Indicators for estimating trends in alcohol-related assault: evaluation using police data from Queensland, Australia

Monitoring levels of alcohol-related harm in populations requires indicators that are robust to extraneous influence. We investigated the validity of an indicator for police-attributed alcohol-related assault. We summarized offence records from Queensland Police, investigated patterns of missing data, and considered the utility of a surrogate for alcohol-related assault. Of 242 107 assaults from 2004-2014, in 35% of cases the drug used by the offender was recorded as 'unknown'. Under various assumptions about non-random missingness the proportion of assaults judged to be alcohol-related varied from 30%-65%. We found a sharp increase in missing data from 2007 suggesting the downward trend from that point is artefactual. Conversely, we found a stable and increasing trend using a time-based surrogate. The volume of missing data and other limitations preclude valid estimation of trends using the police indicator, and demonstrate how misleading results can be produced. Our analysis supports the use of an empirically-based surrogate indicator.