The four major hurricanes that struck Florida during August 13–September 25, 2004, produced electric power outages in several million homes. After the hurricanes, the Consumer Product Safety Commission (CPSC) investigated six deaths in Florida attributed to carbon monoxide (CO) poisoning (CPSC, unpublished data, 2004). The Florida Department of Health and CDC analyzed demographic and CO exposure data from these fatal poisoning cases and from nonfatal poisoning cases among 167 persons treated at 10 hospitals, including two with hyperbaric oxygen (HBO2) chambers. This report describes the results of that analysis, which determined that misplacement of portable, gasoline-powered generators (e.g., indoors, in garages, or outdoors near windows) was responsible for nearly all of these CO exposures. Public health practitioners should recognize that post-hurricane environments present challenges to the safe operation of portable generators and should educate the public on the hazards of CO poisoning in these settings.

All medical records were reviewed from participating hospitals in which a patient received a diagnosis of unintentional CO poisoning (International Classification of Diseases, Ninth Revision code 986) during August 13–October 15, 2004. These dates correspond to landfall of the first hurricane (Charley) and 3 weeks after landfall of the last hurricane (Jeanne), when active surveillance for CO poisoning was discontinued. Nine participating hospitals, including one with an HBO2 chamber, were located in landfall counties and involved in post-hurricane surveillance; a tenth participating hospital, which also had an HBO2 chamber, was located in central Florida. Any case involving a diagnosis of unintentional CO poisoning not related to a fire was included. All available information about the patient’s exposure, clinical presentation, laboratory testing (e.g., result of earliest available measurement of blood carboxyhemoglobin [COHb] level), and medical treatment was collected. In addition, investigations into six deaths from five exposure incidents were reviewed for basic demographic information and details about generator location. Because the six persons who were fatally poisoned died before arrival at a medical facility, no clinical information was recorded for them.

A total of 167 persons had nonfatal CO poisoning diagnosed during the study period, representing a total of 51 exposure incidents. The number of cases and incidents peaked within 3 days after landfall of each hurricane (Figure 1). The mean number of persons poisoned per incident was 3.3 (range: one to eight persons per incident). Fifty-four (32.3%) patients were initially treated at emergency departments (EDs) in hospitals outside the surveillance system but were later transferred to one of the two hospitals with HBO2 chambers.

Of the 167 persons with nonfatal poisoning, 87 (52.1%) were female. The median age was 29 years; 52 (31.1%) were aged <16 years, and 11 (6.6%) were aged >65 years. Seventy-six (45.5%) of the persons with nonfatal poisoning were white,* 47 (28.1%) Hispanic, 36 (21.6%) black, and six (3.6%) Asian; the race/ethnicity of two (1.2%) persons was not known. The percentages of those poisoned who were Hispanic and black were approximately twice the percentages of Hispanics (14.7%) and blacks (9.1%) reported residing in the hurricane-affected counties by the Florida 2004 Behavioral Risk Factor Surveillance System (BRFSS) survey. Among the six persons who were fatally poisoned, all were white, and five (83.3%) were male; the median age was 45 years (range: 30–58 years).

The most frequently reported symptoms of CO poisoning were headache (80.0%), nausea (51.5%), dizziness (50.9%), vomiting (31.5%), shortness of breath (16.4%), and loss of consciousness (14.5%) (Table). Among the 162 patients for whom COHb levels were available, the mean level of COHb was 19.8% (standard deviation: +8.7%); median was 21.1% (range: 0.2%–45.1%). Eighty-one (48.5%) patients were treated and released from the ED without HBO2 treatment. Seventy-three (43.7%) patients were released after HBO2 treatment. Thirteen (7.8%) patients were hospitalized; 11 of those were discharged after one night. Of the 13 hospitalized patients, four received HBO2 treatment. Overall, 77 (46.1%) persons were treated with HBO2.

The majority of nonfatal poisonings occurred overnight, with patients waking in the early morning with symptoms (Figure 2). One hundred eleven (66.5%) patients arrived at the ED during 5:00 a.m.–10:00 a.m. Medical records indicated that patients typically used generators to power refrigerators, fans, and air conditioners while sleeping. Similar exposure patterns and types of powered appliances were reported among the five incidents with fatalities.

Information regarding the source of CO was available for 49 (96.1%) of the 51 incidents with nonfatal poisonings. Use of portable, gasoline-powered generators was implicated in 47 (96.0%) nonfatal incidents and in the five incidents that resulted in the six fatalities. In two other nonfatal incidents, exposure to CO was attributed to use of a gasoline-powered saw and to a vehicle left idling in a garage. In the 47 nonfatal incidents in which a generator was known to be involved, 16 (34.0%) generators were operated outdoors; 16 (34.0%) inside a garage; six (12.8%) inside a home; four (8.5%) on an attached porch, deck, or patio; one (2.1%) inside a business; and one (2.1%) as part of a recreational vehicle. Generator location was unavailable for three (6.4%) incidents. The majority of the 16 generators placed outdoors were reportedly located near windows or window-mounted air conditioners. Medical records for certain patients indicated that generators were placed in homes or garages to protect the devices from the weather or to prevent them from being stolen. Among the five incidents with fatalities, generators were placed inside a home in two incidents, in an office or business in two incidents, and inside a garage in one incident. No mention was made of a home CO detector in any of the medical records.