Intra and interobserver reliability of the interpretation of high-resolution computed tomography on the lungs of premature infants

• Autor(a) principal: Boëchat,Márcia Cristina Bastos
• Data de Publicação: 2010
• Outros Autores: Mello,Rosane Reis de, Dutra,Maria Virgínia Peixoto, Silva,Kátia Silveira da, Daltro,Pedro, Marchiori,Edson
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: São Paulo medical journal (Online)
• Texto Completo: http://old.scielo.br/scielo.php?script=sci_arttext&pid=S1516-31802010000300005
• topic: Tomography, X-ray computed; Infant, premature; Lung; Bronchopulmonary dysplasia; Reproducibility of results; Infant

CONTEXT AND OBJECTIVE: High-resolution computed tomography (HRCT) of the lungs is more sensitive than radiographs for evaluating pulmonary disease, but little has been described about HRCT interpretation during the neonatal period or shortly afterwards. The aim here was to evaluate the reliability of the interpretation of HRCT among very low birth weight premature infants (VLBWPI; < 1500 g). DESIGN AND SETTING: Cross-sectional study on intra and interobserver reliability of HRCT on VLBWPI. METHODS: 86 VLBWPI underwent HRCT. Two pediatric radiologists analyzed the HRCT images. The reliability was measured by the proportion of agreement, kappa coefficient (KC) and positive and negative agreement indices. RESULTS: For radiologist A, the intraobserver reliability KC was 0.79 (confidence interval, CI: 0.54-1.00) for normal/abnormal examinations; for each abnormality on CT, KC ranged from 0.05 to 1.00. For radiologist B, the intraobserver reliability KC was 0.79 (CI: 0.54-1.00) for normal/abnormal examinations; for each abnormality on CT, KC ranged from 0.37 to 0.83. The interobserver agreement was 88% for normal/abnormal examinations and KC was 0.71 (CI: 0.5- 0.93); for most abnormal findings, KC ranged from 0.51-0.67. CONCLUSION: For normal/abnormal examinations, the intra and interobserver agreements were substantial. For most of the imaging findings, the intraobserver agreement ranged from moderate to substantial. Our data demonstrate that in clinical practice, there is no reason for more than one tomographic image evaluator, provided that this person is well trained in VLBWPI HRCT interpretation. Analysis by different observers should be reserved for research and for difficult cases in clinical contexts.