RESUMO
Objective. To assess the accuracy and reliability of yale observation scale (YOS) predicting bacteremia. Methods. 219 consecutive febrile inpatients aged 3-36 months were the subjects. Before giving antipyretics, rectal temperature was recorded. YOS scores were assessed by 2 independent blinded residents. History, clinical examination and investigations followed. Blood cultures were taken in all children before antibiotics. Point estimates and 95% confidence intervals were calculated for sensitivity, specificity, positive & negative predictive values and likelihood ratios for use of YOS as a diagnostic test in prediction of bacteremia. The best cut off value for a positive YOS test was established by calculating these statistical values separately for a cut off YOS score of 8, 10 and 12 and plotting ROC curve. Reliability of YOS was assessed by the inter-observer agreement through kappa statistics. Results. Study population (n=219) had 59.36% males and a mean age of 15.24 months. 28.16% subjects had bacteremia. Mean YOS scores were significantly higher in bacteremic children (14.9 vs 8.78 in non-bacteremic, p=0.00001) Sensitivity, specificity, PPV, NPV, LR+ and LR- of YOS score >10 to predict bacteremia were 87.93%, 83.78%, 68.00%, 94.66%, 5.42 and 0.14 respectively. Those of YOS score >8 were 96.55%, 65.54%, 52.34%, 97.98%, 2.80 and 0.05 respectively and of a YOS score >12 were 48.28%, 91.22%, 68.29%, 81.82%, 5.5 and 0.5 respectively. ROC curve showed YOS score >10 to be the best cut off for prediction of bacteremia. Area under ROC curve was 0.9001. The chance corrected inter-observer agreement (kappa) was 0.7919. Conclusion. YOS is a simple, easy to administer, cost-effective and useful test to predict bacteremia in a febrile child aged 3-36 months due to its high sensitivity and reproducibility.