Jack Short1, Brian Caulfield2. 1. Department of Civil, Structural and Environmental Engineering, Trinity College Dublin, Dublin 2, Ireland. 2. Department of Civil, Structural and Environmental Engineering, Trinity College Dublin, Dublin 2, Ireland. Electronic address: brian.caulfield@tcd.ie.
Abstract
INTRODUCTION: The study of non-fatal road traffic injuries is growing in importance. Since there are rarely comprehensive injury datasets, it is necessary to combine different sources to obtain better estimates on the extent and nature of the problem. Record linkage is one such technique. METHOD: In this study, anonymized datasets from three separate sources of injury data in Ireland: hospitals, police, and injury claims are linked using probabilistic and deterministic linkage techniques. A method is proposed that creates a 'best' set of linked records for analysis, useful when clerical review of undecided cases is not feasible. RESULTS: The linkage of police and hospital datasets shows results that are similar to those found in other countries, with significant police understatement especially of cyclist and motorcyclist injuries. The addition of the third dataset identifies a large number of additional injuries and demonstrates the error of using only the two main sources for injury data. PRACTICAL APPLICATION: The study also underlines the risk in relying on the Lincoln-Petersen capture-recapture estimator to provide an estimate of the total population concerned. CONCLUSION: The data show that road traffic injuries are significantly more numerous than either police or hospital sources indicate. It is also argued that no single measure can fully capture the range of impacts that a serious injury entails.
INTRODUCTION: The study of non-fatal road traffic injuries is growing in importance. Since there are rarely comprehensive injury datasets, it is necessary to combine different sources to obtain better estimates on the extent and nature of the problem. Record linkage is one such technique. METHOD: In this study, anonymized datasets from three separate sources of injury data in Ireland: hospitals, police, and injury claims are linked using probabilistic and deterministic linkage techniques. A method is proposed that creates a 'best' set of linked records for analysis, useful when clerical review of undecided cases is not feasible. RESULTS: The linkage of police and hospital datasets shows results that are similar to those found in other countries, with significant police understatement especially of cyclist and motorcyclist injuries. The addition of the third dataset identifies a large number of additional injuries and demonstrates the error of using only the two main sources for injury data. PRACTICAL APPLICATION: The study also underlines the risk in relying on the Lincoln-Petersen capture-recapture estimator to provide an estimate of the total population concerned. CONCLUSION: The data show that road traffic injuries are significantly more numerous than either police or hospital sources indicate. It is also argued that no single measure can fully capture the range of impacts that a serious injury entails.
Authors: Daniel Eid; Miguel Guzman-Rivero; Ernesto Rojas; Isabel Goicolea; Anna-Karin Hurtig; Daniel Illanes; Miguel San Sebastian Journal: Am J Trop Med Hyg Date: 2018-01-01 Impact factor: 2.345