Oklahoma CODES

Oklahoma CODES
Crash Outcome Data Evaluation System
College of Continuing Education
University of Oklahoma

Crash-Inpatient Linkage

To investigate the health impacts of crashes, probabilistic matching with CODES2000 software is used to link motor vehicle crash data from the Department of Public Safety (DPS) to inpatient hospital billing data from the Oklahoma State Department of Health. Police officers investigate the crash at the scene and complete reports that include information about the crash. Hospitalization data is billing data submitted annually to OSDH by hospitals. To protect the confidentiality of the persons involved and of health case providers, all linkage is carried out onsite at OSDH and a de-identified linked data set is produced for use by the CODES project.

The first step in the linkage process involves identifying shared information in the files to be linked. The data are then transformed into the same code sets (e.g. male and female are coded M and F in both files). CODES2000 software is then used to link the two files. CODES2000 software uses prior probabilities calculated from the distribution of values in the data files and uses that information to calculate the probability that each crash-inpatient record pair is a match by adding weight when two match fields agree and by subtracting weight when two match fields disagree. Probabilities of a match can range from zero to one. The amount of weight added or subtracted is based on the descriminating power of each match field. Unique identifiers contribute more weight than do less unique identifiers. Match fields are chosen to include information about the crash, the vehicle and the person involved in the crash.