Anàlisi metodològic dels estudis de concordança interobservador i reproductibilitat intraobservador en la caracterització de fractures de l’extrem proximal de l’húmer

Abstract

The coefficient kappa, used to value the inter and intraobserver reliability of a classification, presents a bias of prevalence for which when a category of the classification is a lot or a lot little prevailing the value tends to zero independently of that the observers have an overhead agreement. This fact, has posited for some authors like the cause of the low results of agreement obtained in the Neer classification of proximal humerus fractures. Hypothesis of work: The bias of prevalence of the coefficient Kappa of Cohen, influences in the low inter- and intraobserver agreement observed in the Neer classification of proximal humerus fractures. Secondary aims: Analyze the influence of the CT-scan sequence and of the observer’s experience in the interobserver agreement and reproducibility in the characterization and classification of proximal humerus fractures. Material and Method: We selected a sample of 153 consecutive proximal humerus fractures in conservative treatment with simple radiology and CT scan. Six observers with different degrees of experience valued 19 radiological characteristics with simple radiology and axial sequence of the CT scan and with simple radiology and complete sequence of the CT scan and applied the Neer classification simplified in 4 categories. We value the agreement differences between the different radiology studies and as the degrees of experience. A sample of consecutive cases allow us to value the Kappa coefficient with a no swung distribution of the four categories of the Neer classification. We selected a new sample of 92 fractures of the proximal humerus with a balanced distribution of the 4 fractures types of the Neer classification. The differences in reliability of the Neer classification between the two samples are evaluated. Results: In the consecutive sample, only in the characteristic lateromedial cephalodiaphyseal displacement observe significant differences in the interobserver agreement between the axial CT scan and the complete CT scan (Kvalue axial=0,53; Kvalue complete=0,35; p=0,001) with better agreement in axial. In the rest of variables do not observe significant differences. The Kappa value in the Neer classification was 0,43 in the two studies. The intraobserver reproducibility improved in the study with complete CT-scan with significant differences in seven of the variables studied. No differences were observed in the Neer classification. (K value axial=0,74; K value complete= 0,72; p=0,74) As the degree of experience, the most expert observers, in the study with axial CT-scan obtained significantly better agreement in variables of cephalodiaphyseal relationship and of greater tuberosity displacement. In the Neer classification the differences were significant. (Kvalue expert=0,52; Kvalue no expert= 0,36; p= 0,04). In the study with the complete CT scan sequences the differences would be lesser. In the Neer classification did not observe significant differences in complete CTscan. (Kvalue expert=0,43; Kvalue no expert 0,38; p= 0,534) Comparing the interobserver agreement in the Neer classification between the two samples we do not observe significant differences. (K value consecutive sample= 0,43; Kvalue selected sample= 0,45; p= 0,68) Neither observe differences in the intraobserver reproducibility. (Kvalue consecutive sample =0,72; Kvalue selected sample=0,74; p=0,14) Conclusions: The bias of prevalence of the Kappa coefficient does not contribute significantly in the low results observed previously in the inter- and intraobserver reliability of the Neer classification. The use of the complete CT-scan sequence does not help to improve the degree of agreement between observers when value radiological characteristics neither in applying the Neer classification, but it improves the reproducibility. The degree of experience increases the agreement between observers in valuing some characteristics, mainly in the study with simple radiology and axial CT scan. [-]

Date of Award	17 Oct 2014
Original language	English
Awarding Institution	Universitat Autònoma de Barcelona (UAB)
Supervisor	Enric Caceres Palou (Director) & Carlos Torrens Canovas (Director)