CART Model of Aguiar et al for Predicting Tuberculosis in a Hospitalized Patient

Description

Aguiar et al reported a classification and regression tree (CART) model for evaluating a hospitalized patient suspected of having tuberculosis. This can help to estimate the likelihood of the diagnosis and to identify a patient who should be placed into isolation. The authors are from Federal University of Rio de Janeiro in Brazil.

Patient selection: hospitalized patient suspected of having tuberculosis, in a region with high prevalence

Analysis: classification and regression tree (CART)

Prevalence: 27% (84% pulmonary)

Parameters:

(1) X-ray appearance

(2) weight loss

(3) dyspnea

(4) age in years

(5) HIV status

X-ray	Weight Loss	Dyspnea	Age	HIV Status	Percent
normal or sequelae	NA	NA	NA	NA	3.5%
typical or possible	No	NA	NA	NA	60%
typical or possible	Yes	NA	NA	NA	78%
atypical	NA	No	<= 30	NA	50%
atypical	NA	No	> 30	NA	16%
atypical	NA	Yes	NA	positive	2%
atypical	No	Yes	NA	negative or indeterm	0%
atypical	Yes	Yes	NA	negative or indeterm	38%

Sputum positivity: 62%

Performance:

• The area under the ROC curve was 0.80.

To read more or access our algorithms and calculators, please log in or register.

CART Model of Aguiar et al for Predicting Tuberculosis in a Hospitalized Patient

Description

Specialty