Many studies have examined the effect of heat waves on various health outcomes. But few have explored the fine-scale spatial variability over a wide domain while considering the spatial structure of the data. In this study, we compared a few approaches for this problem proposing a novel spatiotemporal method. We propose a spatial extension of the case-crossover design where we match heat wave to similar non-heat wave days for comparison of hospitalization rates in small spatial regions on both absolute and relative scales. We extend to use Bayesian hierarchical models to leverage spatial correlation. For comparison, we apply the Kulldorff method of cluster detection across the state. Our approach detects regions both in urban and rural areas of California, with precision in urban areas. The Bayesian hierarchical model extension increases precision in rural areas. This approach is able to use spatial data to get local estimates all across California that can help guide local level policy making and statewide efforts to reduce heat wave health impacts. This methodology can be used in other regions and with other environmental exposures to help guide policy in a changing climate.