Comparison of the complete filtration method using an automated feces analyzer with three manual methods for stool examinations.

Conventional diagnostic techniques using manual methods for stool examination have important limitations. Hence there is a need for improved technologies in routine clinical practice. This study aimed to compare detection rates, agreements, and diagnostic performances for stool examinations in all parameters of the complete filtration method using the Sciendox Feces Analysis System-50 automated feces analyzer with three manual methods, the direct smear, Kato's thick smear, and formalin ethyl concentration techniques. The 252 routine stool samples were examined for parasites, white blood cells (WBCs), red blood cells (RBCs), fat globules, and yeast cells using the four methods indicated above, and the complete filtration detection rates, Cohen's kappa (κ), and diagnostic performances were evaluated and compared. The detection rates of RBCs, fat globules, and yeast cells examined by the complete filtration automated method were comparable to the manual methods, but the detection rates of parasites and WBCs were significantly lower. Most methods detected the same seven parasite species, Ascaris lumbricoides, hookworm, Trichuris trichiura, Strongyloides stercoralis, Entamoeba histolytica/dispar, Blastocystis spp., and Giardia intestinalis. Pairwise agreements between the complete filtration and other methods were good to very good for all parameters showing κ values of 0.74 to 0.89. The diagnostic performances against the combined results showed complete filtration method sensitivities of 70%, 81.82%, 77.27%, 100%, and 95% for parasites, WBCs, RBCs, fat globules, and yeast cells, respectively, while the complete filtration negative predictive values (NPVs) and accuracies showed higher than 95% for all parameters. The complete filtration method using the automated feces analyzer showed high NPVs and accuracies, and good agreements with the three tested manual methods for stool examination in all parameters.