RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism
By Edward Choi, Mohammad Taha Bahadori, Jimeng Sun, Joshua A. Kulas, Andy Schuetz, Walter F. Stewart. 2016
1,544 citations recorded by the source corpus.
Accuracy and interpretability are two dominant features of successful predictive models. Typically, a choice must be made in favor of complex black box models such as recurrent neural networks (RNN) for accuracy versus less accurate but more interpretable traditional models such as logistic regression. This tradeoff poses challenges in medicine where both accuracy and interpretability are important. We addressed this challenge by developing the REverse Time AttentIoN model (RETAIN) for application to Electronic Health Records (EHR) data. RETAIN achieves high accuracy while remaining clinically interpretable and is based on a two-level neural attention model that detects influential past visits and significant clinical variables within those visits (e.g. key diagnoses). RETAIN mimics physician practice by attending the EHR data in a reverse time order so that recent clinical visits are likely to receive higher attention. RETAIN was tested on a large health system EHR dataset with 14 million visits completed by 263K patients over an 8 year period and demonstrated predictive accuracy and computational scalability comparable to state-of-the-art methods such as RNN, and ease of interpretability comparable to traditional models.
Subfields: Artificial Intelligence, Safety, Robustness & Interpretability, Medical & Health AI
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Mohit Bansal
- Percy Liang
- Yoshua Bengio
- Graham Neubig
- Jianfeng Gao
- Quoc V. Le
- Luke S. Zettlemoyer
- Yoav Goldberg
In Safety, Robustness & Interpretability
- Salvatore Ruggieri
- Dino Pedreschi
- Tomas Pfister
- Sercan Ö. Arik
- Franco Turini
- Bryan Lim
- Bettina Berendt
- Ilya Shpitser
In Medical & Health AI
Citations in the corpus
- Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, neutral): “Aside from the approaches that do not change the inner workings of the RNNs, [285] presents RETAIN (REverse Time AttentIoN) model, which detects influential past patterns by means of a two-level neural attention model.”
- Time-series forecasting with deep learning: a survey (2020, methodology): “Given the flexibility of neural networks, deep neural networks have been used to model both discrete [55] and continuous [37,56] targets—by customizing of decoder and output layer of the neural network to match the desired target type.”
- Reinforcement Learning in Healthcare: A Survey (2019, background): “rvised classification problem, existing machining learning methods on clinical diagnosis heavily rely on a large number of annotated samples in order to infer and predict the possible diagnoses [244], [245], [246]. Moreover, these methods have limits in terms of capturing the underlying dynamics and ”
- Interpretable and Steerable Sequence Learning via Prototypes (2019, methodology): “One example of such models is RETAIN [9], which uses LSTM with attention mechanism for predictive analysis on patient data.”
- G-Net: A Deep Learning Approach to G-computation for Counterfactual Outcome Prediction Under Dynamic Treatment Regimes (2020, background): “RNNs, and in particular LSTMs, have achieved state of the art performance on a wide variety of time series regression tasks, including healthcare related tasks [17–19].”
- Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting (2019, methodology): “Methods combining temporal importance and variable selection have also been considered in [24], which computes a single contribution coefficient based on attention weights from each.”
- Opportunities and obstacles for deep learning in biology and medicine (2017, background): “[353] leveraged attention mechanisms to highlight which aspects of a patient's medical history were most relevant for making diagnoses.”
- Learning to Deceive with Attention-Based Explanations (2019, background): “Similar claims about attention’s utility for interpreting models’ predictions are common in the literature (Li et al., 2016; Xu et al., 2015; Choi et al., 2016; Xie et al., 2017; Martins and Astudillo, 2016; Lai and Tan, 2019).”
- GRAM: Graph-based Attention Model for Healthcare Representation Learning (2016, methodology): “…(HF) cohort, which is a subset of Sutter PAMF data for a heart failure onset prediction study with 3.4K HF cases chosen by a set of criteria described in Vijayakrishnan et al. (2014); Gurwitz et al. (2013) and 27K matching controls chosen by a set of criteria described in Choi et al. (2016e).”
- What Clinicians Want: Contextualizing Explainable Machine Learning for Clinical End Use (2019, methodology): “Uncertainty Distributional shift (Subbaswamy and Saria, 2018) Not evaluated on complex clinical models Parameter uncertainty (Gal and Ghahramani, 2016; Schulam and Saria, 2019) Temporal Explanations RETAIN, RAIM (Xu et al., 2018; Choi et al., 2016) Potential lack of consistency due to the attention ”
- Learning the Graphical Structure of Electronic Health Records with Graph Convolutional Transformer (2019, background): “…organizations enabled deep learning methods to show impressive performance in diverse tasks such as predicting diagnosis (Lipton et al. 2015; Choi et al. 2016a; Rajkomar et al. 2018), learning medical concept representations (Che et al. 2015; Choi, Chiu, and Sontag 2016; Choi et al. 2016b;…”
- Human-in-the-loop machine learning: a state of the art (2022, background): “…understanding what an RNN model has learned, mainly via feature relevance methods (Arras et al. 2017; Che et al. 2015); and (2) explainability by modifying RNN architectures to provide insights about the decisions they make via local explanations (Krakovna and Doshi-Velez 2016; Choi et al. 2016).”
- A Semiparametric Approach to Interpretable Machine Learning (2020, methodology): “[6] make use of an interpretable two-level attention model for prediction purposes.”
- Understanding Multi-Head Attention in Abstractive Summarization (2019, background): “, 2015) has gained popularity as a means of obtaining insight in the inner workings of deep neural networks (e.g., Lei, 2017; Choi et al., 2016; Gilpin et al., 2018; Ghaeini et al., 2018).”
- Exploring Interpretable LSTM Neural Networks over Multi-Variable Data (2019, methodology): “The contribution coefficients defined on attention values is biased as well (Choi et al., 2016).”
- Unbox the black-box for the medical explainable AI via multi-modal and multi-centre data fusion: A mini-review, two showcases and beyond (2021, methodology): “Choi et al. [79] implemented the Reverse Time Attention Model (RETAIN), which incorporated two sets of attention weights, one for visit level to capture the effect of each visit and the other at the variable-level.”