Performance of a Novel Molecular Method in the Diagnosis of Late-Onset Sepsis in Very Low Birth Weight Infants

OBJECTIVE: To compare the use of a generic molecular assay to 'standard' investigations used to assist the diagnosis of late onset bacterial sepsis in very low birth weight infants (VLBW, <1500g).

METHODS: VLBW infants, greater than 48 hours of age, who were clinically suspected to have sepsis were investigated using standard tests (full blood count, C-reactive protein (at presentation) and blood culture), in addition, blood was taken for a universal molecular assay (16S rRNA reverse transcriptase PCR) for comparison. Clinical data were recorded during the suspected infection episode. A validated sepsis score (NEO-KISS) was used to retrospectively determine the presence of sepsis (independent of blood culture). The performance of each of the tests were compared by sensitivity, specificity, positive/negative likihood ratios (+/-LR) and postive/negative predictive values (PPV/NPV).

RESULTS: Sixty-five babies with suspected clinical sepsis were prospectively included. The performance indicators are presented with 95% confidence limits. For the detection of bacteria, blood culture had sensitivity of 0.57 (0.34-0.78), specificity of 0.45 (0.30-0.61); +LR of 1.05 (0.66-1.66) and-LR of 0.94 (0.52-1.7); PPV of 33.3 (18.56-50.97) and NPV of 68.97 (49.17-87.72). Serum CRP had sensitivity of 0.92 (0.64-1) and specificity of 0.36 (0.17-0.59); +LR of 1.45 (1-2.1) and-LR of 0.21 (0.03-1.5); PPV of 44.46 (26.6-66.6) and NPV of 88.9 (51.8-99.7). The universal molecular assay had sensitivity of 0.76 (0.53-0.92), specificity of 0.95 (0.85-0.99); +LR of 16.8 (4.2-66.3) and-LR of 0.25 (0.1-0.5); PPV of 88.9 (65.3-98.6) and NPV of 89.4 (76.9-96.5).

CONCLUSIONS: In VLBW infants this universal molecular assay performed better in the diagnosis of late onset sepsis (LOS) than blood culture and CRP. Further development is required to explore and improve the performance of the assay in real-time diagnosis.

Clustering-Based Ensemble Method as an Alternative to Stacking

One of the most popular techniques of generating classifier ensembles is known as stacking which is based on a meta-learning approach. In this paper, we introduce an alternative method to stacking which is based on cluster analysis. Similar to stacking, instances from a validation set are initially classified by all base classifiers. The output of each classifier is subsequently considered as a new attribute of the instance. Following this, a validation set is divided into clusters according to the new attributes and a small subset of the original attributes of the instances. For each cluster, we find its centroid and calculate its class label. The collection of centroids is considered as a meta-classifier. Experimental results show that the new method outperformed all benchmark methods, namely Majority Voting, Stacking J48, Stacking LR, AdaBoost J48, and Random Forest, in 12 out of 22 data sets. The proposed method has two advantageous properties: it is very robust to relatively small training sets and it can be applied in semi-supervised learning problems. We provide a theoretical investigation regarding the proposed method. This demonstrates that for the method to be successful, the base classifiers applied in the ensemble should have greater than 50% accuracy levels.