Predicting shunt failure on the basis of clinical symptoms and signs in children.

OBJECT: In evaluating pediatric patients for shunt malfunction, predictive values for symptoms and signs are important in deciding which patients should undergo an imaging study, whereas determining clinical findings that correlate with a low probability of shunt failure could simplify management.
METHODS: Data obtained during the recently completed Pediatric Shunt Design Trial (PSDT) were analyzed. Predictive values were calculated for symptoms and signs of shunt failure. To refine predictive capability, a shunt score based on a cluster of signs and symptoms was derived and validated using multivariate methods. Four hundred thirty-one patient encounters after recent shunt insertions were analyzed. For encounters that took place within 5 months after shunt insertion (early encounters), predictive values for symptoms and signs included the following: nausea and vomiting (positive predictive value [PPV] 79%, likelihood ratio [LR] 10.4), irritability (PPV 78%, LR 9.8), decreased level of consciousness (LOC) (PPV 100%), erythema (PPV 100%), and bulging fontanelle (PPV 92%, LR 33.1). Between 9 months and 2 years after shunt insertion (late encounters), only loss of developmental milestones (PPV 83%, LR 36.7) and decreased LOC (PPV 100%) were strongly associated with shunt failure. However, the absence of a symptom or sign still left a 15 to 29% (early encounter group) or 9 to 13% (late encounter group) chance of shunt failure. Using the shunt score developed for early encounters, which sums from 1 to 3 points according to the specific symptoms or signs present, patients with scores of 0, 1, 2, and 3 or greater had shunt failure rates of 4%, 50%, 75%, and 100%, respectively. Using the shunt score derived from late encounters, patients with scores of 0, 1, and 2 or greater had shunt failure rates of 8%, 38%, and 100%, respectively.
CONCLUSIONS: In children, certain symptoms and signs that occur during the first several months following shunt insertion are strongly associated with shunt failure; however, the individual absence of these symptoms and signs offers the clinician only a limited ability to rule out a shunt malfunction. Combining them in a weighted scoring system improves the ability to predict shunt failure based on clinical findings.