Bad Altitude: Categorizing Elevation Produces Spurious Association With Concussions in the National Football League (NFL).

To assess whether prior analyses, where there was a relationship between altitude and concussion rates in American football, would replicate using a larger data set and altitude as a continuous variable. Cohort study replication. We analyzed data from all NFL regular season games from 2012-2019. Concussions were identified from public databases and NFL injury reports. The altitude of each stadium was identified using mapping software. Concussion rates were calculated for each stadium and plotted against continuous altitude. We calculated crude rate ratios for several categorical cut points and used logistic and Poisson regression models to assess associations with continuous altitudes. We identified 867 players (1103 player seasons) who sustained 1159 concussions during the time period 2012-2019. All continuous plots and models showed no evidence of any association between concussions and altitude. A Poisson model found an IRR of 1.00 (95% CI, 0.99-1.01) for every 100-ft increase in altitude. A 644-ft cut point (used in previous studies) produced a significant difference (incidence rate ratio [IRR], 0.71; 95% CI, 0.54-0.94) in 2012-2013, but this did not replicate in 2014-2019 (IRR, 0.99; 95% CI, 0.84-1.14). We found no association between altitude and concussion rates in the NFL when altitude was analyzed continuously rather than inappropriately categorized. Our findings should increase skepticism of any effect of altitude on concussions at the elevations at which most American football is played, as well as clinical interventions based on that theory. It also underscores the importance of keeping continuous variables continuous wherever possible. .