Just did a google search here an interesting read:
Editorial Note: In Florida, most deaths directly attributed to Hurricane Andrew resulted from blunt trauma or asphyxia. In the past, hurricane-associated mortality has included high numbers of drownings (1). However, because of the minimal storm surge in the heavily populated areas, a building code that requires structures to withstand winds of 130 mph, and advanced warning systems and well-coordinated evacuation plans, drowning, as well as deaths from other causes, attributed to the hurricane remained relatively low. Nonetheless, some persons refused to evacuate their homes, and the deaths directly associated with Hurricane Andrew in Florida occurred among such persons. Public health and emergency management agencies need to continue to emphasize the importance of timely evacuation during natural disasters.
Most of the deaths associated with Hurricane Hugo, which struck Puerto Rico and South Carolina in September 1989, occurred in the postimpact phase and included deaths from electrocutions due to contact with energized power lines and the use of portable electric generators; house fires caused by candles used for lighting; and traumatic injuries sustained during clean-up (2,3). Two deaths associated with fires have already occurred in the postimpact phase of Hurricane Andrew. The public and relief workers should be aware of these and other potential dangers during the postimpact, clean-up phase of the hurricane and take appropriate precautions.
Because there is no universally accepted definition of a hurricane-associated death, the cases listed in this report were determined to be hurricane-associated by each district ME office. As a result, each ME office may apply different criteria to determine whether a death resulted from the hurricane. In addition, other organizations that collect information on disaster-associated deaths might apply different criteria. These potential differences reinforce the need to develop a standard definition for disaster-associated deaths.
Since 1989, MEs and coroners have reported important information about deaths associated with several natural disasters, including Hurricane Hugo (1989) (2,3), the Loma Prieta earthquake (1989) (4), the Plainfield tornado in Illinois (1990) (5), and flash floods in Texas (1991). In addition to mortality surveillance, the State of Florida, with the assistance of CDC, is conducting a rapid damage assessment survey to determine the extent of injuries and loss of utilities and health services that resulted from Hurricane Andrew.
http://www.cdc.gov/mmwr/preview/mmwrhtml/00017499.htm
Here is my google search too if you want to browse:
http://www.google.co.uk/search?q=why+most+deaths+occurred+hurricane&hl=en&lr=&start=10&sa=N