RESUMO
BACKGROUND: Missing a fracture in a child on skeletal surveys for suspected non-accidental trauma can have devastating results. Double-read has the potential to improve fracture detection. However the yield of double-read is unknown. OBJECTIVE: To determine the advantage of double-read versus single-read of radiographic skeletal surveys for suspected non-accidental trauma. MATERIALS AND METHODS: The study was performed in two phases. In the first phase (April 2013 to September 2013), double-read was performed for all skeletal surveys obtained during weekday working hours. Because we had no new double-read findings in studies initially read as negative, we conducted a second phase (January 2014 to March 2014). In the second phase we limited double-reads to skeletal surveys found positive on the first read. At the end of this period, we retrospectively performed double-read for all initially negative skeletal surveys. We excluded follow-up skeletal surveys. The difference in discrepancy (new fracture or false diagnosis of a fracture) ratio between negative and positive skeletal surveys was evaluated using the Fisher exact test, and change in discrepancy ratio between the first and second study phases was evaluated using the stratified Cochran-Mantel-Haenszel test. RESULTS: Overall in the two phases, 178 skeletal surveys were performed in 178 children (67 girls) with mean age of 9 months (range 3 days to 3.7 years). Double-read found 16 discrepancies in 8/178 (4.5%) skeletal surveys. Seven of these studies showed additional fractures (n=15). In one study, an initial read of a skull fracture was read as a variant on the second read. There was a significant (P=0.01) difference between rate of disagreement in negative skeletal surveys (1/104, 1.0%) and positive skeletal surveys (7/74, 9.5%). No significant change in disagreement rate was demonstrated between the two phases of the study (P=0.59). CONCLUSION: Double-read of skeletal survey for suspected non-accidental trauma found false-negative fractures in a few cases and rarely found false-positive diagnosis of a fracture. Double-read uncommonly found discrepancies in an initially normal skeletal survey. Limiting double-read to initially positive studies improves the yield of the double-read.