International Journal of Medical and Health Research

Authors
Peter D Hart

Abstract
Background: The purpose of this study was to use a novel data mining technique to identify predictors of hospital admission in adults injured during an exercise-related (ER) activity. Methods: Data for this research came from the 2015 National Electronic Injury Surveillance System (NEISS) which collects data annually from a representative sample of U.S. emergency departments (EDs). Product codes were used to identify adults 18+ years of age who were injured in an ER activity. Variables utilized in the analysis were hospital admission (yes/no), body part (upper/lower), location (recreation or sport facility/other), age group (18-24/25-49/50-64/65+ years), race (white/black/other), and sex (male/female). SAS survey procedures and SPSS CHAID were used for the analyses. Results: An estimated 16,958 (5.4%) out of 311,563 adults were admitted in 2015 after presenting to an ED with upper or lower body ER injury. Multiple logistic regression showed body part, age group, and race as independent predictors of admission. CHAID analysis with 95.3% accuracy showed that the first best predictor of admission was age group. Among the 65+ age group, race and then body part were significant predictors. The three younger age groups showed a similar pattern with body part then age group and location significantly predicting admission. Conclusion: Results from this study support the use of a novel data mining tool to find specialized subgroups of ER injuries predictive of hospital admission.