BACKGROUND: A deeper understanding of how heat wave definition affects the relationship between heat exposure and health, especially as a function of rurality, will be useful in developing effective heat wave warning systems. OBJECTIVE: We compared the relationships between different heat wave index (HI) definitions and preterm birth (PTB) and nonaccidental death (NAD) across urban and rural areas. METHODS: We used a time-stratified case-crossover design to estimate associations of PTB and NAD with heat wave days (defined using 15 HIs) relative to non-heat wave control days in Alabama, USA (1990-2010). ZIP code-level HIs were derived using data from the North American Land Data Assimilation System. Associations with heat wave days defined using different HIs were compared by bootstrapping. We also examined interactions with rurality. RESULTS: Associations varied depending on the HI used to define heat wave days. Heat waves defined as having at least 2 consecutive days with mean daily temperatures above the 98th percentile were associated with 32.4% (95% CI: 3.7, 69.1%) higher PTB, and heat waves defined as at least 2 consecutive days with mean daily temperatures above the 90th percentile were associated with 3.7% (95% CI: 1.1, 6.3%) higher NAD. Results suggest that significant positive associations were more common when relative-compared with absolute-HIs were used to define exposure. Both positive and negative associations were found in each rurality stratum. However, all stratum-specific significant associations were positive, and NAD associations with heat waves were consistently positive in urban strata but not in middle or rural strata. CONCLUSIONS: Based on our findings, we conclude that a relative mean-temperature-only heat wave definition may be the most effective metric for heat wave warning systems in Alabama.