For a study, researchers sought to create a cardiac arrest prediction model utilizing deep learning (CAPD) and to test the system by assessing how the prognosis of out-of-hospital cardiac arrest patients changes as scene time intervals (STI) rise.
Using information from the National Emergency Center’s smart advanced life support experiment, which was conducted from January 2016 to December 2019, they performed a retrospective cohort research. The datasets for derivation and validation were randomly selected from the smart advanced life support data. When predicting a patient’s prognosis, the CAPD model’s performance was compared to that of traditional machine learning techniques using the patient’s age, sex, event witness, bystander cardiopulmonary resuscitation (CPR), administration of epinephrine, initial shockable rhythm, prehospital defibrillation, provision of advanced life support, response time interval, & STI as prediction variables. After correcting other input data values, alterations in the patient’s prognosis about the rise in STI were noticed.
The research comprised 16,992 patients in total. For predicting the prehospital return of spontaneous circulation (ROSC) and positive neurological outcomes, the area under the receiver operating characteristic curve values were 0.828 (95% CI 0.826-0.830) and 0.907 (0.914-0.910), respectively. The program performed substantially better than other artificial intelligence algorithms and traditional techniques. When the STI reached 28 minutes, it was expected that the neurological recovery rate would drop to 1/3 of what it was at the start of cardiopulmonary resuscitation and that the prehospital ROSC would fall to 1/2 of its starting level when the STI was 30 minutes.
The CAPD demonstrated potential and efficacy in identifying patients for prehospital resuscitation who have ROSC and good neurological outcomes.