Night of the sirens: analysis of carbon monoxide-detector experience in suburban Chicago

To determine the pattern and environmental causes of carbon monoxide (CO)-detector alarms.

Data including time, location, detector manufacturer, CO measurements in the home, reported illness, cause, and actions taken were collected between July 15, 1994, and January 26, 1995, on all calls to 17 suburban Chicago fire departments for CO-detector alarms. We used univariate time-series analysis involving joint estimation of model parameters and outlier effects to analyze data and compared data on ambient CO levels from the Illinois Environmental Protection Agency to the number of calls per day.

During the study period, 777 calls for sounding CO detectors were made to the fire departments in question. The median number of calls per day was three. Our univariate time series identified 3 days with a significant excess of calls (December 12, 29 calls; December 21, 69; December 22, 128; P < .001). The average ambient CO readings on these days were 0.99, 3.25, and 3.89 ppm, respectively, compared with an overall mean of 8.8 ppm. In-home CO levels among all 828 measurements taken from the 777 domestic calls ranged from 0 to 425 ppm, 0 in 249 (30%), 1 to 10 in 340 (41%), 11 to 50 in 149 (18%), 51 to 100 in 22 (9%), and more than 100 in 11 (1.3%). No measurement was taken in six cases. Cause of alarm was listed as furnace in 25 cases, auto exhaust in 24, stove/oven in 22, poor location of detector in 14, water heater in 11, outside sources in 7, and multiple sources in 7. Other sources accounted for fewer than 1% each. The participating fire departments considered 242 cases (31%) to be false alarms. Cause was not determined in 400 calls (51%). In 37 calls (4.8%), people reported illness.

Above-average ambient CO levels coincided with a significant increase in the number of calls and may have contributed to the triggering of CO alarms.