Andreas Roposch1, Evangelia Protopapa2, Mario Cortina-Borja2. 1. Institute of Child Health, University College London, London, United Kingdom; Department of Orthopaedic Surgery, Great Ormond Street Hospital for Children, London, United Kingdom. Electronic address: a.roposch@ucl.ac.uk. 2. Institute of Child Health, University College London, London, United Kingdom.
Abstract
OBJECTIVE: To establish clinical diagnostic criteria for developmental dysplasia of the hip (DDH) that model the practices of expert clinicians. STUDY DESIGN: Of 23 clinical criteria for the diagnosis of DDH, ranked in order of diagnostic importance by international consensus, the 7 most highly ranked were placed in all possible combinations to create unique case vignettes. Twenty-six experts rated 52 vignettes for the presence of DDH. We modeled the data to determine which of the 7 criteria were associated with a clinician's opinion that the vignette represented DDH. From the resulting regression coefficients, for each vignette we calculated a probability of DDH. An independent panel rated the same vignettes using a visual analog scale response. We correlated the visual analog scale ratings with probabilities derived from the model. RESULTS: Our model identified 4 of 7 criteria as predictive of DDH (P < .001): Ortolani/Barlow test (β = 3.26), limited abduction (β = 1.48), leg length discrepancy (β = 0.74), and first-degree family history of DDH (β = 1.39). There was substantial correlation between the probability of DDH predicted by the model and that derived from an independent expert panel (r = 0.73; P < .001). CONCLUSION: Weighted clinical criteria for inferring the likelihood of DDH produced consistent results in the judgment of 2 separate groups of experts. Using these weights, nonexperts could establish the probability of DDH in a manner approaching the practice of clinical experts.
OBJECTIVE: To establish clinical diagnostic criteria for developmental dysplasia of the hip (DDH) that model the practices of expert clinicians. STUDY DESIGN: Of 23 clinical criteria for the diagnosis of DDH, ranked in order of diagnostic importance by international consensus, the 7 most highly ranked were placed in all possible combinations to create unique case vignettes. Twenty-six experts rated 52 vignettes for the presence of DDH. We modeled the data to determine which of the 7 criteria were associated with a clinician's opinion that the vignette represented DDH. From the resulting regression coefficients, for each vignette we calculated a probability of DDH. An independent panel rated the same vignettes using a visual analog scale response. We correlated the visual analog scale ratings with probabilities derived from the model. RESULTS: Our model identified 4 of 7 criteria as predictive of DDH (P < .001): Ortolani/Barlow test (β = 3.26), limited abduction (β = 1.48), leg length discrepancy (β = 0.74), and first-degree family history of DDH (β = 1.39). There was substantial correlation between the probability of DDH predicted by the model and that derived from an independent expert panel (r = 0.73; P < .001). CONCLUSION: Weighted clinical criteria for inferring the likelihood of DDH produced consistent results in the judgment of 2 separate groups of experts. Using these weights, nonexperts could establish the probability of DDH in a manner approaching the practice of clinical experts.