The effect of legislation on firearm-related deaths in Canada: a systematic review

As well a deeper delve into the study methods would have been beneficial. The Blais et al. 2011 paper only examines the effects of legislation up until 2004. Which means that licensing is not examined at all since one cannot make inferences about this regulation from 3 years of data as licensing was not fully implemented until 2001. It cannot make any claims about licensing. Studies that examined longer time frames such as Langmann 2012 and Langmann 2020 examined longer time periods. A discussion of this would have improved the review.

Time series linear regression studies require the examination of an impact and trend effect. The Blais et al 2011 study only examines impacts and does not look at trends, leaving it prone to erroneous findings as an increasing trend after legislation can produce an increasing trend with what appears to be a sudden lower impact however resulting in overall higher rates of homicide in later years past the intervention. This error leads one to believe there is a benefit when in fact there is a worsening of homicide rates.

The Blais et al 2011 study and the Bridges 2004 study do not use all available independent variables. Blais mentions under a table that they leave out many variables due to collinearity. This is an important discussion point and should have been fleshed out in the review as for time series analysis it is important to include other independent variables that may account for the effect other than just legislation. Otherwise a false positive result can be obtained. The Blais paper should have received a lower NOS score similar to Bridges 2004.

The Langmann 2020 paper is given a score of 3 when it should have received a score of 4 and a study design of “Greatest design suitability” as it used difference-in differences analysis (DiD), a quasi experimental design that uses a concurrent comparison group pre and post the intervention based on the classification scheme in Appendix 5. I would suggest this be changed in the study along with the Quality Assessment second paragraph in order to correct that error. Discussing the benefits of this method would have enhanced the review article as the review rightly mentions that mental health data is not available for the entirely of Canada and hence left out of the analysis. However, the DiD method allows one to account for that by using a control group that has similar features and trends prior to the intervention thus helping account for confounders that are not included in the analysis.