EVALUATION OF AN ARTIFICIAL INTELLIGENCE HEPARIN DOSING ALGORITHM TO MAINTAIN THERAPEUTIC ANTICOAGULATION DURING ATRIAL FIBRILLATION ABLATION

Abstract

BackgroundIntraprocedural anticoagulation during atrial fibrillation (AF) ablation reduces the risk of cerebral thromboembolism during ablation. Heparin dosing strategies vary, can result in sub-therapeutic levels, and require frequent physician decisions if not protocol driven.ObjectiveTo evaluate a heparin dosing algorithm and determine if it improves maintenance of guideline-based therapeutic activated clotting times (ACT) during AF ablation.MethodsDigital Intern software from iVMD includes a rule-based artificial intelligence algorithm for heparin dosing developed starting from published guidelines and adjusted to limit downward ACT drift and account for individual variation. The initial heparin dose is calculated from the patient’s weight, baseline ACT, and outpatient anticoagulant. Subsequent boluses and an infusion are adapted based on response to the first heparin bolus. Goal ACT was 300-350s. Outcomes from 50 cases prior to algorithm introduction were compared to 50 cases using the algorithm.ResultsProcedures using the algorithm reached goal ACT faster (17.6±9.4 min compared to 33.3±23.6 min pre-algorithm, p<0.0001), ACTs fell below goal less frequently while in the LA (11/50 cases compared to 27/50 cases with typical dosing, p=0.002), and rose above 400 s less frequently (2/50 cases compared to 10/50 cases with typical dosing, p=0.03). When the ACT did drop below goal, time below goal was less when using the algorithm (18.0±10.9 min compared to 29.3±15.5 min for typical dosing, p=0.03). System Usability Scale (SUS) scores were excellent (96±5, n=7, score > 80.3 considered excellent). Survey results demonstrated that the algorithm provided nurses greater autonomy and required infrequent physician interruptions.ConclusionOutcomes for Pre-Algorithm vs AI Algorithm Guided Heparin DosingOutcomesPre-Algorithm (n=50)Artificial Intelligence Algorithm (n=50)P-valueTime to ACT goal (> 300s)33.3 ± 23.6 min17.6 ± 9.4 min<0.0001# of patients with any ACT > 400s1020.03# of patients with any ACT < 300s while in LA27110.002Average time spent with ACT < 300s while in LA29.3 ± 15.5 min18.0 ± 10.9 min0.03(ACT - Activated Clotting Time) Open table in a new tab BackgroundIntraprocedural anticoagulation during atrial fibrillation (AF) ablation reduces the risk of cerebral thromboembolism during ablation. Heparin dosing strategies vary, can result in sub-therapeutic levels, and require frequent physician decisions if not protocol driven. Intraprocedural anticoagulation during atrial fibrillation (AF) ablation reduces the risk of cerebral thromboembolism during ablation. Heparin dosing strategies vary, can result in sub-therapeutic levels, and require frequent physician decisions if not protocol driven. ObjectiveTo evaluate a heparin dosing algorithm and determine if it improves maintenance of guideline-based therapeutic activated clotting times (ACT) during AF ablation. To evaluate a heparin dosing algorithm and determine if it improves maintenance of guideline-based therapeutic activated clotting times (ACT) during AF ablation. MethodsDigital Intern software from iVMD includes a rule-based artificial intelligence algorithm for heparin dosing developed starting from published guidelines and adjusted to limit downward ACT drift and account for individual variation. The initial heparin dose is calculated from the patient’s weight, baseline ACT, and outpatient anticoagulant. Subsequent boluses and an infusion are adapted based on response to the first heparin bolus. Goal ACT was 300-350s. Outcomes from 50 cases prior to algorithm introduction were compared to 50 cases using the algorithm. Digital Intern software from iVMD includes a rule-based artificial intelligence algorithm for heparin dosing developed starting from published guidelines and adjusted to limit downward ACT drift and account for individual variation. The initial heparin dose is calculated from the patient’s weight, baseline ACT, and outpatient anticoagulant. Subsequent boluses and an infusion are adapted based on response to the first heparin bolus. Goal ACT was 300-350s. Outcomes from 50 cases prior to algorithm introduction were compared to 50 cases using the algorithm. ResultsProcedures using the algorithm reached goal ACT faster (17.6±9.4 min compared to 33.3±23.6 min pre-algorithm, p<0.0001), ACTs fell below goal less frequently while in the LA (11/50 cases compared to 27/50 cases with typical dosing, p=0.002), and rose above 400 s less frequently (2/50 cases compared to 10/50 cases with typical dosing, p=0.03). When the ACT did drop below goal, time below goal was less when using the algorithm (18.0±10.9 min compared to 29.3±15.5 min for typical dosing, p=0.03). System Usability Scale (SUS) scores were excellent (96±5, n=7, score > 80.3 considered excellent). Survey results demonstrated that the algorithm provided nurses greater autonomy and required infrequent physician interruptions. Procedures using the algorithm reached goal ACT faster (17.6±9.4 min compared to 33.3±23.6 min pre-algorithm, p<0.0001), ACTs fell below goal less frequently while in the LA (11/50 cases compared to 27/50 cases with typical dosing, p=0.002), and rose above 400 s less frequently (2/50 cases compared to 10/50 cases with typical dosing, p=0.03). When the ACT did drop below goal, time below goal was less when using the algorithm (18.0±10.9 min compared to 29.3±15.5 min for typical dosing, p=0.03). System Usability Scale (SUS) scores were excellent (96±5, n=7, score > 80.3 considered excellent). Survey results demonstrated that the algorithm provided nurses greater autonomy and required infrequent physician interruptions. ConclusionOutcomes for Pre-Algorithm vs AI Algorithm Guided Heparin DosingOutcomesPre-Algorithm (n=50)Artificial Intelligence Algorithm (n=50)P-valueTime to ACT goal (> 300s)33.3 ± 23.6 min17.6 ± 9.4 min<0.0001# of patients with any ACT > 400s1020.03# of patients with any ACT < 300s while in LA27110.002Average time spent with ACT < 300s while in LA29.3 ± 15.5 min18.0 ± 10.9 min0.03(ACT - Activated Clotting Time) Open table in a new tab