A multi-component algorithm achieved 92.6% sensitivity for identifying sport- and recreation-related deaths, substantially outperforming ICD-10 activity codes alone (52.0% sensitivity).
Observational (n=3,719)
Does a three-condition algorithm improve the identification of sport-related and recreation-related deaths compared to ICD-10 activity codes alone in coronial records?
A multi-component algorithm integrating activity codes, cause-of-death codes, and keywords significantly improves the detection of sport- and recreation-related deaths in coronial data compared to ICD-10 activity codes alone.
Absolute Event Rate: 92.6% vs 52%
Background Sport and recreation provide important health benefits but can occasionally result in fatal events, making accurate surveillance essential to estimate their burden and inform prevention strategies. In Québec, Canada, coronial data are commonly used for this purpose, but the performance of different identification approaches remains uncertain. Methods This retrospective observational study used the Bureau du coroner du Québec database to review all 2019 coronial records, including unintentional injury and natural deaths (n=3719). Three methodologies for identifying sport-related and recreation-related deaths were compared: International Classification of Diseases, 10th Revision (ICD-10) activity codes, a three-condition algorithm combining activity codes, underlying cause-of-death codes, and keywords, and a manual review serving as the reference standard. Sensitivity and positive predictive value (PPV) were calculated overall and by death type. Results The manual review identified 175 deaths, including 26 natural deaths and 149 unintentional injury deaths. ICD-10 activity codes identified 91 confirmed cases (52.0% sensitivity), with higher detection for natural deaths (76.9%) than for unintentional injury deaths (47.7%). The PPV of activity codes was 90.1%. The three-condition algorithm identified 162 confirmed cases, achieving 92.6% sensitivity and higher detection of unintentional injury deaths (95.3%), with a PPV of 85.7%. Discussion and conclusions ICD-10 activity codes alone substantially underestimated sport-related and recreation-related deaths in Québec, particularly for unintentional injuries. A multi-component algorithm integrating activity codes, underlying cause-of-death codes, and keywords markedly improves case detection while maintaining acceptable accuracy, supporting its use for routine surveillance. Consistent activity coding and detailed documentation by health records professionals remain essential, especially for identifying natural deaths.
Richard et al. (Tue,) conducted a observational in Sport-related and recreation-related deaths (n=3,719). Three-condition algorithm (activity codes, cause-of-death codes, and keywords) vs. ICD-10 activity codes alone was evaluated on Sensitivity for identifying sport-related and recreation-related deaths. A multi-component algorithm achieved 92.6% sensitivity for identifying sport- and recreation-related deaths, substantially outperforming ICD-10 activity codes alone (52.0% sensitivity).