Predicting hemodialysis mortality utilizing blood pressure trends

Area of Science:

• Nephrology
• Cardiovascular Medicine
• Biostatistics

Background:

• Mean Systolic Blood Pressure (SBP) is a key mortality predictor in hemodialysis (HD) patients.
• Current prediction models may not fully capture the dynamic nature of SBP.

Purpose of the Study:

• To investigate if transforming SBP measurements to reflect trends improves mortality prediction in HD patients.
• To compare the predictive performance of SBP trend models against a baseline model using mean SBP.

Main Methods:

• Utilized data from 4,500 US-based HD patients with at least six months of follow-up.
• Developed six transformed SBP trend variables using Relative Difference in Percentage.
• Employed Support Vector Machine (SVM) models, incorporating demographic and clinical factors, and a generalized person-month approach for pooled data.

Main Results:

• The model incorporating SBP trends achieved an Area Under the Curve (AUC) of 0.70 on unseen data, significantly outperforming the baseline model (AUC 0.63, p<0.00001).
• A pooled data model also demonstrated strong predictive performance with an AUC of 0.69.

Conclusions:

• Incorporating SBP trends into predictive models substantially enhances mortality risk assessment for hemodialysis patients.
• This methodology offers a more nuanced understanding of SBP's impact on patient outcomes.