Heat-Related Emergency Department Visits In Relation To Measures Of Potential Heat Stress In New Jersey

BACKGROUND:  The purposes of this project were to: 1) determine whether real-time syndromic surveillance data using patient’s emergency room (ER) chief complaint can identify episodes of heat-related illness (HRI) in New Jersey; and 2) examine the relationship between counts of HRI (from patient diagnosis data) and maximum daily heat index (HI), and whether this relationship changes by region of the state.

METHODS:  In the first phase, we characterized the relationship between a patient’s initial ER chief complaint potentially related to HRI with the patient’s final coded primary and secondary diagnoses, in the May-September periods of 2009-2011.  An algorithm for identifying HRI from chief complaint was developed based on keyword classifications developed by New York City.  ICD-9 diagnostic codes for HRI included 992.0-992.9 and external cause of injury codes E900.0 and E900.9.  In the second phase, we examined HRI diagnosis data in relation to daily maximum HI during May-September of 2010-2011.  We computed relative risks of HRI at various cut-points of maximum daily HI by region of the state.

RESULTS:  In the May-September periods in 2009-2011, there were 871 HRI cases identified from chief complaints among ERs participating in the syndromic surveillance system.  In the same ERs over that time, there were 2,146 HRI cases determined from ICD-9 codes.  The sensitivity of the HRI definition using chief complaint in comparison to diagnostic coding was 16% (348/2,146) and the positive predictive value was 40% (348/871); these measures improved during a major heat wave.  The temporal pattern of HRI from chief complaint tracked well with HRI from diagnostic codes.  Non-concordant cases typically involved non-specific keywords or codes.  There were 3,479 HRI diagnoses state-wide in the May-September periods of 2010-2011.  Compared to days with the maximum HI < 85, the RRs of HRI were: 16.7 for HI > 95, 28.6 for HI > 100; and 41 for HI > 105.  Similar patterns were seen across regions.

CONCLUSIONS:  Though chief complaint data to monitor HRI was relatively insensitive, all major episodes of HRI were identified, suggesting that syndromic data reveals the temporal pattern of HRI diagnosis.  Non-concordant cases did not suggest ways to improve sensitivity of the algorithm.  Risk of HRI increased smoothly with increasing HI.  Using these data, triggers for forecast warnings may be considered in relation to local HRI and meteorological data.  Real-time HRI surveillance provides an opportunity to supplement forecast warnings with actual impact data to reinforce protective public health messages during a prolonged heat wave.