ABSTRACT
BACKGROUND: Nosocomial outbreaks of tuberculosis (TB) have been attributed to unrecognized pulmonary TB. Accurate assessment in identifying index cases of active TB is essential in preventing transmission of the disease. OBJECTIVES: To develop an artificial neural network using clinical and radiographic information to predict active pulmonary TB at the time of presentation at a health-care facility that is superior to physicians' opinion. DESIGN: Nonconcurrent prospective study. SETTING: University-affiliated hospital. PARTICIPANTS: A derivation group of 563 isolation episodes and a validation group of 119 isolation episodes. INTERVENTIONS: A general regression neural network (GRNN) was used to develop the predictive model. MEASUREMENTS: Predictive accuracy of the neural network compared with clinicians' assessment. RESULTS: Predictive accuracy was assessed by the c-index, which is equivalent to the area under the receiver operating characteristic curve. The GRNN significantly outperformed the physicians' prediction, with calculated c-indices (+/- SEM) of 0.947 +/- 0.028 and 0.61 +/- 0.045, respectively (p < 0.001). When the GRNN was applied to the validation group, the corresponding c-indices were 0. 923 +/- 0.056 and 0.716 +/- 0.095, respectively. CONCLUSION: An artificial neural network can identify patients with active pulmonary TB more accurately than physicians' clinical assessment.
Subject(s)
Diagnosis, Computer-Assisted , Neural Networks, Computer , Patient Admission , Tuberculosis, Pulmonary/diagnosis , AIDS-Related Opportunistic Infections/diagnosis , AIDS-Related Opportunistic Infections/epidemiology , Cross Infection/diagnosis , Cross Infection/epidemiology , Disease Outbreaks , Hospitals, University , Humans , New York , Patient Admission/statistics & numerical data , Predictive Value of Tests , Prospective Studies , Regression Analysis , Reproducibility of Results , Tuberculosis, Pulmonary/epidemiologyABSTRACT
The recent outbreaks of multidrug-resistant strains of M. tuberculosis in health care facilities has increased concern over its transmission in health care facilities. Isolation has been recommended for all patients suspected to have tuberculosis even though the feasibility and the cost of this recommendation can be substantial. We have developed a classification tree using clinical and radiographic data from 277 isolation episodes in patients admitted between August 1992 and March 1994 who required isolation for suspicion of tuberculosis. The classification tree was developed with a sensitivity and negative predictive value of 100% by binary recursive partitioning to predict those patients who are unlikely to require isolation. The predictor variables were upper zone disease on chest radiograph, a history of fever, weight loss, and CD4 count. The tree was validated in a separate cohort of 286 isolation episodes between April 1994 and December 1995. In this validation cohort, no erroneous prediction was made of not isolating a patient with active pulmonary tuberculosis. The classification tree had a sensitivity of 100% (95% confidence interval [CI]: 92.5 to 100%), a specificity of 48.1% (95% CI: 43.8 to 52.4%), and a negative predictive value of 100% (95% CI: 98.5 to 100%). We estimate that the use of the tree could have reduced the number of patients requiring isolation by more than 40% without increasing the risk of cross infection.
