Investigating the impact of measurement uncertainty in the radiographic surveillance of hip displacement in cerebral palsy: A Monte Carlo simulation.

Hip surveillance programmes have greatly improved the management of hip dysplasia in children with cerebral palsy. Reimer's migration percentage is the most common index for quantifying hip dysplasia from planar radiographs. However, measurement uncertainty could undermine the diagnostic accuracy. A Monte Carlo simulation was created to investigate the impact of measurement error on decision making in hip surveillance programmes. The simulation was designed to mimic the annual surveillance of children with cerebral palsy (Gross Motor Functional Classification System levels III-V) between 2 and 8 years of age. Simulation parameters for the natural history of hip dysplasia and measurement error were derived from published data. At each measurement interval, the influence of uncertainty in the measurement of Reimer's migration percentage on decision-making was investigated. The probability of a child being indicated for intervention in error during the course of the simulation was relatively high, particularly in the highest functioning cohort where the positive predictive value of Reimer's migration percentage was at best 70% and at worse less than 20%. Including a rate of progression term within the decision-making algorithm had a negative effect on positive predictive power. This simulation suggests that hip surveillance programmes are sensitive to detecting genuine hip dysplasia but can have poor positive predictive power, potentially resulting in unnecessary indication for intervention.