Objective: To evaluate the accuracy of World Health Organization (WHO) method of interpreting chest radiographs on identifying young children with bacterial pneumonia, and to compare its accuracy with other method.
Methods: Chest radiographs from children aged under 5 years old hospitalized for pneumonia, with microbiological evidence of bacterial or viral infection, were evaluated by 3 observers blinded to other data (pediatrician [P], pulmonologist [N], radiologist [R]) according to WHO and Khamapirad scores. A WHO score=1 and a Khamapirad score >2 were selected as the thresholds suggesting bacterial pneumonia. The relationship between radiographic scores and etiology was evaluated using chi square. Sensitivity (Se), specificity (Sp), positive (PPV) and negative (NPV) predictive values for predicting bacterial pneumonia were calculated. Interobserver agreement (P vs. N vs. R) was calculated (kappa).
Results: 108 chest radiographs were evaluated (87 viral, 21 bacterial). WHO score= 1 was associated with bacterial pneumonia (p <0.001; OR= 6.4; CI 95%= 1.629.7), achieving a Se= 85%, Sp= 51%, PPV= 30%, NPV= 93%. Khamapirad score >2, evaluated by P, was also associated with bacterial pneumonia (p <0.0008; OR= 6.31; CI 95%= 1.8-24.4), achieving a Se= 80%, Sp= 59%, PPV= 32% NPV= 92%. Interobserver agreement was slightly better using WHO score (P vs. N= 0.82, P vs. R= 0.69, N vs. R= 0.85) than Khamapirad score (P vs. N= 0.48, P vs. R= 0.69, N vs. R= 0.82).
Conclusions: Both methods showed similar accuracy in order to identify bacterial pneumonia. WHO score is simpler than Khamapirad score and showed a better interobserver agreement.