Predictive diagnostic value of the tourniquet test for the diagnosis of dengue infection in adults

OBJECTIVE To examine the accuracy of the admission tourniquet test in the diagnosis of dengue infection among Lao adults. METHODS Prospective assessment of the predictive diagnostic value of the tourniquet test for the diagnosis of dengue infection, as defined by IgM, IgG and NS1 ELISAs (Panbio Ltd, Australia), among Lao adult inpatients with clinically suspected dengue infection. RESULTS Of 234 patients with clinically suspected dengue infection on admission, 73% were serologically confirmed to have dengue, while 64 patients with negative dengue serology were diagnosed as having scrub typhus (39%), murine typhus (11%), undetermined typhus (12%), Japanese encephalitis virus (5%), undetermined flavivirus (5%) and typhoid fever (3%); 25% had no identifiable aetiology. The tourniquet test was positive in 29.1% (95% CI = 23.2-34.9%) of all patients and in 34.1% (95% CI = 27.0-41.2%) of dengue-seropositive patients, in 32.7% (95% CI = 23.5-41.8) of those with dengue fever and in 36.4% (95% CI = 24.7-48.0) of those with dengue haemorrhagic fever. Interobserver agreement for the tourniquet test was 90.2% (95% CI = 86.4-94.0) (Kappa = 0.76). Using ELISAs as the diagnostic gold standard, the sensitivity of the tourniquet test was 33.5-34%; its specificity was 84-91%. The positive and negative predictive values were 85-90% and 32.5-34%, respectively. CONCLUSIONS The admission tourniquet test has low sensitivity and adds relatively little value to the diagnosis of dengue among Lao adult inpatients with suspected dengue. Although a positive tourniquet test suggests dengue and that treatment of alternative diagnoses may not be needed, a negative test result does not exclude dengue.