Can Machine Learning Personalize Cardiovascular Therapy in Sepsis?

Large randomized trials in sepsis have generally failed to find effective novel treatments. This is increasingly attributed to patient heterogeneity, including heterogeneous cardiovascular changes in septic shock. We discuss the potential for machine learning systems to personalize cardiovascular resuscitation in sepsis. While the literature is replete with proofs of concept, the technological readiness of current systems is low, with a paucity of clinical trials and proven patient benefit. Systems may be vulnerable to confounding and poor generalization to new patient populations or contemporary patterns of care. Typical electronic health records do not capture rich enough data, at sufficient temporal resolution, to produce systems that make actionable treatment suggestions. To resolve these issues, we recommend a simultaneous focus on technical challenges and removing barriers to translation. This will involve improving data quality, adopting causally grounded models, prioritizing safety assessment and integration into healthcare workflows, conducting randomized clinical trials and aligning with regulatory requirements.

Highlighting uncertainty in clinical risk prediction using a model of emergency laparotomy mortality risk.

Clinical prediction models typically make point estimates of risk. However, values of key variables are often missing during model development or at prediction time, meaning that the point estimates mask significant uncertainty and can lead to over-confident decision making. We present a model of mortality risk in emergency laparotomy which instead presents a distribution of predicted risks, highlighting the uncertainty over the risk of death with an intuitive visualisation. We developed and validated our model using data from 127134 emergency laparotomies from patients in England and Wales during 2013-2019. We captured the uncertainty arising from missing data using multiple imputation, allowing prospective, patient-specific imputation for variables that were frequently missing. Prospective imputation allows early prognostication in patients where these variables are not yet measured, accounting for the additional uncertainty this induces. Our model showed good discrimination and calibration (95% confidence intervals: Brier score 0.071-0.078, C statistic 0.859-0.873, calibration error 0.031-0.059) on unseen data from 37 hospitals, consistently improving upon the current gold-standard model. The dispersion of the predicted risks varied significantly between patients and increased where prospective imputation occurred. We present a case study that illustrates the potential impact of uncertainty quantification on clinical decision making. Our model improves mortality risk prediction in emergency laparotomy and has the potential to inform decision-makers and assist discussions with patients and their families. Our analysis code was robustly developed and is publicly available for easy replication of our study and adaptation to predicting other outcomes.

Temporal convolutional networks allow early prediction of events in critical care.

OBJECTIVE: Clinical interventions and death in the intensive care unit (ICU) depend on complex patterns in patients' longitudinal data. We aim to anticipate these events earlier and more consistently so that staff can consider preemptive action. MATERIALS AND METHODS: We use a temporal convolutional network to encode longitudinal data and a feedforward neural network to encode demographic data from 4713 ICU admissions in 2014-2018. For each hour of each admission, we predict events in the subsequent 1-6 hours. We compare performance with other models including a recurrent neural network. RESULTS: Our model performed similarly to the recurrent neural network for some events and outperformed it for others. This performance increase was more evident in a sensitivity analysis where the prediction timeframe was varied. Average positive predictive value (95% CI) was 0.786 (0.781-0.790) and 0.738 (0.732-0.743) for up- and down-titrating FiO2, 0.574 (0.519-0.625) for extubation, 0.139 (0.117-0.162) for intubation, 0.533 (0.492-0.572) for starting noradrenaline, 0.441 (0.433-0.448) for fluid challenge, and 0.315 (0.282-0.352) for death. DISCUSSION: Events were better predicted where their important determinants were captured in structured electronic health data, and where they occurred in homogeneous circumstances. We produce partial dependence plots that show our model learns clinically-plausible associations between its inputs and predictions. CONCLUSION: Temporal convolutional networks improve prediction of clinical events when used to represent longitudinal ICU data.