Discussion
Firearm injury is a public health crisis in the USA claiming more than 300 000 persons in the last decade and injuring far more.1 2 4 14 The annual estimated cost of injuries exceeded $174 billion in 20103 and NF injuries are 40 times more common than fatal injuries.14 Despite small decreases, the incidence of gun violence still far outpaces other nations.3 4 14 Australia adopted a panel of strict gun laws in 1996 after a mass shooting.15 16 As a result, the case fatality rate is now 134 times smaller than in the USA despite only a 13.7-fold population difference.15 16 Likewise, firearm-related homicide is 27 times higher in the USA.15 16
The underlying causes are a complicated, multifaceted issue that spans cultural, economic, behavioral including substance abuse, firearm safety, and legislative factors that contribute to the staggering statistics.2 15 17 Although no single law is likely to have an overwhelming effect, continuing to assess and implement strategies that contribute to driving this rate down is a key principle for curbing this public health emergency. Even former Representative Jay Dickey, the author of the amendment banning gun violence research, has evolved to state that funding is needed and that “doing nothing is no longer an acceptable solution.”6
This study examines the role of legislative action on injuries through expanded restriction of open carry laws. California has the strongest gun laws, ranking highest on the Brady scale in the nation with 104 separate laws.13 18 19 The impact of open carry laws on firearm injuries had not previously been investigated. Demonstrating the expansion of the open carry law had a measurable effect on decreasing firearm death and injury in a state already with the most restrictive gun regulations in the nation suggests that this type of gun control measure has incremental added benefit.
The ABA and multiple physician stakeholder groups have jointly advocated for both increased research and regulations.3 20 To be accepted, legislation must respect the second amendment, be rational in implementation, and based on fundamentally sound research. Fatality rates are lower in states that have more restrictive overall gun laws.21–23 However, simply having more laws is not the answer22; uncovering which specific laws actually make a difference is important for both public safety and advocacy efforts. For example, the ban on open carry had only a small effect on suicide rates and this is not surprising. Overall access to guns is a far more important factor in suicide prevention.1 24 25
California also has a sizeable number of gun owners with over 2.9 million handguns purchased in the last 9 years, with men having the highest rate at 242 handguns per 10 000 persons compared with a rate of 25 per 10 000 for women.1 There is also racial differences (white: 209 per 10 000; black: 80 per 10 0001 26). Multiple studies have shown that firearm ownership is a risk factor for both suicide and homicide.1 3 7–9 19 27 28 Interestingly, the largest reduction in firearm fatalities and hospital utilization in this study was seen in white men. This suggests that this law was particularly effective in this subgroup.
Homicide and NF injury disproportionately affect black men and women.14 22 29 The case fatality rate was greater than fourfold higher in both black men and women compared with white men and women, respectively. Unfortunately, the implementation of the open carry ban on unloaded guns had minimal effect on blacks. Although only speculative, this further highlights that there are likely differing risk factors across racial groups, making subgroup analyses of legislative changes extremely important.
Although suicide makes up the largest proportion of firearm deaths, those from homicide have fueled the debate on tougher gun control legislation. As mass shootings have become more common, the debate over gun control has found renewed visibility.28 The reality is that the number of persons killed or injured each year is astronomically higher from interpersonal violence than from mass shootings. In fact, mass shootings make up less than 1% of all firearm deaths.1 Contributing to the controversy is a lack of literature examining the specific impact of the new legislation.
The DID regression analysis is a commonly employed strategy in health policy research when assessing the longitudinal outcomes after changes in legislation.11 Research approaches that compare outcomes with differing state gun laws can be biased by the multitude of causative factors affecting gun violence that are not able to be controlled for in the analyses.18 Reporting before and after results in the intervention group would not be sufficient to draw conclusions as there are multiple other unmeasured confounders that could have effective changes in case numbers.
The DID is a robust alternative method for overcoming these potential biases, and any differences found between groups are interpreted as being a causal effect of the policy.11 This technique is particularly important when observational studies or randomized control trials are not available. The findings are most helpful when comparison groups are large and the data span enough time to see the longitudinal impact of a legislative change.18 This study included a large number of states in the control group with differing overall strength of gun laws and used multiyear data. This provided enough observations and statistical power to uncover important differences in case fatality rates and hospitalization usage.
The DID also relies on two key assumptions. The first assumption is parallel trends. The trends in outcomes between the treated and comparison groups must be the same prior to the intervention. If they are not, the DID cannot be used. Regression modeling was used to determine if the trends are statistically different across the years of the preintervention. There was no statistically significant difference in trends in the preintervention period within or between groups. In the sensitivity analysis, the results remained statistically significant, supporting the difference seen was not attributable to a violation of the parallel trend assumption.
The second key assumption is that of “common shocks assumptions.” This assumption states “that any events occurring during or after the time of the policy changed will equally affect the treatment and comparison groups.”30 These are considered unexpected and unpredictable events that are unrelated to the policy.30 In our study time period, there was several highly visible events in the USA involving firearms, including the Fort Hood shooting (Texas 2009), Binghamton shootings (New York 2009), the Geneva county shooting (Alabama 2009), the Sandy Hook Elementary School shooting (Connecticut 2012), the Aurora Theater shooting (Colorado 2012), and the Washington Naval Yard shooting (DC 2013).
There is no plausible reason to think that the control or treatment groups would have been differentially affected by these events which were widely carried on national news coverage. However, it is acknowledged that there could be unaccounted variables that changed over time that could not be measured or controlled for that differentially affected a control state. The DID analysis is strongest when control groups are as large as possible to minimize the effect of this unknown “common shock.” Therefore, the control group includes all the states with stable gun laws during the study period.
The 2012 ban on open carry resulted in a significant decrease in both firearm-related fatalities and hospital utilization. Changing the law led to 3.7% less fatalities and 6.5% fewer hospital visits in California. This translates to 337 saved lives and 1285 fewer hospital visits. Although these decreases are modest, extrapolated nationwide, this would represent almost 1200 fewer deaths per year and over 4800 fewer costly hospital visits. Although the data demonstrate the net effect of banning open carry was a reduction in fatalities and hospital utilization, the vast majority of this effect was from the decrease in assaults. During the last 20 years, efforts have been under way, including legislative actions, to curb loss of life from interpersonal gun violence. These data suggest that these legislative actions have an effect on at least a portion of the intended at-risk groups, but that effect varies by race.
This study has several limitations. The most significant is the limited number of Brady grade A states with no changes to open carry during the study period. New York had no appreciable change in their Brady grade, but did enact a controversial law, the Secure Ammunition and Firearms (SAFE) Act, during 2013. This law was widely challenged in court and was not settled until 2015. The act contained multiple provisions including the ban on assault weapons and expanded background checks, and made internet sales of ammunition illegal. Although there was no change in Brady grade, the inclusion of New York in the controls could have affected our results; however, it would be expected to bias the results toward the null as these provisions strengthened laws. We also performed a sensitivity analysis without New York in the controls. The DID analysis remained significant (p=0.002) and the net effect of the difference in fatality rates between California and the control states was even greater. Therefore, we think that our inclusion of New York provides a conservative estimate of the net effect of the ban on open carry.
Further limitations include the small number of states reporting data for NF injuries in publicly available data sets. The control group for these analyses relies on only two comparison states. Although this is considered an adequate sample size in a DID analysis, the limited variety of the included states could have biased the results. Additionally, the categorization of race is limited to only white, black, and other due to the restrictions in reporting for CDC data. Finally, case numbers are quite small, especially for women by race categories, making reporting relative changes somewhat misleading. Large relative changes can be seen for incidence rates with quite small absolute changes.