Martin Gellerstedt1, Nina Rawshani2, Johan Herlitz3, Angela Bång4, Carita Gelang5, Jan-Otto Andersson6, Anna Larsson4, Araz Rawshani7. 1. University West, School of Business, Economics and IT, Trollhättan, Sweden. Electronic address: martin.gellerstedt@hv.se. 2. Department of Emergency Medicine, University of Gothenburg, Göteborg, Sweden. 3. The Pre-hospital Research Centre of Western Sweden, Prehospen, University College of Borås, Borås, Sweden; Department of Medicine, University of Gothenburg, Göteborg, Sweden. 4. The Pre-hospital Research Centre of Western Sweden, Prehospen, University College of Borås, Borås, Sweden. 5. The Pre-hospital Research Centre of Western Sweden, Prehospen, University College of Borås, Borås, Sweden; Department of Ambulance and Prehospital Emergency Care, University of Gothenburg, Göteborg, Sweden. 6. Department of Ambulance and Prehospital Emergency Care, Skaraborg, Sweden. 7. Department of Medicine, University of Gothenburg, Göteborg, Sweden.
Abstract
BACKGROUND: To evaluate whether a computer-based decision support system could improve the allocation of patients with acute coronary syndrome (ACS) or a life-threatening condition (LTC). We hypothesised that a system of this kind would improve sensitivity without compromising specificity. METHODS: A total of 2285 consecutive patients who dialed 112 due to chest pain were asked 10 specific questions and a prediction model was constructed based on the answers. We compared the sensitivity of the dispatchers' decisions with that of the model-based decision support model. RESULTS: A total of 2048 patients answered all 10 questions. Among the 235 patients with ACS, 194 were allocated the highest prioritisation by dispatchers (sensitivity 82.6%) and 41 patients were given a lower prioritisation (17.4% false negatives). The allocation suggested by the model used the highest prioritisation in 212 of the patients with ACS (sensitivity of 90.2%), while 23 patients were underprioritised (9.8% false negatives). The results were similar when the two systems were compared with regard to LTC and 30-day mortality. This indicates that computer-based decision support could be used either for increasing sensitivity or for saving resources. Three questions proved to be most important in terms of predicting ACS/LTC, [1] the intensity of pain, [2] the localisation of pain and [3] a history of ACS. CONCLUSION: Among patients with acute chest pain, computer-based decision support with a model based on a few fundamental questions could improve sensitivity and reduce the number of cases with the highest prioritisation without endangering the patients.
BACKGROUND: To evaluate whether a computer-based decision support system could improve the allocation of patients with acute coronary syndrome (ACS) or a life-threatening condition (LTC). We hypothesised that a system of this kind would improve sensitivity without compromising specificity. METHODS: A total of 2285 consecutive patients who dialed 112 due to chest pain were asked 10 specific questions and a prediction model was constructed based on the answers. We compared the sensitivity of the dispatchers' decisions with that of the model-based decision support model. RESULTS: A total of 2048 patients answered all 10 questions. Among the 235 patients with ACS, 194 were allocated the highest prioritisation by dispatchers (sensitivity 82.6%) and 41 patients were given a lower prioritisation (17.4% false negatives). The allocation suggested by the model used the highest prioritisation in 212 of the patients with ACS (sensitivity of 90.2%), while 23 patients were underprioritised (9.8% false negatives). The results were similar when the two systems were compared with regard to LTC and 30-day mortality. This indicates that computer-based decision support could be used either for increasing sensitivity or for saving resources. Three questions proved to be most important in terms of predicting ACS/LTC, [1] the intensity of pain, [2] the localisation of pain and [3] a history of ACS. CONCLUSION: Among patients with acute chest pain, computer-based decision support with a model based on a few fundamental questions could improve sensitivity and reduce the number of cases with the highest prioritisation without endangering the patients.
Authors: Charles Richard Knoery; Janet Heaton; Rob Polson; Raymond Bond; Aleeha Iftikhar; Khaled Rjoob; Victoria McGilligan; Aaron Peace; Stephen James Leslie Journal: Crit Pathw Cardiol Date: 2020-09