7&2,* KC.
review a sample of patient notes manually to
E9&08&7.32 &2) 1&2&,*1*27 6*59.(*6. A
confirm the reliability of coding in their
(314&5.632 3+ 1*).(&0 5*(35) )3(81*27&7.32
electronic environment.
:.7- &(78&0 '.00.2, .2 (31182.7; +&1.0; 45&(7.(*.
A5(- F&1 M*) 2000 J&2;9(1):68-71.
2. Your data collection and analysis plan should be
based on sound methodology. To achieve valid,
3. C-&3 J, G.00&2)*56 %G, F03(/* #A, *7 &0. B.00.2,
robust results, consider using the input of a
+35 4-;6.(.&2 6*59.(*6: & (314&5.632 3+ &(78&0
trained statistician to determine sample size and
'.00.2, :.7- C!$ (3)*6 &66.,2*) '; ).5*(7
appropriate statistical techniques. It is not
3'6*59&7.32. J F&1 !5&(7 1998 J80;47(1):28-32.
uncommon to begin analyzing data, only to
4. 'N*.00 L, K0*4&(/ %. E0*(7532.( 1*).(&0 5*(35)6
find the original statistical plan was flawed,
+35 & 585&0 +&1.0; 45&(7.(*: & (&6* 678); .2 6;67*16
leaving you with data that is inadequate for
)*9*0341*27. J M*) #;67 2007 F*';31(1):25-33.
analysis.
5. B&503: #, J3-2632 J, #7*(/ J. $-* *(3231.( *++*(7
3. A simple chart or graph that visually displays
3+ .140*1*27.2, &2 EM" .2 &2 3874&7.*27 (0.2.(&0
changes in coding over time is an effective way
6*77.2,. J H*&07-( I2+ M&2&, 2004
to communicate this information to
%.27*5;18(1):46-51.
stakeholders.
AHRQ Publication No: 09-0095
September 2009