[The application of decision trees for constructing an algorithm for the differential diagnosis of zoonotic infections].

AIM: To attempt to construct an algorithm using the routine epidemiological, clinical and laboratory data for the differential diagnosis of ixodid tick-borne borreliosis (ITBB) caused by Borrelia miyamotoi (BM-ITBB) and other zoonotic infections that are endemic in Russia.

SUBJECTS AND

METHODS

The investigation enrolled the adult patients treated at the Republican Hospital for Infectious Diseases (Izhevsk) in 2010-2012 with diagnoses of BM-ITBB (n = 71), Lyme disease (n = 38), tick-borne encephalitis (TBE) (n = 25), and hemorrhagic fever with renal syndrome (n = 27). The Decision Tree procedure in IBM SPSS Statistics was used to analyze more than 65 variables characterizing a disease case.

RESULTS

The final decision tree had 7 dichotomous fissions in accordance with the values of several indices (presence of erythema migrans, tick bite, goat's milk consumption, sweating, vertigo, nausea, abdominal pain, as well as blood concentrations of platelets, alanine aminotransferase, and count, and urea) and formed 8 terminal nodes. The proposed algorithm provides correct classification in 95% of disease cases.

CONCLUSION

ITBB caused by B. miyamotoi can be successfully discriminated from other widespread zoonotic infections. Thus, practitioners have an additional opportunity to detect and diagnose the "new" infection BM-ITBB.