Non-linear association between AKI alert detection rate by physicians and medical costs

Background: Acute kidney injury (AKI) is associated with high mortality rates and long-term adverse outcomes and significantly increases medical costs. The AKI electronic alert system built the AKI diagnostic algorithm into the medical system, along with automated collection of key indications and generation of alerts. However, the relationship between the AKI electronic alert system and medical costs is still unknown. Methods: An exploratory secondary analysis of data from a double-blinded, multicenter, parallel, randomized controlled trial to investigate the association between the AKI electronic alert system and medical costs. Results: Finally, a total of 6030 patients were enrolled in this study. Multivariate logistic regression analysis revealed that the alert group was not significantly associated with medical costs (all p-values > 0.05). However, the rate of alert detection by an attending physician demonstrated a notable negative correlation with medical costs; adjusted effects for direct and total costs were −126.78$ and −236.82$, respectively. The curve fitting and threshold effect analysis revealed that when the rate of alert detection by an attending physician was between 18% and 59%, each unit increase in the rate corresponded to decreases in direct cost by 363.94 (−463.34, −264.55) $ and in total cost by 698.93 (−885.78, −512.07) $. Our subgroup analysis also found a significant relationship between the rate and medical costs. Conclusion: The alert group did not significantly reduce medical costs compared to the usual care group. However, the rate of alert detection by an attending physician had a significant negative association with medical costs, and there was a threshold effect between them. When the rate was between 18% and 59%, medical costs decreased as the rate increased, and when the rate was