Probabilistic broken-stick model: A regression algorithm for irregularly sampled data with application to eGFR

We propose a new regression algorithm known as the probabilistic broken-stick model. Using a set of locally linear line segments, i.e., the 'broken sticks', it can model any complex, non-linear function. Therefore, the model can balance both short term interpretability and long term flexibility simultaneously. It is parametric and completely generative, providing rate change as additional output. Furthermore, it can seamlessly handle any irregularly sampled clinical time series. In this paper, we show how the broken-stick model can be applied to modelling estimated glomerular filtration rate (eGFR) https://youtu.be/nS1X5OEulDY.176