Accident Analysis and Prevention 40 (2008) 714–718

Coding external causes of injuries: Problems and solutions

Kirsten McKenzie ∗ , Emma Enraght-Moony, Leith Harding, Sue Walker,
Garry Waller, Linping Chen
National Centre for Classification in Health, School of Public Health and Institute for Health and Biomedical Innovation,
Queensland University of Technology, Kelvin Grove, Qld. 4059, Australia
Received 8 March 2007; received in revised form 4 September 2007; accepted 10 September 2007

Abstract
Complete and accurate information about hospitalised injuries is essential for injury risk and outcome research, though the accuracy and reliability
of hospital data for injury surveillance are often questioned. To ascertain clinical coders’ views of the reasons for a lack of specificity in external
cause code usage and ways to improve external cause coding, a nationwide survey of coders was conducted in Australia in 2006. Four hundred
and two coders participated in the questionnaire. The results of this study show that discharge summaries and doctors’ notes were the poorest
source of information regarding external causes, place of injury occurrence, and activity at the time of injury. Coders viewed missing external cause
information and missing documentation as having the greatest impact on the quality of external cause coding. A large majority of coders suggested
that improving clinical documentation in the emergency department and introducing a centralised structured form for external cause information
would improve the quality of external cause coding. Clinical coders are a valuable source of information regarding problems with, and solutions to
the collection of high quality data and this research has highlighted several areas where improvements can be made and further research is needed.
© 2007 Elsevier Ltd. All rights reserved.

Keywords: Injury surveillance; External cause coding; Clinical coders; Survey

1. Introduction standardised format, enabling the collection, storage, and anal-

Injuries are an important public health issue in Australia with
injuries costing an estimated $4 billion across health sectors and
$2.8 billion for hospitals annually (Mathers and Penn, 1999).
Complete and accurate information about hospitalised injuries
is essential for injury risk and outcome research, and policy-
makers rely on high quality data regarding causes of serious
injury in the community.
Nationally and internationally, the International Statistical
Classification of Diseases and Related Health Problems (ICD)
system is used to categorise diagnoses and other health-related
data recorded in patient hospital records into alphanumerical
ICD codes (World Health Organization, 1994; National Centre
for Classification in Health, 2004). Use of ICD coded data facil-
itates the statistical aggregation of morbidity information in a
∗ Corresponding author at: National Centre for Classification in Health, School
of Public Health, Queensland University of Technology, Kelvin Grove, Qld.
4059, Australia. Tel.: +61 7 3138 9753; fax: +61 7 3138 5515.
E-mail address: k.mckenzie@qut.edu.au (K. McKenzie).
ysis of comparable data over time and between countries (Walker
and McEvoy, 2004). The external causes chapter (Chapter XX)
of the ICD describes the causes of injury, poisoning, and adverse
events. External cause codes capture different elements of the
circumstances surrounding injury events, including the causal
mechanism (e.g. transport accident, fall, etc.), the place of occur-
rence (e.g. school, work, etc.), the activity at the time of injury
(e.g. working for income, playing sport, etc.), and the intent
of the injured person and others (e.g. accidental, intentional
self-harm, assault, etc.).
Comprehensive information about hospitalised injuries
requires accurate clinical documentation and precise clinical
coding (Langlois et al., 1995; Lorence and Ibrahim, 2003). How-
ever, the accuracy and reliability of hospital data for injury
surveillance are often questioned, with researchers reporting
errors in the coding of cause of injury information (or exter-
nal causes) of between 13 and 18% of records (MacIntyre et
al., 1997; LeMier et al., 2001; Langley et al., 2006). Our pre-
vious study on the quality of cause-of-injury data indicated
that there is extensive use of ‘Other Specified’ and ‘Unspeci-
fied’ external cause codes in hospital morbidity data (McKenzie

0001-4575/$ – see front matter © 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aap.2007.09.008
 K. McKenzie et al. / Accident Analysis and Prevention 40 (2008) 714–718 715

et al., 2006). For example, about 11.5% of ‘Accidents’ and Table 1

18.6% of ‘Assaults’ were assigned to the ‘Other Specified’ and Summary of participant’s characteristics
‘Unspecified’ external cause codes, with non-specific external Variable N %
cause codes being assigned for over 47,000 hospitalised injuries Sex (missing n = 4)
per year and therefore, resulting in a loss of valuable injury Male 30 7.5
information. Overuse of ‘Other Specified’ and ‘Unspecified’ Female 368 91.5
codes may indicate a lack of clinical documentation to facilitate Work status (missing n = 6)
code selection, a lack of specific code availability, and/or coding Full-time 244 60.7
error. Part-time 145 36.1
Clinical coders are employed in hospital settings to assign Casual 7 1.7
ICD codes to accurately describe the patients’ diagnoses, pro- Educational level (missing n = 26)
cedures, and external causes as documented in the medical On the job 12 3.0
record. Clinical coders are trained professionals with expertise in Health department training 20 4.9
Distance education coding course 48 11.9
medical terminology, anatomy and physiology, clinical classifi- Health Information Management 123 30.6
cations, and health information systems. As such, clinical coders Association course
possess valuable expertise to provide insight into the reasons for Undergraduate degree 168 41.8
invalid or unreliable coded data, though there is a considerable Postgraduate degree 5 1.2
lack of research utilising the views of these health professionals.
Work place facilities
The aims of the current study were to ascertain clinical
coders’ views of the reasons for a lack of specificity in exter- Location Size Sector
nal cause code usage, particularly problematic areas for external Urban N = 192 Largea Public 85 23.0
cause coding and to ascertain coders’ views of ways to improve Private 19 5.1
external cause coding. Mediuma Public 34 9.2
Private 29 7.8
Smalla Public 3 0.8
2. Methodology Private 20 5.4
Regional N = 161 Largea Public 25 6.8
The survey questionnaire was based on the Australian clin- Private 4 1.1
ical coder survey (McKenzie and Walker, 2002), and modified Mediuma Public 46 12.4
in consultation with key project stakeholders to address areas Private 23 6.2
Smalla Public 41 11.1
of concern for external cause coding. A pilot survey was con- Private 18 4.9
ducted with a small sample of trauma coders and the survey was
Remote N = 25 Mediuma Public 4 1.1
modified and shortened based on feedback from these coders.
Private 0 0
The survey was web-based, with email and paper versions Smalla Public 13 3.5
available to allow surveys to be posted or emailed to clinical Private 6 1.6
coders where access to the Internet was limited. Clinical coders Total (missing n = 32) 370 100
were recruited via multiple channels with multiple follow-ups
a Large, >300 beds; medium, 100–300 beds; small, <100 beds.
to ensure the large majority of the coders in Australia were
informed of the survey and able to contribute to the survey.
Emails advertising the survey with a link to the website, where tralia. Four hundred and two coders Australia-wide participated
the survey could be completed, were sent by State Health Depart- in the questionnaire, representing an overall response rate of
ment representatives to all coders listed on their most current around 40%.
mailing list. The same email advertisement was sent to Code L,
which is an email discussion list available to coders for coding- 3. Results
related issues. An advertisement was placed both on the NCCH
website and in Coding Matters which is the NCCH publication 3.1. Respondent characteristics
for coders and Health Information Managers. An advertisement
was also placed in State Health Department correspondence Table 1 provides a summary of demographic characteristics
aimed at coders/health information managers. of participants. The average age of the coders was 41 years (range
Information about documentation sources in terms of the 21–65 years). Of the 402 coders, 62% (n = 244) were working
quality of the external cause codes, factors influencing cod- full-time and 37% (n = 145) worked part-time. Only 7 coders
ing, and opinions about improving external cause coding were indicated that they worked on a casual basis. A total of 169
obtained. Participants’ demographic information, their coding (41.9%) coders had experienced two or more different modes
history, workplace environment, and coding processes were also of coder training, and the most common education for coders
recorded. was an undergraduate university course (45% of respondents).
At the time of the survey, based on personal consultation with The average length of coding experience was about 10 years
key state health department representatives, it was estimated that (range 1–34 years). The participants spent 28% of their time on
there were approximately 1000 clinical coders employed in Aus- average in coding external causes in their overall coding task.
716 K. McKenzie et al. / Accident Analysis and Prevention 40 (2008) 714–718

Table 2
Quality of source documents for external cause information
Variable No information Poor information Average information Good in formation

n % n % n % n %

External causes
Ambulance report 8 2.1 36 9.4 121 31.8 216 56.7
Emergency department notes 19 5.1 72 19.3 152 40.8 130 34.9
Nurses notes 8 2.1 99 25.4 189 48.5 94 24.1
Doctors notes 10 2.6 140 35.9 160 41.0 80 20.5
Discharge summary 33 8.5 185 47.8 124 32.0 45 11.6
Place of injury occurrence
Ambulance report 12 3.2 53 14.1 117 31.2 193 51.5
Emergency department notes 22 5.9 121 32.7 146 39.5 81 21.9
Nurses notes 20 5.2 179 46.7 139 36.3 45 11.7
Doctors notes 19 4.9 214 55.6 124 32.2 28 7.3
Discharge summary 64 16.8 227 59.6 74 19.4 16 4.2
Activity at the time of injury
Ambulance report 21 5.6 90 23.9 143 38.0 122 32.4
Emergency department notes 23 6.2 135 36.2 143 38.3 72 19.3
Nurses notes 22 5.7 176 45.8 140 36.5 46 12.0
Doctors notes 24 6.3 213 55.5 121 31.5 26 6.8
Discharge summary 69 18.0 217 56.5 80 20.8 18 4.7

About 68% of participants worked in the public sector in large tion about place of injury occurrence and activity when injury
or medium hospital facilities. occurred.

3.2. Quality of external cause documentation sources 3.3. Factors affecting external cause coding

Table 2 shows the coders’ responses regarding the quality
of source information for external cause coding. Coders ranked
the quality of source documentation for external cause, place
of injury occurrence, and activity as providing good informa-
tion, average information, poor information, or no information.
Over half of the respondents (56.7%) stated that the highest
quality external cause documentation came from ambulance
reports, which were rated as a good source of external cause
information. In contrast, almost half of the respondents stated
that discharge summaries were a poor source of information for
external causes (47.8%). Similar results were found for informa-
To explore the factors that influence a coders’ ability to code
external causes effectively, coders were asked to consider a list
of factors that may have an impact on the accuracy, completeness
and timeliness of this coding. Coders indicated the severity of
each factor for coding external cause information on a scale from
no impact, low impact, medium impact to high impact. Table 3
shows the coders’ views of factors affecting the quality of coding
of external cause information, in order of the factors deemed to
have the highest impact. Coders rated missing external cause
information and missing documentation as the factors with the
greatest impact for external cause coding. These issues were

Table 3
Factors influencing external cause coding
Variable No impact Low impact Medium impact High impact

n % n % n % n %

Missing external cause information 1 0.3 32 8.2 52 13.4 304 78.1

Missing/absent medical documentation 13 3.3 87 22.3 89 22.8 202 51.7
Lack central forma 40 10.3 72 18.6 93 24.0 182 47.0
Illegible medical record entries 23 5.9 121 31.2 111 28.6 133 34.3
Pressure to maintain coding throughput 71 18.4 150 38.9 75 19.4 90 23.3
Lack of specific codes 34 8.7 181 46.5 93 23.9 81 20.8
Missing /incomplete ambulance reports 61 16.1 181 47.6 60 15.8 78 20.5
Structure medical recordb 93 23.8 176 45.1 63 16.2 58 14.9
Insufficient time and resources 73 18.8 184 47.4 78 20.1 53 13.7
Insufficient management support 108 28.1 177 46.1 52 13.5 47 12.2
Lack of ongoing education 97 25.2 206 53.5 51 13.2 31 8.1
Inexperience coding external cause data 130 33.3 194 49.7 37 9.5 29 7.4
a Lack of centralised form where all external cause information is recorded.
b Structure of records (e.g. emergency department or ambulance documents separate from inpatient).
 K. McKenzie et al. / Accident Analysis and Prevention 40 (2008) 714–718 717

Table 4
Coder satisfaction with ICD-10-AM by code blocks
Variable Not at all satisfied Low satisfaction Medium satisfaction High satisfaction

n % n % n % n %

Number of codes available

Accidents 1 0.3 44 11.4 173 44.8 168 43.5
Intentional self-harm 2 0.5 39 10.2 185 48.6 155 40.7
Assault 2 0.5 38 10.0 186 48.8 155 40.7
Place of injury 3 0.8 106 27.5 151 39.1 126 32.6
Activity at the time of injury 4 1.0 99 25.8 162 42.2 119 31.0
Specificity of codes available
Accidents 3 0.8 49 12.7 162 41.9 173 44.7
Intentional self-harm 2 0.5 42 11.0 156 40.8 182 47.6
Assault 2 0.5 37 9.7 164 42.8 180 47.0
Place of injury 4 1.0 130 33.9 141 36.7 109 28.4
Activity at the time of injury 5 1.3 117 30.6 153 40.1 107 28.0

rated 78 and 51%, respectively, by coders as factors that have a
high impact on external cause coding. These were followed by a
lack of a centralised form where all external cause information is
recorded and illegible medical record entries, with these being
rated as high impact by 47 and 34% of coders, respectively.
Similar patterns were found for place of injury occurrence and
activity at the time of the injury.
high impact) external cause coding quality could be improved
through various different measures. Over 85% of coders consid-
ered improving the quality of documentation in the emergency
department would have a high impact on the quality of coded
data, and 78% of coders considered the introduction of a struc-
tured form for external cause information would have a high
impact.

3.4. Satisfaction with external cause classification 4. Discussion

Table 4 shows coder satisfaction with the external cause
classification system. Coders were asked to rank their level of
satisfaction (highly satisfied, medium, low, and not at all satis-
fied) with the ICD-10-AM code blocks in terms of the number
and the specificity of the codes available. Coders reported lower
levels of satisfaction with the number and specificity of codes
for injury place and activity blocks compared to all other code
blocks. Between 26 and 34% of coders reported low satisfaction
with these code blocks. However, 28–32% coders reported high
levels of satisfaction with these specific code blocks.
3.5. Ways to improve external cause coding
Table 5 reports coders’ views of ways to improve external
cause coding. Coders were asked to what extent (no impact to
This study ascertained coders’ views of the reasons for a lack
specificity in external cause code usage to provide greater insight
into the findings from a previous study regarding external cause
code specificity (McKenzie et al., 2006). This research found
that a lack of external cause information in medical records
and poor clinical documentation regarding injury circumstances
were the major factors impacting on the specificity and resultant
quality of external cause code usage, particularly with regards
to place of occurrence and activity at the time of injury. Miss-
ing external cause information and missing documentation were
rated by 78 and 51% of respondents, respectively, as the factors
with the greatest impact for external cause coding.
The results of this study show that coders viewed dis-
charge summaries followed by doctors’ notes as the poorest
sources of information regarding external causes, place of injury

Table 5
Coders’ views of ways to improve external cause coding
Measure to improve coding quality No impact Low impact Medium impact High impact

n % n % n % n %

Improve quality of EDa documentation 1 0.3 11 2.9 42 11.2 321 85.6

Structured forms for external cause information 5 1.3 19 5.0 61 16.1 294 77.6
Standardise coding system in EDa 11 3.0 51 13.8 92 24.9 215 58.3
Improve ambulance documentation 10 2.7 49 13.0 113 30.0 205 54.4
Standardise coding system in ambulance coding 9 2.4 61 16.1 116 30.7 192 50.8
Research to investigate proposed changes 12 3.2 88 23.3 123 32.6 154 40.8
Increase number of specific ICD codes 17 4.5 84 22.0 133 34.9 147 38.6
More education in external cause coding 29 7.6 95 25.0 133 35.0 123 32.4
More time for external cause coding 40 10.8 131 35.2 123 33.1 78 21.0
a ED: emergency department.
718 K. McKenzie et al. / Accident Analysis and Prevention 40 (2008) 714–718
occurrence, and activity at the time of injury. The best source
of information regarding external causes, place, and activity
according to the coders was the ambulance report form with
over half of the coders stating that the ambulance report form
was a good source of information.
A lack of clinical documentation has been identified by previ-
ous researchers as a significant problem to obtaining high quality
coded data (Curtis et al., 2002; O’Malley et al., 2005). Curtis
et al. (2002) recommended that clinicians (medical, nursing,
and medical records staff) are made aware of the documentation
requirements for accurate coding, with various suggestions such
as redesigning discharge summaries, development of guidelines
for discharge summaries, and a pocket size guide outlining these
requirements being suggested.
The study also found that coders were considerably dissat-
isfied with the number and specificity of codes for place of
occurrence and activity when injured, with between 26 and 34%
of coders reporting low satisfaction with these code blocks. This
indicates that, beyond documentation issues, work is needed
to refine the external cause classification system and provide
more specific codes, particularly in regard to place of occur-
rence and activity. Developments and improvements to the
ICD-10-AM external cause classification are managed through
the National Centre for Classification in Health, through col-
laboration with the National Injury Surveillance Unit. Recent
developments and refinements to the external causes chapter
have been largely informed by the International Classification
of External Causes of Injuries (ICECI) (ICECI, 2004). The
ICECI uses a multi-axial system to capture detailed information
regarding various aspects of injury circumstances for multiple
uses.
A majority of coders suggested that two key approaches
to improve the quality of external cause coding would be to
improve clinical documentation in the emergency department
and to introduce a centralised structured form for external cause
information. Previous research has highlighted the need for more
detailed documentation in the emergency department (Schwartz
et al., 1995; Houry et al., 1999; Kaida et al., 2006) The World
Health Organization produced guidelines in 2001 for the collec-
tion of injury information using data collection forms (Holder
et al., 2001).
5. Conclusion
Clinical coders are a valuable source of information regarding
problems and solutions to the collection of high quality coded
data. This research is the first study to ascertain the professional
opinions of those individuals who are uniquely placed to provide
insight into the coding process, pitfalls, and potentials regarding
the use of administrative data for injury surveillance. High qual-
ity cause of injury data is a crucial element to building a strong
evidence base for injury prevention policy and practice.
Acknowledgements
We thank Associate Professor James Harrison, Geoffrey
Henley, and Professor Roderick McClure for their advice and
assistance in conducting this study. The study was funded by
the Australian Research Council.
References
Curtis, K., Bollard, L., Dickson, C., 2002. Coding errors and the trauma
patient—is nursing case management the solution? Aust. Health Rev. 25
(4), 73–80.
Holder, Y., Peden, M., Krug, E., Lund, J., Gururaj, G., Kobusingye, O. (Eds.),
2001. Injury Surveillence Guidelines. World Health Organisation, Geneva.
Houry, D., Feldhaus, K.M., Nyquist, S.R., Abbott, J., Pons, P.T., 1999. Emer-
gency department documentation in cases of intentional assault. Ann. Emerg.
Med. 34 (6), 715–719.
ICECI Coordination and Maintenance Group, 2004. International Classification
of External Causes of Injuries (ICECI) version 1.2. Consumer Safety Insti-
tute, Amsterdam and AIHW National Injury Surveillance Unit, Adelaide.
Kaida, A.K., Marko, J., Hagel, B., Lightfoot, P., Sevcik, W., Rowe, B.H., 2006.
Unspecified falls among youth: predictors of coding specificity in the emer-
gency department. Inj. Prev. 12 (5), 302–307.
Langley, J., Stephenson, S., Thorpe, C., Davie, G., 2006. Accuracy of injury
coding under ICD-9 for New Zealand public hospital discharges. Inj. Prev.
12 (1), 58–61.
Langlois, J.A., Buechner, J.S., O’Connor, E.A., Nacar, E.Q., Smith, G.S., 1995.
Improving the E coding of hospitalizations for injury: do hospital records
contain adequate documentation? Am. J. Public Health 85 (9), 1261–1265.
LeMier, M., Cummings, P., West, T.A., 2001. Accuracy of external cause of
injury codes reported in Washington State hospital discharge records. Inj.
Prev. 7, 334–338.
Lorence, D.P., Ibrahim, I.A., 2003. Disparity in coding concordance: do physi-
cians and coders agree? J. Health Care Finance 29 (4), 43–53.
MacIntyre, C.R., Ackland, M.J., Chandraraj, E.J., 1997. Accuracy of injury
coding in Victorian hospital morbidity data. Aust. N. Z. J. Public Health 27
(7), 779–783.
Mathers, C., Penn, R., 1999. Health System Costs of Injury, Poisoning and
Musculoskeletal Disorders in Australia 1993–1994. AIHW, Canberra.
McKenzie, K., Walker, S., 2002. National Coder Workforce Survey 2002.
National Centre for Classification in Health, Brisbane.
McKenzie, K., Harding, L., Walker, S., Harrison, J., Enraght-Moony, E., Waller,
G., 2006. The quality of cause-of-injury data: where hospital records fall
down. Aust. N. Z. J. Public Health 30 (6), 509–513.
National Centre for Classification in Health, 2004. The International Statisti-
cal Classification of Diseases and Related Health Problems, 10th Revision,
Australian Modification (ICD-10-AM), fourth ed. University of Sydney,
Sydney.
O’Malley, K.J., Cook, K.F., Price, M.D., Wildes, K.R., Hurdle, J.F., Ashton,
C.M., 2005. Measuring diagnoses: ICD code accuracy. Health Serv. Res. 40
(5p2), 1620–1639.
Schwartz, R.J., Nightingale, B.S., Boisoneau, D., Jacobs, L.M., 1995. Accuracy
of e-codes assigned to emergency department records. Acad. Emerg. Med.
2 (7), 615–620.
Walker, S., McEvoy, S., 2004. Injury classification systems. In: McClure, R.,
Stevenson, M., McEvoy, S. (Eds.), The Scientific Basis of Injury Prevention
Control. IP Communications, Melbourne, pp. 51–61.
World Health Organization (WHO), 1994. International Statistical Classification
of Diseases and Related Health Problems, 10th Revision (ICD-10), WHO,
Geneva.