Citation

Miller M, Swanson SA, Azrael DR. Epidemiol. Rev. 2016; 38(1): 62-69.

Copyright

(Copyright © 2016, Oxford University Press)

DOI

10.1093/epirev/mxv011

PMID

26769723

Abstract

Despite the magnitude and consistency of risk estimates in the peer-reviewed literature linking firearm availability and suicide, inferring causality has been questioned on the theoretical basis that existing studies may have failed to account for the possibility that members of households with firearms differ from members of households without firearms in important ways related to suicide risk. The current bias analysis directly addresses this concern by describing the salient characteristics that such an unmeasured confounder would need to possess in order to yield the associations between firearm availability and suicide observed in the literature when, in fact, the causal effect is null. Four US studies, published between 1992 and 2003, met our eligibility criteria. We find that any such unmeasured confounder would need to possess an untenable combination of characteristics, such as being not only 1) as potent a suicide risk factor as the psychiatric disorders most tightly linked to suicide (e.g., major depressive and substance use disorders) but also 2) an order of magnitude more imbalanced across households with versus without firearms than is any known risk factor. No such confounder has been found or even suggested. The current study strongly suggests that unmeasured confounding alone is unlikely to explain the association between firearms and suicide.

Language: en