Heat waves are the product of what is usually referred to as subseasonal atmospheric variability. Typically the search for sources of subseasonal midlatitude predictability has primarily focused on atmospheric phenomena driven by tropical convection because midlatitude dynamics are highly chaotic. Here, based on a 12,000-year integration of an atmospheric general circulation model, we present that US heat waves tend to be preceded by an anomalous atmospheric zonal wavenumber-5 planetary wave pattern by 15-20 days. This circulation pattern resembles the observed leading pattern of subseasonal variability and does not rely on tropical heating. When it is present with high amplitude in the initial state of a prediction, one may anticipate an increased probability of a subsequent US heat wave on subseasonal time scales. Given the high impacts of heat waves, even if only a modest portion of midlatitude circulation anomalies that influence heat wave likelihood are predictable beyond the typical weather forecast range, there would be the potential for substantial benefits to society.