Validation of the "HAMP" mapping algorithm: a tool for long-term trauma research studies in the conversion of AIS 2005 to AIS 98.

BACKGROUND: There are significant changes in the abbreviated injury scale (AIS) 2005 system, which make it impractical to compare patients coded in AIS version 98 with patients coded in AIS version 2005. METHODS: Harborview Medical Center created a computer algorithm "Harborview AIS Mapping Program (HAMP)" to automatically convert AIS 2005 to AIS 98 injury codes. The mapping was validated using 6 months of double-coded patient injury records from a Level I Trauma Center. HAMP was used to determine how closely individual AIS and injury severity scores (ISS) were converted from AIS 2005 to AIS 98 versions. The kappa statistic was used to measure the agreement between manually determined codes and HAMP-derived codes. RESULTS: Seven hundred forty-nine patient records were used for validation. For the conversion of AIS codes, the measure of agreement between HAMP and manually determined codes was [kappa] = 0.84 (95% confidence interval, 0.82-0.86). The algorithm errors were smaller in magnitude than the manually determined coding errors. For the conversion of ISS, the agreement between HAMP versus manually determined ISS was [kappa] = 0.81 (95% confidence interval, 0.78-0.84). CONCLUSION: The HAMP algorithm successfully converted injuries coded in AIS 2005 to AIS 98. This algorithm will be useful when comparing trauma patient clinical data across populations coded in different versions, especially for longitudinal studies.

The impact of the AIS 2005 revision on injury severity scores and clinical outcome measures.

BACKGROUND: The abbreviated injury scale (AIS) was updated in 2005 from the AIS 1998 version. The purpose of this study is to describe the effects of this change on injury severity scoring and outcome measures. MATERIALS AND METHODS: Analyses were performed on all trauma patients consecutively admitted over a 6-month period at two geographically separate Level I trauma centers. Injuries were manually double-coded according to the AIS 05 and the AIS 98. Changes in AIS, ISS, and new ISS (NISS) were analysed using paired t-tests. Apparent differences in outcome by ISS strata (<16, 16-24, >24) were compared for AIS 05 versus AIS 98 using the Wald-type statistic. Lastly, the percent of patients with a change in ISS strata are reported. RESULTS: There were 2250 patients included in the study. Nearly half (46.4%) of AIS codes changed, resulting in a different AIS score for 18.9% of all codes. The mean ISS was significantly lower using the AIS 05 (11.7) versus the AIS 98 (13.3, p<0.001). Similarly, the mean NISS was significantly lower (16.3 versus 18.7, p<0.001). In the ISS strata 16-24 an apparent increase in mortality, length of stay, and percent of patients not discharged home was observed for the AIS 05 versus AIS 98. Changes in outcome measures for this stratum were as follows (AIS 98 versus AIS 05): mortality, 4.3% versus 7.7% (p=0.002); hospital length of stay, 5.2 days versus 7.3 days (p<0.001); percent of patients not discharged home, 39.2% versus 49.3% (p<0.001). Finally, there was a 20.5% reduction in patients with an ISS>or=16 and a 26.2% reduction in patients with an ISS>or=25 using the AIS 05. CONCLUSIONS: The AIS revision had a significant impact on overall injury severity measures, clinical outcome measures, and percent of patients in each ISS strata. Therefore, the AIS revision affects the ability to directly compare data generated using AIS 05 and AIS 98 which has implications in trauma research, reimbursement and ACS accreditation.

Effects of a nonsurgical hospitalist service on trauma patient outcomes.

BACKGROUND: The American College of Surgeons criteria for Level I trauma centers calls for >90% of trauma patients to be admitted directly by a trauma surgeon or surgical subspecialist; however, the efficiency of the trauma system may be increased if patients presenting with comorbid conditions and minor injuries are treated by a hospitalist team (nonsurgical Trauma MEDical [TMED] service). We hypothesized outcomes would be equivalent for patients treated under TMED versus a surgical service. METHODS: This retrospective review compared mortality, hospital length of stay (LOS), Emergency Department (ED) LOS, placement to rehabilitation facilities, and complication rates for patients who could have been treated by TMED as identified by an algorithm. The study population for 2003 (pre-TMED) was compared with the study population for 2006 (post-TMED). Univariate analyses and multivariate logistic and linear regression were used to identify outcomes that were different for patients treated in 2003 versus 2006. Sensitivity, specificity, and percent kappa agreement were calculated for patients who were treated by the TMED team in 2006 versus patients in 2006 who were identified using the algorithm. RESULTS: The algorithm had reasonable sensitivity (78%) and specificity (90%); the kappa agreement was excellent (0.88). No differences were found in mortality (P = .31), rate of complications (P = .08), ED LOS (P = .77), or placement to rehabilitation facilities (P = .29) for patients identified in 2003 versus 2006. Hospital LOS was increased in 2006 (3.7 vs 4.1 days; P = .02). CONCLUSION: These data support admission of trauma patients with nonsevere, single-system injuries to a nonsurgical hospitalist service. We hypothesize that overall system efficiency may be improved by applying this alternative model in other trauma centers.