Related Experiment Videos
Predicting smear negative pulmonary tuberculosis with classification trees and logistic regression: a cross-sectional
Fernanda Carvalho de Queiroz Mello1, Luiz Gustavo do Valle Bastos, Sérgio Luiz Machado Soares
1Tuberculosis Research Unit, Clementino Fraga Filho Hospital, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil. fcqmello@hucff.ufrj.br
BMC Public Health
|March 1, 2006
Summary
Developing prediction models for smear-negative pulmonary tuberculosis (SNPT) can help screen outpatients in resource-limited areas. These models offer a cost-effective approach to managing tuberculosis cases efficiently.
Area of Science:
- Medical research
- Public health
- Epidemiology
Background:
- Smear-negative pulmonary tuberculosis (SNPT) constitutes 30% of annual pulmonary tuberculosis cases in Brazil.
- Effective screening tools are needed for SNPT diagnosis, particularly in resource-scarce settings.
Purpose of the Study:
- To develop and validate predictive models for SNPT in outpatient settings.
- To create a clinical and radiological prediction score for SNPT risk assessment.
Main Methods:
- Utilized logistic regression and classification and regression tree models.
- Data from 551 patients with suspected SNPT in Rio de Janeiro, Brazil, were used.
- Models were evaluated using the area under the receiver operator characteristic curve, sensitivity, and specificity.
Main Results:
- Achieved sensitivity ranging from 64% to 71%.
- Achieved specificity ranging from 58% to 76%.
Conclusions:
- The developed models show potential as screening tools for estimating SNPT risk.
- These models can optimize the use of expensive diagnostic tests and reduce unnecessary anti-tuberculosis treatment.
- The predictive models offer a cost-effective solution for healthcare networks with limited resources.