Using a genetic algorithm to predict evaluation of acute coronary syndromes

Background: Because rapid therapy can improve the unfavorable prognosis of individuals with acute coronary syndromes (ACSs), it is critical that nurses accurately associate the cues of ACS and quickly and aggressively initiate interventional strategies that reduce mortality. Objectives: To determine if genetic algorithms (GAs) can be used to decipher the prediction rules that emergency department (ED) nurses use to triage persons suspected of having ACS and to determine whether these rules differ based on patient gender. Methods: A nonexperimental, descriptive study was conducted. Three thousand ED nurses were selected randolmly to receive a mailed clinical vignette questionnaire, and 840 questionnaires were returned. Data analysis included binary logistic regression (BLR), development of GA, sensitivity, specificity, receiver-operator characteristic curves, and Monte Carlo simulations. Results: Nurses use different prediction rules for triaging male and female vignette patients with possible ACS. Accuracies were similar between BLR and GA. Both suffered a loss of predictive accuracy when algorithms or equations developed on one sex were tested on the other sex. Monte Carlo simulations showed that similar cues were used in triaging both men and women but they were combined differently in producing GA. Discussion: Using GA was as accurate as results found by BLR and can be used to predict nurses triage deck sions for ACS. The GA presented as flow charts may be user-friendly.