Prognostic Model for Progradient Tuberculosis Course in HIV-Infected Children

The aim of the study was to define independent predictors and create a prognostic model for the progradient course of tuberculosis in HIV-infected children. MATERIALS AND METHODS: The prospective observational study of tuberculosis cases of HIV-infected children under 15 years of age has been carried out over 2000–2014 using a continuous sampling method (n=65). Two observation groups were formed: the main group comprising children with a progradient (progressive) type of tuberculosis (n=37) and the comparison group of children with a regradient (regressive) type of the disease (n=28). The logistic regression method was used to create a prognostic model. The quality of model approximation was assessed using maximum likelihood function. Indicators of model goodness of fit are the coefficient of concordance (a permissible level of frequency disagreement is less than 20%) and Hosmer–Lemeshow test. RESULTS: As a result of a paired comparative analysis based on 17 medico-biological, epidemiological and social signs, 11 statistically different parameters have been distinguished. On their basis, a functional prognostic model has been created including six independent predictors: flaws in children observation in the general medical network (b=23.962), absence of Bacillus Calmette–Guérin vaccination (b=20.404), fatal course of tuberculosis in the human source of infection (b=2.762), tuberculosis identification in children under 3 years of age (b=2.620), absence or low adherence to therapy of the latent tuberculosis infection before tuberculosis detection (b=1.859), marked or severe immunodeficiency (b=1.693). The degrees of the risk factors for the progradient tuberculosis course have been established at the following probability values (decile): at 0.3–0.4 the risk is minimal; at 0.5–1.0 the risk is high; at values of 0–0.2 there is no risk of the disease. Programs for quantitative and qualitative assessment of the risk of progradient tuberculosis course in children with HIV infection have been designed to facilitate the model use. CONCLUSION: The presented prognostic model is based on the analysis of the obligatory data in the diagnostic search making its use convenient at any stage of rendering medical aid to HIV-infected children.