Interobserver agreement in the clinical assessment of children with blunt abdominal trauma
Kenneth Yen,Nathan Kuppermann,Kathleen Lillis,David Monroe,Dominic Borgialli,Benjamin T Kerrey,Peter E Sokolove,Angela M Ellison,Lawrence J Cook,James F Holmes,Intra-abdominal Injury Study Group for the Pediatric Emergency Care Applied Research Network (PECARN),M Tunik,L Lee,P Mahajan,M Kwok,F Nadel,S Atabaki,B Kerrey,J Kooistra,D Monroe,D Borgialli,S Blumberg,K Yen,B Bonsu,N Kuppermann,J Holmes,J Menaker,A Rodgers,M Garcia,K Adelgais,K Quayle,K Lillis,E Alpern,J Callahan,J Chamberlain,P Dayan,J M Dean,M Gerardi,M Gorelick,J Hoyle,E Jacobs,D Jaffe,R Lichenstein,R Maio,R Ruddy,R Stanley,A Walker,D Kavanaugh,H Park,R Holubkov,S Knight,A Donaldson,S Zuspan,M Miskin,J Wade,A Jones,M Fjelstad,T Singh,A Drongowski,L Fukushima,E Kim,G O'Gara,H Rincon,K Brown,L Cimpello,G Foltin,F Moler,S Teach,A Cooper,N C Mann,K Shaw,P Ehrlich,R Enriquez,B Millar,M Shults
DOI: https://doi.org/10.1111/acem.12132
Abstract:Objectives: The objective was to determine the interobserver agreement of historical and physical examination findings assessed during the emergency department (ED) evaluation of children with blunt abdominal trauma. Methods: This was a planned substudy of a multicenter, prospective cohort study of children younger than 18 years of age evaluated for blunt abdominal trauma. Patients were excluded if injury occurred more than 24 hours prior to evaluation or if computed tomography (CT) imaging was obtained at another hospital prior to transfer to a study site. Two clinicians independently recorded their clinical assessments of a convenience sample of patients onto data collection forms within 60 minutes of each other and prior to CT imaging (if obtained) or knowledge of laboratory results. The authors categorized variables as either subjective symptoms (i.e., patient history) or objective findings (i.e., physical examination). For each variable recorded by the two observers, the agreement beyond that expected by chance was estimated, using the kappa (κ) statistic for categorical variables and weighted κ for ordinal variables. Variables with 95% lower confidence limits (LCLs) κ ≥ 0.4 (moderate agreement or better) were considered to have acceptable agreement. Results: A total of 632 pairs of physician observations were obtained on 23 candidate variables. Acceptable agreement was achieved in 16 (70%) of the 23 variables tested. For six subjective symptoms, κ ranged from 0.48 (complaint of shortness of breath) to 0.90 (mechanism of injury), and only the complaint of shortness of breath had a 95% LCL κ < 0.4. For the 17 objective findings, κ ranged from -0.01 (pelvis instability) to 0.82 (seat belt sign present). The 95% LCL for κ was <0.4 for flank tenderness, abnormal chest auscultation, suspicion of alcohol or drug intoxication, pelvis instability, absence of bowel sounds, and peritoneal irritation. Conclusions: Observers can achieve at least acceptable agreement on the majority of historical and physical examination variables in children with blunt abdominal trauma evaluated in the ED. Those variables are candidates for consideration for development of a clinical prediction rule for intra-abdominal injury in children with blunt trauma.