COMMENT & RESPONSEIn Reply We thank Cazes et al for their comments on our recently published study. 1 We wish to respond to their concerns.First, they raise concern for unmeasured confounding. We discuss this at length in our study. Foremost, while specific factors may be unknown (frequently the case in observational studies), this does not automatically render them a confounder. A confounder is defined as a factor associated both with the exposure (in our study, geospatial access to care) and the outcome (in our study, mortality). There is no evidence that injury severity or mode of transport differs across the spectrum of access to care within urban Philadelphia, Pennsylvania. Furthermore, our hierarchical model accounted for between-neighborhood variation in mortality that might conceivably arise due to other unmeasured differences.Second, they note that location of death was unknown. While certainly interesting to examine, there is no statistical or biological rationale to exclude based on, or adjust for, prehospital deaths since they represent a legitimate outcome for critically injured patients with poor access to care.Third, they incorrectly suggest that patients in the category of worst access to care (predicted ground transport time >15 minutes; n = 24) should be disregarded because "it seems illusory to carry out extensive statistics on such a small number of people…" All analyses were performed on the entire study population of 10 105 individuals with firearm injuries, first in a logistic regression model that considered access to care as a continuous variable, and second in a negative binomial model that considered each 1-minute increment of predicted access to care in a categorical variable. These approaches are robust to small numbers within categories.Fourth, they propose the subgroup of 24 patients with worst access to care might have experienced higher-lethality injuries "without any relation to geospatial access to care." By definition, only factors related to access to care (the exposure) can be confounders. Regardless, any occurrence affecting 0.2% of the study population would not impact our overall findings.Fifth, they refute a claim that we never made: speed is all that matters ("mortality would be lower if patients were initially cared for by racing drivers than by advanced life support teams in an ambulance"). We took great care to frame our findings as pertaining to geospatial access to care, not simply speed. However, as we discuss in the article, there are several