Unfolding Physiological State: Mortality Modelling in Intensive Care Units

By Marzyeh Ghassemi, Tristan Naumann, Finale Doshi-Velez, Nicole Brimmer, Rohit Joshi, Anna Rumshisky, Peter Szolovits. 2014

233 citations recorded by the source corpus.

Accurate knowledge of a patient's disease state and trajectory is critical in a clinical setting. Modern electronic healthcare records contain an increasingly large amount of data, and the ability to automatically identify the factors that influence patient outcomes stand to greatly improve the efficiency and quality of care. We examined the use of latent variable models (viz. Latent Dirichlet Allocation) to decompose free-text hospital notes into meaningful features, and the predictive power of these features for patient mortality. We considered three prediction regimes: (1) baseline prediction, (2) dynamic (time-varying) outcome prediction, and (3) retrospective outcome prediction. In each, our prediction task differs from the familiar time-varying situation whereby data accumulates; since fewer patients have long ICU stays, as we move forward in time fewer patients are available and the prediction task becomes increasingly difficult. We found that latent topic-derived features were effective in determining patient mortality under three timelines: in-hospital, 30 day post-discharge, and 1 year post-discharge mortality. Our results demonstrated that the latent topic features important in predicting hospital mortality are very different from those that are important in post-discharge mortality. In general, latent topic features were more predictive than structured features, and a combination of the two performed best. The time-varying models that combined latent topic features and baseline features had AUCs that reached 0.85, 0.80, and 0.77 for in-hospital, 30 day post-discharge and 1 year post-discharge mortality respectively. Our results agreed with other work suggesting that the first 24 hours of patient information are often the most predictive of hospital mortality. Retrospective models that used a combination of latent topic features and structured features achieved AUCs of 0.96, 0.82, and 0.81 for in-hospital, 30 day, and 1-year mortality prediction. Our work focuses on the dynamic (time-varying) setting because models from this regime could facilitate an on-going severity stratification system that helps direct care-staff resources and inform treatment strategies.

Subfields: Artificial Intelligence, Epidemiology, Signal Processing

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

In Epidemiology

In Signal Processing

Citations in the corpus

Natural Language Processing for EHR-Based Computational Phenotyping (2018, background): “[106], [142], [178], [179], applied topic models and extended tensor-based topic models to learn better coherent features.”
Continuous State-Space Models for Optimal Sepsis Treatment: a Deep Reinforcement Learning Approach (2017, background): “The incorporation of time in a supervised setting could be implicit within the feature space construction (Hug and Szolovits, 2009; Joshi and Szolovits, 2012), or captured with multiple models for different time-points (Fialho et al., 2013; Ghassemi et al., 2014).”
Predicting Complications in Critical Care Using Heterogeneous Clinical Data (2016, methodology): “[40] F. Doshi-Velez, B. C. Wallace, R. Adams, ‘‘Graph-sparse LDA: A topic model with structured sparsity,’’ in Proc.”
Multitask learning and benchmarking with clinical time series data (2017, background): “ality prediction. Other related works model the risk of individual conditions such as Clostridium di˝cile infection [58] or sepsis [27] over time. Caballero Barajas and Akella [9] and Ghassemi et al. [22] both train temporal risk models but use them to forecast eventual, rather than imminent, mortal”
A Multivariate Timeseries Modeling Approach to Severity of Illness Assessment and Forecasting in ICU with Sparse, Heterogeneous Clinical Data (2015, methodology): “This is usually accomplished by windowing or aggregating the structured numerical data so that a single feature matrix can be fed into a structured deterministic classifier (Hug and Szolovits 2009; Lehman et al. 2012; Joshi and Szolovits 2012; Ghassemi et al. 2014).”
Supervised topic models for clinical interpretability (2016, methodology): “Topic models such as latent Dirichlet allocation (LDA) (Blei, 2012) are a popular tool for such dimensionality reduction (e.g. Paul and Dredze (2014) or Ghassemi et al. (2014)).”
State of the art review: the data revolution in critical care (2015, background): “These include the heterogeneity of treatment effect of red blood cell (RBC) transfusion [13], the impact of pre-admission selective serotonin reuptake inhibitors on mortality in the ICU [14], the interplay between clinical notes and structured data on mortality prediction [15], optimization of hepar”
Why Is My Classifier Discriminatory? (2018, methodology): “Following prior work on enrichment of topics in clinical notes (Marlin et al., 2012; Ghassemi et al., 2014), we estimate the probability of patient mortality Y given a topic c as p̂(Y |C = c) := ( ∑n i=1 yiqic)/( ∑n i=1 qic) where yi is the hospital mortality of patient i.”
Benchmark of Deep Learning Models on Large Healthcare MIMIC Datasets (2017, background): “The availability of large healthcare databases such as Medical Information Mart for Intensive Care (MIMIC-II and III) databases [4, 5] has accelerated the research in this important area as evidenced by a lot of recent publications [6, 7, 8, 2, 9, 10, 11, 1, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 2”