Heat wave warning systems and related research define heat waves using various indices and there exists no standard definition for a heat wave. Despite various weather forecast services for heat stress in South Korea, it is unclear how different thermal indices affect the designation of heat waves and health effect estimates. We aimed to analyze trends of heat wave characteristics and mortality associations using various criteria for the warm season (June-September) in 2011-5 for the most populated two cities in South Korea, Seoul and Busan. Hourly weather monitoring data and daily mortality data in each city were obtained. The following indices were estimated to define heat waves: air temperature, heat index (HI), humidex, apparent temperature (AT), effective temperature (ET), and wet-bulb globe temperature (WBGT). The thresholds of each index for heat wave definitions were obtained by statistical distribution (95th percentiles) and minimum mortality temperature (MMT). Thermal indices showed clustering of accumulation of excess heat above thresholds for northeast regions in the cities while air temperature showed it for central regions. Compared to 95th percentiles, the MMTs resulted 14 times longer heat wave days for thermal indices except for air temperature. When MMTs were used, nine times larger excess mortality from heat waves occurred for all indices compared to that from heat waves defined by the 95th percentiles. The thermal indices with the highest association between heat and mortality varied between the two cities: air temperature for Seoul and WBGT for Busan. Heat wave warnings should be based on a thorough comparison of how different heat wave criteria will affect the public health impact of heat wave warnings in terms of identifying a heat wave and degree of health impacts due to it.