Agreement of burn size estimations: result of five data sets

Objectives The determination of the burned TBSA is done by different methods. Regardless the used method there are 2 key factors for quality: (1) the average error of estimation of real wound sizes, as well as (2) the agreement of estimation of different raters (Inter-Rater-Reliability, IRR). Especially for comparisons and for aggregations of size assessments the IRR has an important role. The IRR for five datasets will be investigated with regards to method, experience of raters and education of raters. As measure of the IRR the Intra-Class-Correlation will be used. Methods For assessment of the TBSA-B mannequins with artificial wounds have been prepared. By use of 2D and 3D Scans the TBSA of the mannequin and the TBSA-B of each wound was measured as a gold standard. In one data set the TBSA-B was marked on photos with highlights borders only. The IRR of the individual TBSA-B estimations were determined by Intra-Class-Correlation, variant ICC(2,1) [1] . In contrast to the classic Pearson-Correlation the ICC(2,1) has the advantage, to handle offset errors as well as multiplicative errors. An ICC of 1,0 can only be reached by exact agreement, an over-estimation of 100% would mean an ICC of 0,49. Results In the years from 2010 to 2016 multiple data-sets of TBSA-B assessments investigations have been created within the scope of scientific conferences (German Speaking Burns Meeting, European Burns Association) as well as in a separate study with paramedics. The raters have been asked to use different Methods (Hand-Rule, Rule-of-Nines, Lund-Browder Chart, 2D/3D computer aided methods). In relation to the gold-standard the resulting average ICC(2,1) ranges from 0,25 to 0.97. Discussion / Conclusion Even unexperienced raters (e.g. paramedics, SAN 2015) can reach good agreement with the gold standard. By high qualified burn experts, the usage of computer-aided methods showed a significant improvement of the Inter-Rater-Reliability.