Balancing Interpretability and Flexibility in Modeling Diagnostic Trajectories with an Embedded Neural Hawkes Process Model
This work addresses the need for interpretable yet flexible models in healthcare for analyzing event sequences like EHRs, offering an incremental improvement over existing neural network approaches.
The authors tackled the trade-off between interpretability and flexibility in modeling diagnostic trajectories using Hawkes processes, proposing an embedded neural Hawkes process model that balances these aspects. Results showed accurate impact function recovery in simulations, competitive performance on MIMIC-IV, and clinically meaningful interpretations on Duke-EHR, maintaining interpretability without performance loss.
The Hawkes process (HP) is commonly used to model event sequences with self-reinforcing dynamics, including electronic health records (EHRs). Traditional HPs capture self-reinforcement via parametric impact functions that can be inspected to understand how each event modulates the intensity of others. Neural network-based HPs offer greater flexibility, resulting in improved fit and prediction performance, but at the cost of interpretability, which is often critical in healthcare. In this work, we aim to understand and improve upon this tradeoff. We propose a novel HP formulation in which impact functions are modeled by defining a flexible impact kernel, instantiated as a neural network, in event embedding space, which allows us to model large-scale event sequences with many event types. This approach is more flexible than traditional HPs yet more interpretable than other neural network approaches, and allows us to explicitly trade flexibility for interpretability by adding transformer encoder layers to further contextualize the event embeddings. Results show that our method accurately recovers impact functions in simulations, achieves competitive performance on MIMIC-IV procedure dataset, and gains clinically meaningful interpretation on Duke-EHR with children diagnosis dataset even without transformer layers. This suggests that our flexible impact kernel is often sufficient to capture self-reinforcing dynamics in EHRs and other data effectively, implying that interpretability can be maintained without loss of performance.