This is another "someday, in the future" article that made the rounds with the news releases yesterday - where, supposedly, within a few hours of infection, there are significant differences in phagocyte chemiluminescence that allow researchers to differentiate between viral and bacterial infections.
As usual, the breathless commentary is a little ahead of the actual research results. What the authors did was a data-mining experiment from 69 patients, each of whom had been diagnosed (through standard clinical practice) with either a viral infection, or a bacterial infection. They ran all the polymorphonuclear leukocytes through their assay, recorded several different sorts of chemoluminescence, and then let computer software do a partitioning analysis to determine the most predictive patterns for bacterial and viral infections.
The software trained to 94.7% accuracy on the "knowns", and then, when tested on a confusion sample with 18 "unknowns" it was 88.9% accurate.
So, still not good enough for clinical use as a dichotomous result, but if it were allowed to return an equivocal range that quantified the assay uncertainty, then perhaps it could have a role in clinical practice. In theory, an assay such as this might otherwise reduce additional testing and help reduce the number of viral infections that receive antibiotics.
"Differentiation Between Viral and Bacterial Acute Infectious Using Chemiluminescent Signatures of Circulating Phagocytes"