摘要
Background: Antibiotics resistance threats Tuberculosis control, being crucial to work on unbiased MDR-TB images. The decision of testing is clinical, non-random, raising extrapolation problems. Aim: To evince and describe non-random testing practices;develop and apply a coherent and intuitive method for estimating global corrected resistance prevalences (2000-2009). Methods: A quantitative approach upon National Tuberculosis Database was undertaken, to assess testing potential predicting factors. Different factors structures in tested and non-tested cases were characterized (regarding socio-demographic and clinical variables), through binary logistic regressions. Estimated multirresistance prevalences were corrected using the essayed model. Results: Only 32% of cases had been tested, where MDR-TB prevalence was 2.38%. All factors influenced the practice of testing (p < 0.05). Corrected resistance estimates in non-tested ranged 1.96% - 2.71%, and the global weighted average found ranged 2.07% - 2.51%, depending on the chosen strata structure. Conclusions: MDR-TB prevalence representation must consider patients’ characteristics influencing testing. The correction method improved prevalences interpretation substantially;corrected and conventional values were close, because tested and non-tested had similar structures. But in other settings or health problems, correcting such estimates can make a relevant difference.
Background: Antibiotics resistance threats Tuberculosis control, being crucial to work on unbiased MDR-TB images. The decision of testing is clinical, non-random, raising extrapolation problems. Aim: To evince and describe non-random testing practices;develop and apply a coherent and intuitive method for estimating global corrected resistance prevalences (2000-2009). Methods: A quantitative approach upon National Tuberculosis Database was undertaken, to assess testing potential predicting factors. Different factors structures in tested and non-tested cases were characterized (regarding socio-demographic and clinical variables), through binary logistic regressions. Estimated multirresistance prevalences were corrected using the essayed model. Results: Only 32% of cases had been tested, where MDR-TB prevalence was 2.38%. All factors influenced the practice of testing (p < 0.05). Corrected resistance estimates in non-tested ranged 1.96% - 2.71%, and the global weighted average found ranged 2.07% - 2.51%, depending on the chosen strata structure. Conclusions: MDR-TB prevalence representation must consider patients’ characteristics influencing testing. The correction method improved prevalences interpretation substantially;corrected and conventional values were close, because tested and non-tested had similar structures. But in other settings or health problems, correcting such estimates can make a relevant difference.
