Citation

Amorim M, Ferreira S, Couto A. Transp. Res. Proc. 2014; 3: 224-233.

Copyright

(Copyright © 2014, Elsevier Publications)

DOI

10.1016/j.trpro.2014.10.108

PMID

unavailable

Abstract

Road safety research, in particular road and traffic safety evaluation research, is highly applied and carried out mostly to help reducing the number of road accidents and the injuries resulting from them. This subject has been continuously studied, and in developed countries road safety is improved in a way that, more and more, new measures have less visible impact. Although measures are usually taken directly in the source, which makes all the sense, it is possible to reduce the accident impact if improvements are made "a posteriori"; this is, improving the emergency system to minimize the socioeconomic impact of the accident. In order to study accidents impact it is necessary to merge two different datasets - police and hospital. This process is known as data linkage and besides a manual linkage process there are three main numerical methodologies: deterministic record linkage, fuzzy matching and probabilistic record linkage. Because these types of datasets are usually protected by anonymity, unique identifiers are not possible to achieve, thus the probabilistic record linkage is usually the chosen method. This paper presents a concept for an algorithm based on the databases' demographics. By analyzing the various demographic fields it is possible for the algorithm to calculate individual weights that depend on the occurrence of each fields' values among a specific dataset. The demographics are based on the case of Gaia's city road record accidents.

Language: en

Keywords

traffic injury; linkage algorithm; numerical methods; police data; hospital discharge data; healthcare data