Hong Kong Med J 2010;16(Suppl 3):S34-7
Evaluation of risk assessment tools and infectious aetiology in cancer patients with fever and neutropaenia in Hong Kong
EP Hui, LKS Leung, F Mo, VTC Chan, ATW Ma, A Poon, EK Hui, SS Mak, M Lai, KIK Lei, BBY Ma, TSK Mok, W Yeo, B Zee, ATC Chan
The Chinese University of Hong Kong: Department of Clinical Oncology, Prince of Wales Hospital
1. A total of 227 cancer patients who developed fever and neutropaenia after chemotherapy were prospectively evaluated according to the Talcott and the Multinational Association of Supportive Care in Cancer (MASCC) risk models.
2. The positive predictive value (PPV) of low-risk prediction by the Talcott model was 84%. The sensitivity (SE), specificity (SP), negative predictive value (NPV) and misclassification rate (MR) were 50%, 72%, 33% and 44%, respectively.
3. The MASCC score of ≥21 identified low-risk patients with a PPV of 86%, SE of 81%, SP of 60%, NPV of 52% and MR of 24%. Of the 160 (70%) low-risk patients, 12.5% developed complications and 1.9% died. In contrast, of the 67 (30%) high-risk patients, 43.3% developed complications and 9% died (P<0.0001).
4. The MASCC risk index was superior to the Talcott risk model in terms of a higher discriminative power for identifying low-risk patients.
5. An infective aetiology was microbiologically documented in 21% of the 227 patients. Gram-negative bacteria were more commonly implicated than Gram-positive bacteria (58% vs 31%).
2. The positive predictive value (PPV) of low-risk prediction by the Talcott model was 84%. The sensitivity (SE), specificity (SP), negative predictive value (NPV) and misclassification rate (MR) were 50%, 72%, 33% and 44%, respectively.
3. The MASCC score of ≥21 identified low-risk patients with a PPV of 86%, SE of 81%, SP of 60%, NPV of 52% and MR of 24%. Of the 160 (70%) low-risk patients, 12.5% developed complications and 1.9% died. In contrast, of the 67 (30%) high-risk patients, 43.3% developed complications and 9% died (P<0.0001).
4. The MASCC risk index was superior to the Talcott risk model in terms of a higher discriminative power for identifying low-risk patients.
5. An infective aetiology was microbiologically documented in 21% of the 227 patients. Gram-negative bacteria were more commonly implicated than Gram-positive bacteria (58% vs 31%).