Systematic review with meta-analyses and critical appraisal of clinical prediction rules for pulmonary tuberculosis in hospitals.

OBJECTIVE

To systematically review studies evaluating clinical prediction rules (CPRs) for adult inpatients suspected to have pulmonary tuberculosis.

DESIGN

Systematic review with meta-analyses.

SETTING

Hospitals. Patients Inpatients at least 15 years of age admitted to acute care.

METHODS

A search was conducted in 5 indexed electronic databases with no language or year of publication restrictions. We performed a meta-analysis for those CPRs with at least 2 validation studies. Results were reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

RESULTS

Of the 461 abstracts selected, 36 articles were fully analyzed and 11 articles were included, yielding 8 CPRs derived in 4 countries. Broad validation studies were identified for 2 CPRs. The most frequent clinical predictors were fever and weight loss. All CPRs included chest imaging signs. Most CPRs were derived in countries with a low prevalence of pulmonary tuberculosis and included homeless, immigrants, and those who reacted to the purified protein derivative test. Both of the CPRs derived in countries with a high prevalence of pulmonary tuberculosis strongly relied on chest radiograph predictors. Accuracy of the different CPRs was high (area under receiver operating characteristic curve, 0.79-0.91). Meta-analysis of 4 validation studies for Wisnivesky's CPR indicates optimistic pooled results: sensitivity, 94.1% (95% CI, 89.7%-96.7%); negative likelihood ratio, 0.22 (95% CI, 0.12-0.40).

CONCLUSION

On the basis of a critical appraisal of the 2 best validated CPRs, the presence of weight loss and/or fever in inpatients warrants obtaining a chest radiograph, regardless of the presence of productive cough. If the chest radiograph is abnormal, the patient should be placed in isolation until more specific test results are available. Validation in different settings is required to maximize external generalization of existing CPRs.