Evaluation of Tornado-Related Mortality Tracking Using the Electronic Death Registration System – Oklahoma, May 2013

Wednesday, June 7, 2017: 11:20 AM
430A, Boise Centre
Anindita Issa , Centers for Disease Control and Prevention, Atlanta, GA
Rebecca Noe , Centers for Disease Control and Prevention, Atlanta, GA
Royal Law , Centers for Disease Control and Prevention, Atlanta, GA
Derek Pate , Oklahoma State Department of Health, Oklahoma City, OK
Kelly Baker , Oklahoma State Department of Health, Oklahoma City, OK

BACKGROUND: The majority of U.S. states use electronic death registration systems (EDRS). A few have leveraged their EDRS for disaster-related mortality surveillance by creating flags or notifications. Our objectives were to evaluate 1) Oklahoma’s EDRS flagging process and tornado-related terminology used during the May 2013 tornadoes and 2) the accuracy of ICD-10 coding of these deaths.

METHODS:  During our site visit to Oklahoma in September 2016, we coordinated stakeholder discussions and reviewed EDRS data (including event flags) using Oklahoma’s tornado-related death case definition.

RESULTS:  The Oklahoma EDRS flagging process had a sensitivity of 89.6% (43/48) and a positive predictive value of 100% (43/43). The median time from date of death to date of certification was 3.5 (range: 1-21) days. Seventy-five percent of tornado-related deaths (36/48), contained the term “tornado” in the final death certificate data file, designating the death was attributed to the tornadoes. Of these 36 cases, 72.2% (26/36) were accurately coded with the ICD-10 code X37 (Cataclysmic Storm).

CONCLUSIONS:  In addition to providing real-time disaster mortality surveillance, Oklahoma’s EDRS flag can ensure that death data files receive proper disaster-related terminology and ICD-10 coding. The majority of tornado-related deaths were flagged as they were entered into Oklahoma’s EDRS. Of note, over a quarter of the deaths attributed to the tornadoes was not coded with X37 despite the presence of a “tornado” term. Coding errors can lead to underreporting of tornado-related deaths at state and national levels and can affect tornado prevention research.