Poor yield of Clostridium difficile testing algorithms using glutamate dehydrogenase antigen and C. difficile toxin enzyme immunoassays in a pediatric population with declining prevalence of clostridium difficile strain BI/NAP1/027

We compared the performance of algorithmic Clostridium difficile infection (CDI) diagnosis with four molecular tests in children. Stool samples in patients 1-18 years old were tested with an algorithm (C. Diff Quik Chek Complete (QCC) reflexed to illumigene C. difficile); AmpliVue C. difficile (ACD); Lyra Direct C. difficile (Lyra); BD MAX C diff (BDM); and Xpert C. difficile (XCD). The gold standard was positivity by two tests. Sensitivity, specificity, positive predictive value, and negative predictive value were 85%, 99%, 93%, 97% for the algorithm; 21%, 99%, 78%, 87% for QCC’s toxin component; 94%, 99%, 94%, 99% for ACD; 88%, 99%, 94%, 98% for Lyra; 94%, 100%, 100%, 99% for BDM, and 94%, 99%, 94% and 99% for XCD. 9.6% of samples were ribotype 027. Algorithms may detect CDI with lower sensitivity compared to molecular methods in children. This may be related to low prevalence of NAP-1/ribotype 027. •Sensitivity and specificity of currently available molecular methods of detecting toxigenic Clostridium difficile is satisfactory and generally similar in pediatric stool specimens.•“Algorithmic” detection of toxigenic C. difficile appears to lack sensitivity in children.•The prevalence of the NAP-1/ribotype 027 strain of C. difficile was low in children.