Why Didn't We Get Them All?: Analyzing Unlinked Records in California's Linked Homicide File

• NCJ Number: 206510
• Journal: Homicide Studies Volume: 8 Issue: 3 Dated: August 2004 Pages: 311-321
• Author(s): Jason Van Court; Roger B. Trent
• Date Published: August 2004
• Length: 11 pages
• Publication Format: Article
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America
• Note: Special Issue on Missing Data, for additional articles see NCJ-206505-509.

Annotation

This study analyzed unlinked records in California’s linked homicide file.

Abstract

The California Linked Homicide File links two major files containing information on homicides. The basic file includes 34,542 cases from Supplementary Homicide Reports (SHR). Using a probabilistic matching program, 32,122 (93.0 percent) of these cases were linked with statistics from Vital Statistics (VS) death records. This study tried to expand the effort to monitor and study homicide by linking records from the California SHR with records in the estate’s VS death records from 1990 through 1999. By linking two data sources into one file, researchers can study a variety of never before studied characteristics of homicide. The purpose of this study was to reduce the number of missing cases, that is, to increase the number of SHR and VS matched cases. Specifically, this report examined 2,420 unlinked SHR's to determine whether they were a random subset of all California homicides or whether they were distinct and introduced bias into the Linked Homicide File. After including noninjury deaths in the linkage, it was found that there were not significantly more matches found, but these cases are important for future linkages. The 399 added cases improved the match by slightly more than 1 percent but added 7.1 percent more child abuse cases. This is consistent with past research on misclassification of child abuse homicide. The effects of including more than 2 million noninjury cases on false positives is not known. The question remains: are the extra 399 cases obtained worth the potential false matches received when doing this sort of linkage at once or should there be two separate passes where only injury deaths in the first pass, then expand to noninjury deaths to increase the match? It is this problem that needs further research. In future linkages, noninjury deaths should be included, but researchers also need to find out how many false matches are found and revisit match criteria.