Prediction of intradialytic morbid events in hemodialysis patients by monitoring the second derivative of relative blood volume

Intradialytic morbid events (IME) are a major issue affecting hemodialysis (HD) patients. Change in relative blood volume (RBV) reflects change in plasma refilling rate (PRR) and variability of ultrafiltration rate (UFR). However, degree of fluid overload influences this relationship. We propose a method using a second derivative of RBV (SDRBV) model to detect decrease in PRR during HD which is a major factor associated with IME. Forty-five HD patients (age 55.7±14 years, height 167.7±10 cm, Pre HD weight 79.3±16.6 kg) were studied with a total of 245 treatments, with IME occurring in 124 treatments. Predictions of IME with the SDRBV model showed 92 % sensitivity for 124 IME and 43 % specificity of 121 non IME treatments respectively. The average time between prediction and IMEs was -30.8±36 minutes. A high positive predictive value was due to 1) using a low-pass filter to reduce possible interferences in the RBV curve and 2) use of 5 minutes sample frequency to apply the SDRBV model. These results indicate that the SDRBV model is a useful tool in clinical practice to predict and prevent IME.