A time-series analysis of trends in firearm-related homicide and suicide.

Abstract

In order to determine whether firearm use was an increasing component of the general pattern of homicide in Cape Town, South Africa, hidden Markov time-series models were used to examine a week-by-week count of firearm homicides, non-firearm homicides, firearm suicides and non-firearm suicides for the 6-year period from 1986 to 1991. Of several models fitted to the proportion of homicides that involved firearms, the one which incorporated a discrete upward shift in the middle of 1991 was the most successful. There was no evidence of a similar upward shift in the proportion of the suicides that involved firearms. The sharp increase in 1991 in the probability that a homicide involved the use of a firearm is consistent with a reported upsurge in violence related to the so-called 'taxi wars'. Hidden Markov models, as a general methodology for the analysis of discrete-valued time series, may be a useful and flexible means of identifying time trends or points of transition related to events or interventions in a wide range of public health contexts.