Impact of prolonged elevated heart rate on sepsis-associated acute kidney injury patients: a causal inference and prediction study.

Abstract

Background: This study investigates the relationship between prolonged elevated heart rate (pe-HR) and mortality among patients with sepsis-associated acute kidney injury (S-AKI) using causal inference and machine learning.

Methods: Pe-HR was defined as a heart rate exceeding 100 beats/min for 11 hours within a 12-hour interval. The average treatment effect (ATE) was a measure of the therapeutic effect of β-blockers. The primary outcome measure of the study was the 90-day survival and hospitalization survival as the secondary endpoint. Four machine learning algorithms were employed to assess whether an increase in pe-HR enhanced the performance of predictions.

Results: The study included a total of 14,388 patients with S-AKI from MIMIC (Medical Information Mart for Intensive Care)-IV. The results revealed that both 90-day and in-hospital survival were lower in patients with pe-HR than in those without (hazard ratio: 1.62 for 90-day survival and 1.22 for in-hospital survival). The ATE of development pe-HR was 56.3 days survival higher than without pe-HR. The ATE of β-blocker use was 40.2 days survival higher than no use. Four machine learning algorithms showed that the inclusion of pe-HR improved the accuracy of 90-day survival predictions (p < 0.05), with the best area under the receiver operating characteristic curve reaching 0.76.

Conclusion: In S-AKI patients, pe-HR was significantly associated with lower survival rates. Including pe-HR in prediction models improved their accuracy for 90-day survival. Our results suggest a causal relationship, highlighting the potential therapeutic benefit of β-blockers in managing S-AKI patients.