Why low-altitude aviation accidents become fatal

Abstract

Low-altitude aviation accidents can become fatal when pilots have little time or altitude to recover after an abnormal event. Fatal outcomes, however, are rarely explained by altitude alone: they often depend on how human, aircraft, environmental, and operational factors combine within the accident sequence. This study asks whether the structure of these documented factor combinations helps explain injury severity in U.S. low-altitude aviation accidents. The data comprise 11,639 low-altitude accidents with completed Findings investigations, drawn from the National Transportation Safety Board (NTSB) Aviation Accident/Incident Data System (AVALL) for 2013–2024; throughout, “Findings” denotes the NTSB’s investigator-coded causal and contributing factors. The method represents Findings as a co-occurrence network that links factor types documented in the same investigation. The global network is constructed from a broader base of 23,847 completed-investigation accidents and is used only to define the co-occurrence structure; it is not the analytical sample. From this network, five accident-level measures describing the breadth and structure of each accident’s documented failure pattern are added to a 43-covariate multinomial logit severity model. The five measures significantly improve model fit (LR(15) = 236.2, p < 10^-6[jls-end-space/]). The strongest fatal-severity associations are found for factor diversity, the number of distinct factor types involved (odds ratio OR = 1.46 per standard deviation, 95 % CI 1.31–1.64), and multi-source spread, the extent to which those factors span broader co-occurrence groups (OR = 1.37, 95 % CI 1.25–1.51). The multi-source-spread association is strong for fixed-wing aircraft (OR = 1.66) but weaker and not statistically significant for helicopters and agricultural Part 137 operations. We interpret these results as evidence that, among completed investigations of conventional low-altitude aviation accidents, fatal outcomes are associated with broader documented failure configurations, whereas some specialized operations may follow narrower, operation-specific fatality pathways. Because Findings are coded after investigation, the framework is a retrospective safety-review tool rather than a real-time warning or causal-prediction method.