Projecting the Burden of Injury Related to Tropical Cyclone Impacts in Florida, 2020-2040

Background Florida is vulnerable to the impacts of tropical cyclones (TCs) and experiences more landfalls than any other state. The storm surge and high winds associated with TCs contribute to a variety of direct and indirect effects on human health. How these impacts on health may change in the future is a key question for public health practitioners and emergency responders. Therefore, understanding projected TC and associated health impacts is important for effective public health adaptation and preparedness planning. Two important public health problems associated with TC impacts in Florida are injury and carbon monoxide (CO) poisoning. We used available data sources to project the future burden of disease associated with TC impacts among Florida residents out to 2040 in ten-year increments using an attributable risk model. Methods To effectively project the burden of these health outcomes, several inputs are needed including baseline rate of disease, exposure-response functions between the health outcome and hazard, future population projections, and projected changes in hazard impacts. Baseline rates of the diseases of interest were based on hospital (2004-2012) and emergency department (ED; 2005-2012, due to data availability) billing records. Primary and secondary diagnosis codes were used to define visits related to all-cause injury and CO poisoning. Poison control center calls related to CO exposure (2004-2012) were also assessed separately from hospital billing records. Exposure-response functions were assessed using a matched cohort study design with the unit of analysis defined as daily counts of visits or poison center calls in a specific county. Exposure was defined as counties experiencing tropical storm-force (39-74 MPH) or hurricane-force (≥74 MPH) winds or storm surge greater than 1.2m. The Southeast Regional Climate Center provided maximum wind speeds observed in all Florida counties for each TC. Storm surge data for the Gulf Coast was obtained from the Louisiana State University SURGEDAT database. Impact periods were outcome-specific. For injury, we were interested in an impact period to include pre-storm preparation and post-storm clean-up activities (-2 to 14 days). For CO poisoning, the time period of interest included the days of impact and post-impact (0 to 14 days). Impact periods were matched to two control periods, both pre- and post-hurricane season, by county and system. Exposure-response functions were analyzed for combined TC impacts and separately for hurricane impacts using Poisson regression models. Rate ratios (RR) and 95% confidence intervals (CI) were calculated. Population projections for Florida by county and year were obtained from the Florida Bureau of Economic and Business Research. We used the total population projections for Florida for 2020 to 2040 in ten-year increments. Because of the interannual variability and the complexity of factors associated with TC formation, there is wide variation in TC projections for the 21st century. The general consensus, however, is a tendency toward decreasing frequency, increasing intensity, and increasing precipitation. Therefore, we present projections for both combined TC impacts and hurricane only impacts. We used an attributable risk model to project disease burden for our health outcomes of interest. RRs and associated 95% CIs were converted to attributable fractions (AF) in order to obtain the estimated number of events that may occur due to TC or hurricane impacts above baseline. Results From 2004 to 2012, eight hurricanes and eleven tropical storms made landfall in or impacted all of Florida's 67 counties in some manner. AFs varied by health outcome and exposure type (Table). For all-cause injury, AFs ranged from about 3.5% for TC impacts to 19.7% for hurricane impacts. For CO poisoning, AFs were higher and ranged from 70.9-84.8% for TC impacts and from 90.4-93.3% for hurricane impacts. The estimated number of events above baseline for each health outcome and event type are presented in the table below. The range is based on the upper and lower 95% confidence limits from the exposure-response functions. Conclusions CO poisoning had the highest AFs associated with TC and hurricane impacts. By 2040, hospital visits attributable to TCs range from 101 to 126, while ED visits range from 225 to 359. These numbers increase when examining hurricanes only and are also higher for ED visits and poison center exposure calls. Even though the proportion of events attributable to TCs and hurricanes were much lower for all-cause injury than CO poisoning, the public health impacts are greater given the higher baseline rates of injury in the population. Further, the majority of deaths in Florida during post-impact periods are due to unintentional injury. By 2040, the estimated number of hospital visits for injury above baseline attributable to TC impacts range from about 4,000 to over 10,000 and are even higher for hurricane impacts. Some limitations to this method must be noted. Our attributable risk models assume that there are no changes in baseline rates over time and no additional hurricane adaptation measures are implemented during this period. Further, we did not include direct estimates of projected changes in TC impacts, but indirectly estimated projected increases in intensity by focusing on hurricane impacts only. Additional work is on-going to further improve the methods used to project TC-related disease burden among Floridians.