Outcome measures must be chosen after considering the need to find enough events of interest, the value of having rich data for analysis, and the match between research goals and the type of information gathered. The five commonly used sources of information about patient violence include police and court records, treatment records, unit incidents and seclusion reports, interviews with patients, and collateral interviews. While triangulating methods may alleviate many of the potential biases of these methods, use of multiple sources can also raise the issue of handling conflicting reports about specific incidents. Techniques used in the complex task of categorizing violent events include coding acts, coding the act in a situational context, coding the processes accompanying a violent incident, and coding opportunities for violence. 41 references