Professional practice and innovation: level of agreement between coding sources of percentage total body surface area burnt (%TBSA).
Abstract: The percentage of total body surface area burnt (%TBSA) is a critical measure of burn injury severity and a key predictor of burn injury outcome. This study evaluated the level of agreement between four sources of %TBSA using 120 cases identified through the Victorian State Trauma Registry. Expert clinician, ICD-10-AM, Abbreviated Injury Scale, and burns registry coding were compared using measures of agreement. There was near-perfect agreement (weighted Kappa statistic 0.81-1) between all sources of data, suggesting that ICD-10-AM is a valid source of %TBSA and use of ICD-10-AM codes could reduce the resource used by trauma and burns registries capturing this information.

Keywords (MeSH): Burns; Trauma Severity Indices; Registries; Information Storage and Retrieval; ICD-10
Article Type: Report
Subject: Burn care units (Information management)
Registries (in medicine) (Management)
Authors: Watterson, Dina
Cleland, Heather
Picton, Natalie
Simpson, Pam M.
Gabbe, Belinda J.
Pub Date: 02/01/2011
Publication: Name: Health Information Management Journal Publisher: Health Information Management Association of Australia Ltd. Audience: Academic Format: Magazine/Journal Subject: Health Copyright: COPYRIGHT 2011 Health Information Management Association of Australia Ltd. ISSN: 1833-3583
Issue: Date: Feb, 2011 Source Volume: 40 Source Issue: 1
Topic: Event Code: 200 Management dynamics; 260 General services Computer Subject: Company business management; Company systems management
Geographic: Geographic Scope: Australia Geographic Code: 8AUST Australia
Accession Number: 252944963
Full Text: Introduction

The percentage of total body surface area burnt (%TBSA) is a critical measure of burn injury severity, has been identified as a key predictor of burn injury outcome (Germann et al. 1997; Mahar et al. 2008), and is an important variable for appropriate remuneration of clinical service provision (Turner et al. 1996; Wallis et al. 2009). The %TBSA has also been identified as one of the most important burn injury measures for burn care research, quality assurance and benchmarking activities (Perry et al. 1996; Wachtel et al. 2000; Howard et al. 2007).

Following burn injury, the %TBSA can be estimated or assessed at multiple time points post-injury and by a number of clinicians. Difficulties can arise where multiple %TBSA estimates are recorded, particularly with respect to identifying the definitive or most accurate %TBSA assessment for coding purposes. Variable coding of %TBSA can have implications for the validity and reliability of burn care research and quality activities. The potential variation in assessment, documentation and resultant coding of %TBSA means caution is required when interpreting research findings and evaluating trends in burn injury (Perry et al. 1996; Wachtel et al. 2000).

The patient medical record is the primary source of %TBSA information and is used to enable routine coding and classification for hospital and health system administrative processes and capture of information for clinical databases such as burns and trauma registries. The International Statistical Classification of Disease and Health Related Health Problems, Tenth Revision Australian Modification (ICD-10-AM) and the Abbreviated Injury Scale (AIS) are standardised classifications for coding burns injury (Association for the Advancement of Automotive Medicine 2008; Commonwealth of Australia 2008). The ICD-10-AM is routinely used for coding and classifying burn injury for administrative data collection purposes and/or funding within Australian and New Zealand hospitals. The Abbreviated Injury Scale (AIS) is used by trauma registries worldwide to code injuries sustained and calculate the Injury Severity Score (ISS) for trauma patients, including burn patients. These classification systems categorise the %TBSA, while burns registries routinely record a %TBSA without categorisation. Each method relies on medical record documentation to enable retrospective coding. Given the multiple assessments of %TBSA recorded during a burns patient admission, there is the potential for disagreement in burn size coding across the datasets, but agreement between the methods has not been previously investigated.

The aim of this project was to determine the level of agreement between different sources of %TBSA, and to establish whether %TBSA is coded consistently from the medical record irrespective of the source of coded information. Improved understanding of the relationship between sources of %TBSA coding is needed to inform the development of data collection methodologies for burns, and to provide context for the evaluation and interpretation of benchmarking and research findings.

Method

Setting

The State of Victoria, Australia, has one designated adult burn care facility, the Victorian Adult Burns Service (VABS) located at The Alfred Hospital, Melbourne. All patients admitted to the VABS are recorded on the VABS registry, which is developed and managed within the VABS service and used in 'real time' for burn unit management and weekly case audits. Junior medical staff are responsible for registry data collection, including recording the %TBSA estimate. The Victorian State Trauma Registry (VSTR) is a population-based registry which captures data about all major trauma patients in Victoria, including those admitted to the VABS (Cameron et al. 2005). The VSTR captures %TBSA using the ICD10-AM and AIS diagnosis codes. The ICD-10-AM codes are collected at discharge for burn patients through medical record review by the hospital's clinical coders. The AIS codes are assigned by VSTR data collectors through medical history review post-discharge.

Procedures

Ethics approval was obtained from The Alfred Human Research Ethics Committee. Burns cases captured by the VSTR with a date of injury from 1 January 2007 to 31 December 2008 were extracted for inclusion. The ICD10-AM and AIS data related to %TBSA, and demographic data for adult burn cases identified from the VSTR and definitively managed at the VABS were retrieved for analysis. The %TBSA recorded by the VABS registry was merged with the VSTR data, using patient and admission identifiers to link. In addition, the individual patient medical records were retrieved for expert burn clinician review and separate coding of the definitive %TBSA. This information was then entered into the combined VSTR and VABS dataset for analysis. The expert burn clinician was blinded to the VSTR and VABS data.

To enable comparison of the continuous coding of the VABS and expert burn clinician %TBSA with the categorical AIS and ICD-10-AM coding, the VABS and expert burn clinician %TBSA was categorised into the corresponding six AIS %TBSA categories and the ten ICD 10-AM %TBSA categories.

Scatter plots and Bland-Altman plots were generated to visually represent the level of agreement between the expert burn clinician and VABS, as the %TBSA was recorded on the full scale from 0 to 100%. Bland-Altman plots display the mean difference between two variables with the solid line representing the overall mean difference and the dashed lines representing the limits of agreement (Bland and Altman 1986).

A Wilcoxon signed rank test, a non parametric test equivalent to a paired T-Test, was used to test the hypothesis that there were differences in agreement between the expert burn clinician coding of %TBSA and the VABS coding due to the skewed distribution of the data. Mean differences, and 95% confidence intervals (CI) of the difference, as well as limits of difference were calculated.

Agreement greater than that expected by chance between the categorised classifications of the %TBSA was examined using the weighted kappa statistic ([K.sub.w]). The weighted kappa is an extension of a simple kappa for nominal data where less weight is assigned to large differences between ratings than to small differences (Sim & Wright 2005). The weighted Kappa statistics were interpreted using established conventions for Kappa where; [less than or equal to] 0 = poor, 0.01-0.20=slight, 0.21-0.40=fair, 0.41-0.60=moderate, 0.61-0.80=substantial and 0.81-1 = almost perfect (Landis & Koch 1977). Ninety-five percent confidence intervals (95% CI) were calculated for [K.sub.w] using the Bootstrap method. All analyses were performed using Stata 11.0 (StataCorp 2009).

Results

One hundred and twenty cases were included. Table 1 outlines the cases available for each coding procedure. There was an AIS %TBSA code for all cases. The ICD10-AM, VABS and expert burn clinician %TBSA data was unable to be coded for one case where the AIS code was <10% TBSA. There were two cases where the VABS and expert burn clinician identified an inhalation injury only, while AIS and ICD-10-AM codes of <10% TBSA were reported. There were two cases where the expert burn clinician was unable to code the %TBSA; however codes from all other sources were available. There were three cases were VABS did not have a %TBSA coded when the other coding sources did. The medical histories were not able to be retrieved for seven cases for expert burn clinician coding.

Figure 1 shows the scatterplot of the expert burn clinician coding of %TBSA against the VABS coding of %TBSA. Points on the line represent perfect agreement between VABS and the expert burn clinician coding. The mean (95% CI) difference between the %TBSA coding of the expert burn clinician and VABS database was -0.05 (-0.87 to 0.76). There was no significant difference (p=0.81) between the coding procedures, and this is confirmed in the Bland-Altman plot (Figure 2) as there is an even spread of points around the mean difference line with just a few cases near the limits of agreement (-8.453 to 8.349).

Agreement between the ICD-10-AM and AIS, AIS and expert burn clinician and ICD-10-AM and expert burn clinician was almost perfect based on the established conventions for the kappa statistic (Landis & Koch 1977), and substantial for the agreement between the VABS and AIS, and VABS and ICD-10-AM (Table 2).

[FIGURE 1 OMITTED]

While not significant, the lowest levels of agreement for comparisons with the AIS were in the 10-19 %TBSA category, with the burn generally considered more extensive using the AIS code. For the ICD-10-AM and VABS coding, the agreement was lowest in the 40-49% category (28%). For the ICD-10-AM and expert burn clinician comparisons, agreement was lowest in the <10% category (65%). These results were not significant.

Discussion

Percentage total body surface area burnt is routinely collected by trauma and burn registries worldwide and used widely in research, quality and benchmarking activities. The collection of the %TBSA remains a time consuming and complicated process for registry data collectors in terms of the multiple sources of data and time for data retrieval (Howard et al. 2007). This study is believed to be the first to explore the agreement between multiple sources of %TBSA. The results suggest coding procedures for hospital reporting, the local burn unit registry and the separate population-based trauma registry have substantial or near perfect agreement.

A primary function of many registries is to stimulate the use of the registry data for research. The importance of collecting data systematically to ensure data quality and reduce data collector burden is considered fundamental to the ongoing sustainability of registries (Australian Commission on Safety and Quality in Health Care 2008; McNeil et al. 2010). The retrieval of %TBSA creates additional data collection burden, given it is already routinely collected by hospital clinical coders for administrative requirements. Results of this study suggest the administrative data could be a valid alternative for sourcing the %TBSA by trauma and burn registries.

[FIGURE 2 OMITTED]

While both the ICD-10-AM and the AIS were found to agree with the other coding sources, the ICD-10-AM has greater potential to provide more specific %TBSA data for registries and clinical research. The AIS is less specific than the ICD-10-AM, with six %TBSA categories and provides no additional information, limiting its usefulness. The ICD-10-AM has 10 %TBSA categories providing more detailed data. Unless more specific burn size data is required, the use of the ICD-10-AM appears to be a valid source for %TBSA data. One limitation to using the ICD-10-AM coding for %TBSA data is the time of data collection. Clinical coding of ICD-10-AM data is routinely completed following discharge. For registries that require %TBSA data for regular caseload management or clinical audits, relying on ICD-10-AM data would be impractical.

The literature available on %TBSA coding has highlighted the importance of good documentation of burn variables, including %TBSA for accurate coding. Tuner et al. highlighted the challenge with identifying the most accurate or definitive %TBSA for coding purposes and the impact this had on assignment of ICD-10-AM and Diagnostic Related Group codes related to reimbursement processes for hospital payment systems. They introduced a system of revising %TBSA diagrams to correctly depict the %TBSA and therefore obtaining optimal and accurate financial reimbursement for burn service (Turner et al. 1996).

Similarly, Wallis et al. (2009) created a document to improve the coding processes in their burns unit; the improved documentation ensured a high quality of coding which was considered to have a possible direct impact on the financial resources accrued for the burn care. At the time of data collection for this project, the VABS did not have a systematic data collection form to accurately collect burn variables used for coding. Following the identification of the challenge in retrieving the %TBSA from medical history documentation for registry coding a burn assessment form has been proposed for the VABS with plans to implement this form in the near future. This will improve retrieval of burn data for clinical coders including the %TBSA.

There are some limitations to this study. The study hospital is casemix funded and relies heavily on accurate coding for financial remuneration. There are processes and work practices in place for clinical coder and clinician consensus on %TBSA and other burn variables for ICD-10-AM codes primarily used in hospital financial reporting. This may limit the ability for generalisation of results beyond the health care system of the study hospital if other systems do not have the same incentive for accurate clinical coding. In Australia all states used ICD-10-AM coding; however, other countries use the ICD-9 version, which is structurally different from ICD-10-AM. This also creates difficulties in generalising results. The final limitation is that the expert burn clinician who completed the retrospective data audit was a senior staff member of VABS. While at least a year had passed since the patients' admissions, the possibility that the clinician would recall the individual cases could not be excluded. However, given that there are more than 250 admissions per year to the VABS, the potential for recall to influence the results is likely to be small. Nevertheless, this may have reduced the possibility of poor agreement between coding sources.

Conclusion

Results of this study suggest substantial or near perfect agreement between coding sources of %TBSA. If the specificity of the %TBSA codes from the ICD-10-AM is considered sufficient for registry and clinical research purposes, further coding by burns and trauma registry staff may be unnecessary. This would decrease data collector burden and increase registry data quality.

References

Association for the Advancement of Automotive Medicine (AAAM) (2008). Abbreviated Injury Scale 2005 Update 2008. Barrington, IL USA, Association for the Advancement of Automotive Medicine.

Commonwealth of Australia (2008). The International Statistical Classification of Disease and Related Health Problems (ICD-10-AM). Tenth Revision. Sydney, National Centre for Classification in Health.

Bland, J. and Altman, D. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 8: 307-310.

Cameron, P., Gabbe, B., McNeil, J., Finch, C., Smith, K., Cooper, J., Judson, R. and Kossman, T. (2005). The Trauma Registry as a statewide quality improvement tool. Journal of Trauma, Injury, Infection & Critical Care 59(6):1469-79.

Australian Commission on Safety and Quality in Health Care (ACSQHC) (2008). Operating principles and technical standards for Australian clinical quality registries. Available at: http://www. safetyandquality.gov.au/internet/safety/publishing.nsf/Content/ PriorityProgram-08_CQRegistries (accessed 20 June 2010).

Germann, G., Barthold, U., Lefering, R., Raff, T. and Hartmann, B. (1997). The impact of risk factors and pre-existing conditions on the mortality of burn patients and the precision of predictive admission-scoring systems. Burns 23(3):195-203.

Howard, P. A., Jeng, J. C. and Miller, S. F. (2007). Is the glass really half empty? A closer look at the TBSA data in the National Burn Repository. Journal of Burn Care & Research 28(4):542-543. Landis, J. and Koch, G. (1977). The measurement of observer agreement for categorical data. Biometrics 33: 159-174.

Mahar, P., Wasiak, J., Bailey, M. and Cleland, H. (2008). Clinical factors affecting mortality in elderly burn patients admitted to a burns service. Burns 34: 629-636.

McNeil, J., Evans, S., Johnson, N. and Cameron, P. (2010). Clinical-quality registries: their role in quality improvement. Medical Journal of Australia 192(5):244-245.

Perry, R. J., Moore, C. A., Morgan, B. D. G. and Plummer, D. L. (1996). Determining the approximate area of a burn: an inconsistency investigated and re-evaluated. British Medical Journal 312(7042):1338.

Sim, J. and Wright, C. C. (2005). The Kappa statistic in reliability studies: use, interpretation, and sample size requirements. Physical Therapy 85(3): 257-268.

StataCorp (2009). Stata 11.0. Texas, College Station.

Turner, D. G., Berger, N., Weiland, A. P. and Jordan, M. H. (1996). The revised burn diagram and its effect on diagnosis-related group coding. Journal of Burn Care & Rehabilitation 17(2):169-174.

Wachtel, T. L., Berry, C. C., Wachtel, E. E. and Frank, H. A. (2000). The inter-rater reliability of estimating the size of burns from various burn area chart drawings. Burns 26: 156-170.

Wallis, K., Malic, C., Littlewood, S., Judkins, K. and Phipps, A. (2009). Surviving 'Payment by Results: a simple method of improving clinical coding in burn specialised services in the United Kingdom. Burns 35: 232-236.

Corresponding author:

Dina Watterson BOT(Hons), MPH

Research Fellow

Department of Epidemiology and Preventive Medicine

Monash University

The Alfred Centre

99 Commercial Road

Melbourne VIC 3004

AUSTRALIA

Tel: +61 408583947

email: Dina.Watterson@monash.edu

Heather Cleland MBBS, FRACS

Director

Victorian Adult Burns Service

Alfred Hospital

Department of Surgery

Central & Eastern Clinical School

Monash University

99 Commercial Road

Melbourne VIC 3004

AUSTRALIA

Natalie Picton BHS(Nurs)

Project Co-Ordinator

Department of Epidemiology and Preventive Medicine

Monash University

99 Commercial Road

Melbourne VIC 3004

AUSTRALIA

Pam M Simpson BSc(Hons)

Biostatistician

School of Public Health and Preventive Medicine

Monash University

99 Commercial Road

Melbourne VIC 3004

AUSTRALIA

Belinda J Gabbe BPhysio(Hons), GradDipBiostat, MAppSc, PhD

Senior Research Fellow

Department of Epidemiology and Preventive Medicine

Monash University

99 Commercial Road

Melbourne VIC 3004

AUSTRALIA
Table 1: Number of cases by coding procedure

CODING PROCEDURE        NO. OF CASES

AIS                         120
ICD-10-AM-AM                119
VABS                        114
Expert burn clinician       108

Table 2: Agreement between coding procedures for %TBSA categories

                                             WEIGHTED
CODING                         % AGREEMENT    KAPPA       95% CI

ICD vs AIS                        95.8         0.88     (0.78-0.91)
AIS vs VABS                       92.0         0.77     (0.69-0.84)
ICD vs VABS                       93.9         0.78     (0.70-0.83)
AIS vs Expert burn clinician      94.1         0.83     (0.77-0.87)
ICD vs Expert burn clinician      95.2         0.83     (0.74-0.90)
Gale Copyright: Copyright 2011 Gale, Cengage Learning. All rights reserved.