State Gun Laws and Pediatric Firearm-Related Mortality Available to Purchase

Goyal et al [1] use cross-sectional data to assess whether firearm mortality among US children and young adults is related to state-level firearm-related characteristics, including a summary score of overall firearm legislation strength, select firearm laws, and a survey-based measure of household firearm ownership. They conclude that their findings “suggest that strict firearm legislation may be protective of children even in areas of high gun ownership.” 

For several reasons, their approach cannot provide insight into the effect of legislation on mortality. The reported association does, however, raise the question:  what likely causal factor makes states that have passed stricter gun laws more likely to have lower firearm-related deaths?  To Goyal et al’s credit, they include a variable in their models for the strongest correlate of state-level firearm mortality that has been identified:  household firearm ownership. Unfortunately, the data they use, and the way they operationalize their adjustment for firearm prevalence, is flawed. First, the data come from a survey of only a few thousand people (while nationally representative it is not representative at the state-level). Second, their decision to dichotomize gun ownership, rather than use it as a continuous measure, severely undercuts the very adjustment inclusion of this variable was intended to accomplish.

The problems that these data analytic decisions entail are illustrated by considering Hawaii, a state that validated surveys have consistently found to have one of the lowest firearm ownership rates in the US and that Goyal et al’s underlying data indicates has the lowest firearm mortality rate in the entire dataset and the third highest Brady Score (see Table). YouGov estimates a firearm prevalence rate of 45%--well over the national average. This discrepancy is consequential: using YouGov as a continuous variable and excluding Hawaii from analyses attenuates Goyal et al’s findings to statistical non-significance, and, substantively, from an inverse association with a correlation coefficient of -0.5 to a correlation coefficient of -0.2. Using a validated cross-sectional proxy for firearm prevalence, the percentage of suicides involving firearms [2], likewise attenuates the legislation-mortality association to non-significance with a coefficient of approximately 0.2. Further, had they stratified their main finding by their (flawed) classification of high and low gun states, they would see that the legislative effect they report is apparent only in the high gun states --only one of which, Colorado, passed any of the laws that they point to as of particular importance.

In short, Goyal et al's study does not show, one way or the other, whether legislation affects firearm mortality. Instead, because it is easier to pass firearm laws in states with fewer guns (see Table), what their study points to is the well-established relationship between firearm prevalence and firearm mortality [3,4,5]. The danger in publishing findings like these, when misinterpreted as suggesting causality (in either direction), is that it diverts attention away from the availability of the firearm itself, which, in many instances (eg, suicide prevention), is more likely to be affected by non-legislative pathways, such as counselling at risk families to remove guns from their homes.

REFERENCES

[1] Goyal MK, Badolato GM, Patel SJ, et al. State Gun Laws and Pediatric Firearm-Related Mortality. Pediatrics. 2019;144(2):e20183283

[2] Azrael D, Cook P, Miller M.  State and local prevalence of firearm ownership measurement, structure and trends. J Quantitative Criminology 2004; 20(1):43-62

[3] Miller M, Azrael D, Hemenway D. Firearm availability and unintentional firearm deaths, suicide and homicide among 5-14 year olds. J of Trauma 2002; 52:267-275.

[4] Miller M, Barber C, Azrael D, White R.  Firearms and suicide in the United States: is risk independent of underlying suicidal behavior? Am J Epidemiol 2013 Sep 15;178(6):946-55. doi: 10.1093/aje/kwt197. Epub 2013 Aug 23.

[5] Anglemyer A, Horvath T, Rutherford G. The accessibility of firearms and risk for suicide and homicide victimization among household members: a systematic review and meta-analysis. Ann Intern Med. 2014;160(2):101–110pmid:24592495

TABLE

Goyal et al. use cross-sectional data (2011-2015) to assess whether firearm mortality among US children and young adults is related to several state-level firearm-related characteristics, including a summary score of overall firearm legislation strength, select firearm laws, and a survey-based measure of household firearm ownership. They conclude that the data they present “suggest that strict firearm legislation may be protective of children even in areas of high gun ownership.”

For several reasons, Goyal et al.’s cross-sectional approach, as they acknowledge but then fail to reckon with, cannot in fact provide insight into the effect of legislation on their outcome. Their observation of an association between the strength of firearm legislation and firearm mortality does, however, raise the question: what is it about states that have passed stricter gun laws, compared with those that have passed less stringent laws, that is likely to be causally related to lower rates of gun deaths? To Goyal et al.’s credit, they include a variable in their models for the strongest correlate of state-level firearm mortality that has been identified: state-level household firearm ownership. Unfortunately, the data they use, and the way they operationalize their measure of firearm prevalence, is flawed. First, the data come from a survey of only a few thousand people that the firm conducting the survey warns should not be used to estimate state-level firearm prevalence. Second, even were the data appropriate to use (or had they used state level firearm ownership or proxy data that are available and validated), Goyal et al.’s decision to dichotomize gun ownership, rather than use it as a continuous measure at the state level, severely undercuts the very adjustment that inclusion of this variable is intended to accomplish.

One discrepancy that should have given the authors pause is that Hawaii, a state that validated state-level surveys have consistently found to have a low firearm ownership rate (in the 10-15% range, see Table 1), has, in YouGov, a firearm prevalence rate of 45%, well over the national average. It turns out that this discrepancy is also consequential: if one uses YouGov as a continuous variable (note that Goyal et al. provide no rationale for dichotomizing) and excludes Hawaii from analyses, the association they report between the strictness of firearm legislation and firearm mortality attenuates not only to statistical non-significance, but from an inverse association with a correlation coefficient of -0.5 to a correlation coefficient of -0.2 (using a validated cross-sectional proxy for firearm prevalence, the percentage of suicides involving firearms , instead of the YouGov data, the correlation coefficient likewise attenuates to non-significance with a coefficient of approximately 0.2). Further, if Goyal et al. looked at their findings stratified by their classification of high and low gun states, they would see that the legislative effect they report overall can be seen only in the high gun states --only one of which, Colorado, passed any of the laws that they point to as of particular importance.

In the end, Goyal's study does not show, one way or the other, whether legislation affects firearm mortality. Instead, because it is easier to pass firearm laws in states with fewer guns (the Brady Law Scorecard turns out to be a modestly good proxy for household firearm ownership, rho >-0.8 –Table 1), what Goyal's study points to is the well-established relationship between firearm prevalence and firearm mortality. , , The danger in looking to unsubstantiated legislative effect and not to the availability of the firearm itself, which can be affected by non-legislative as well as legislative pathways, is that we will congratulate ourselves on having saved lives by passing legislation when, in fact, other approaches, such as counselling at risk families to remove guns from their homes, are more likely in many instances (e.g., suicide prevention) to prove effective.

Matthew Miller, MD, MPH, ScD,
Northeastern University
Deborah Azrael, PhD
Harvard University

Table 1:
Please note I was unable to upload the table of data that I want to and that adds an important empirical data set for readers. Please help. Thanks.

This is stated in the background section but it disagrees with the two most recent studies on CBCs from JHU/Bloomberg and UC Davis/VPRP:

"Recently, 3 state laws were associated with a reduction in overall deaths from firearms: universal background checks for firearm purchases, universal background checks for ammunition purchases, and identification requirement for firearms."

The two most-recent studies on CBCs state:

1. The simultaneous implementation of CBC and MVP policies was not associated with a net change in the firearm homicide rate over the ensuing 10 years in California. The decrease in firearm suicides in California was similar to the decrease in nonfirearm suicides in that state.
https://www.sciencedirect.com/science/article/abs/pii/S1047279718306161

2. The average rates of firearm homicide and suicide in Indiana and Tennessee following repeal were within the range of what could be expected, given natural variation
https://www.ncbi.nlm.nih.gov/pubmed/29613872

What data are you using that shows universal background checks are associated with a reduction in firearm deaths?

Goyal et al use cross-sectional data (2011-2016) to assess the association between firearm mortality and several state-level firearm-related characteristics, including a measure of the strength of firearm legislation, presence of select firearm laws, and an unvalidated survey based measure of state-level household firearm ownership. In the authors’ words, their objective was “to determine if stricter firearm legislation at the state level is associated with lower pediatric firearm-related mortality.” Unfortunately, the study is unable to meet even this limited objective because, fundamentally, one cannot assess the effect of legislation, firearm or other, using cross sectional data.

The authors, relative new-comers to the firearm research field, also appear to be unacquainted with the limitations of the data upon which they rely, such as using a 2013 YouGov survey of a few thousand people to assess state-level firearm prevalence. As stated by the market research company that conducted the survey, while nationally representative, survey estimates are not representative of gun ownership at the state level. That the data are similarly misused in the widely challenged paper they cite as the source of these data (and of the particular laws Goyal et al focus on as a secondary outcome), is one reason we felt compelled to call out this common misappropriation. Further, even if the You.gov survey were, by chance, representative at the state level, Goyal et al’s decision to dichotomize states into high vs. low gun ownership as a way to adjust for firearm prevalence, when many states have prevalence estimates that cluster in the mid-range, necessarily compromises all but rhetorically the very adjustment that inclusion of this variable is intended to accomplish. Likewise, the author’s decision to pool all firearm deaths (suicide, homicide, unintentional firearm deaths) rather than look at outcomes separately betrays an uninformed perspective of the ways and extent to which these deaths might reasonably be affected by legislation. Take for example their decision to examine ballistic fingerprinting laws (along with background check laws for ammunition and firearms), that, on their face, are unlikely to produce changes in youth firearm mortality. Whatever value one might postulate ballistic fingerprinting might provide with regard to preventing homicide among adults, what possible, let alone measurable value could it have in preventing suicide when the gun used in the suicide is itself ends up lying next to the decedent (and is usually a gun that belongs not to the child but to the child’s parent)? Lastly, including 21-year-old adults along with younger adolescents and children strongly skews the death data towards older youth (and 21-year-old adults), undercutting the putative motivation for the study itself – to better understand understudied pediatric deaths.

In the end, what the Goyal et al's published data show is that in states where more and stricter laws are passed, more youth die by gunfire. Because it is easier to pass laws in states with fewer guns (an empirical observation borne out by the high correlation between firearm prevalence and measures of gun law strictness), what Goyal's study points to is the well-established relationship between firearm prevalence and firearm mortality among children.

Sincerely,
Matthew Miller, MD, MPH, ScD
Northeastern University
Deborah Azrael, PhD
Harvard University