Background CD64 (a neutrophil surface marker) upregulates in infection. How markers function as a point of care test (POCT) is influenced by condition prevalence. Proportionate reduction in uncertainty (PRU) is then the clinically relevant factor.

Objective To show, using PRU curves, how CD64 performs as a POCT in preterm neonates screened for late onset infection (LOI).

Methods CD64 was measured in preterm infants screened for LOI. Ethics and parental consent were obtained. Episodes were retrospectively classified as infected if blood culture (BC) positive (BC+), BC negative but 5 days treatment given (BC-), there was local infection, pneumonia or NEC.

Results 50 screenings were obtained in 38 infants, median gestation 26 weeks, birthweight 825 g. 58% of episodes were infective (BC+ (9), BC- (6), local (4), NEC (11)), falling to 38% if local infections and conservatively managed NEC were excluded. Clinicians at screening felt that 82% were definitely or probably infected, 18% being screened to rule out infection. PRU curves were developed to show how CD64 functions to reduce remaining uncertainty at various cut-off levels. A CD64 value of >3.4 reduced the remaining uncertainty of infection by a further 25%. In contrast a value of <3.4 was poor at ruling out infection, reducing residual uncertainty by only 10%.

Conclusions CD64 may assist decisions, but correct application of this and other POCT depends on clinicians' understanding of residual uncertainty, dependent on an understanding of disease prevalence and test performance.