Road Injury Information Program - Report Series, Number 7

Report information
Report No. RIIP-7 Date April 1995 ISBN 0 642 22704 7

Title:
The Linkage of Hospital and Police Information on Road Crash Casualties: An Investigation of Alternative Methods

Author(s):
Rosman D.L.

Performing Organisations:
Road Accident Prevention Research Unit
Department of Public Health
University of Western Australia
Nedlands W.A. 6009
Australia

Sponsor:
AIHW National Injury Surveillance Unit
Mark Oliphant Building
Laffer Drive
Bedford Park S.A. 5042
Australia

Abstract:

By linking together data routinely collected by the various authorities dealing with road crash casualties, comprehensive information can be assembled for the analysis of outcomes of road crashes.

The objective of this study was to assess the feasibility of linking hospital and police records which do not contain casualty names.

A three month sample of records from the Western Australian Road Injury Database (RID), linked previously using named data and the Generalised Interactive Record Linkage System (GIRLS), was used to investigate different linkage methods. The effect of the level of identifying information on the quantity and quality of links was tested by re-linking the records under different conditions. The GIRLS links were used as the 'gold standard' against which all subsequent linkages using different methods were assessed.

It was found that the dates of crash occurrence and hospital admission, even when used in conjunction with age and sex for matching hospital and police records, was insufficient to produce reliable links without the added discriminating power of the family name of the casualty. The false positive rate of linkage of hospital and police records was 56%, under conditions which derived a true positive link rate of 52% using variables other than name.

Based on the results of this study, it was recommended that if hospital and police data are to be linked both data sets should include family name ideally in full form or, less desirably, in coded form.

TABLE OF CONTENTS:

EXECUTIVE SUMMARY:

INTRODUCTION:

Background

Objectives

METHODS:

Probabilistic linkage

Internal linkage

Data sources

Uniqueness of identifiers

Comparison of SAS/Links and GIRLS

Comparison of identifiers

RESULTS:

Internal links

Uniqueness of identifiers

Deterministic linkage

Probabilistic linkage

Linkage using full names

Linkage using NYSIIS

Linkage using SOUNDEX

Linkage using unnamed data

DISCUSSION:

CONCLUSIONS AND RECOMMENDATIONS.

REFERENCES.

APPENDIX I - Phonetic Codes

APPENDIX II - Methods

APPENDIX III - Results