•We review current work in the field of automated machine learning (AutoML) from a computer science and biomedical perspective.•AutoML is a growing field that seeks to automatically select, compose, ...and parametrize machine learning models.•AutoML reduces the demand for human experts, making it a valuable tool for healthcare.•While there have been some use cases of AutoML in healthcare, more work needs to be done in order for there to be widespread adoption.
This work aims to provide a review of the existing literature in the field of automated machine learning (AutoML) to help healthcare professionals better utilize machine learning models “off-the-shelf” with limited data science expertise. We also identify the potential opportunities and barriers to using AutoML in healthcare, as well as existing applications of AutoML in healthcare.
Published papers, accompanied with code, describing work in the field of AutoML from both a computer science perspective or a biomedical informatics perspective were reviewed. We also provide a short summary of a series of AutoML challenges hosted by ChaLearn.
A review of 101 papers in the field of AutoML revealed that these automated techniques can match or improve upon expert human performance in certain machine learning tasks, often in a shorter amount of time. The main limitation of AutoML at this point is the ability to get these systems to work efficiently on a large scale, i.e. beyond small- and medium-size retrospective datasets.
The utilization of machine learning techniques has the demonstrated potential to improve health outcomes, cut healthcare costs, and advance clinical research. However, most hospitals are not currently deploying machine learning solutions. One reason for this is that health care professionals often lack the machine learning expertise that is necessary to build a successful model, deploy it in production, and integrate it with the clinical workflow. In order to make machine learning techniques easier to apply and to reduce the demand for human experts, automated machine learning (AutoML) has emerged as a growing field that seeks to automatically select, compose, and parametrize machine learning models, so as to achieve optimal performance on a given task and/or dataset.
While there have already been some use cases of AutoML in the healthcare field, more work needs to be done in order for there to be widespread adoption of AutoML in healthcare.
Clinicians document cancer patients' symptoms in free-text format within electronic health record visit notes. Although symptoms are critically important to quality of life and often herald clinical ...status changes, computational methods to assess the trajectory of symptoms over time are woefully underdeveloped.
To create machine learning algorithms capable of extracting patient-reported symptoms from free-text electronic health record notes.
The data set included 103,564 sentences obtained from the electronic clinical notes of 2695 breast cancer patients receiving paclitaxel-containing chemotherapy at two academic cancer centers between May 1996 and May 2015. We manually annotated 10,000 sentences and trained a conditional random field model to predict words indicating an active symptom (positive label), absence of a symptom (negative label), or no symptom at all (neutral label). Sentences labeled by human coder were divided into training, validation, and test data sets. Final model performance was determined on 20% test data unused in model development or tuning.
The final model achieved precision of 0.82, 0.86, and 0.99 and recall of 0.56, 0.69, and 1.00 for positive, negative, and neutral symptom labels, respectively. The most common positive symptoms were pain, fatigue, and nausea. Machine-based labeling of 103,564 sentences took two minutes.
We demonstrate the potential of machine learning to gather, track, and analyze symptoms experienced by cancer patients during chemotherapy. Although our initial model requires further optimization to improve the performance, further model building may yield machine learning methods suitable to be deployed in routine clinical care, quality improvement, and research applications.
Increasing recognition of biases in artificial intelligence (AI) algorithms has motivated the quest to build fair models, free of biases. However, building fair models may be only half the challenge. ...A seemingly fair model could involve, directly or indirectly, what we call "latent biases." Just as latent errors are generally described as errors "waiting to happen" in complex systems, latent biases are biases waiting to happen. Here we describe 3 major challenges related to bias in AI algorithms and propose several ways of managing them. There is an urgent need to address latent biases before the widespread implementation of AI algorithms in clinical practice.
Mutation in the TSC2 tumor suppressor causes tuberous sclerosis complex, a disease characterized by hamartoma formation in multiple tissues. TSC2 inhibits cell growth by acting as a GTPase-activating ...protein toward Rheb, thereby inhibiting mTOR, a central controller of cell growth. Here, we show that Wnt activates mTOR via inhibiting GSK3 without involving β-catenin-dependent transcription. GSK3 inhibits the mTOR pathway by phosphorylating TSC2 in a manner dependent on AMPK-priming phosphorylation. Inhibition of mTOR by rapamycin blocks Wnt-induced cell growth and tumor development, suggesting a potential therapeutic value of rapamycin for cancers with activated Wnt signaling. Our results show that, in addition to transcriptional activation, Wnt stimulates translation and cell growth by activating the TSC-mTOR pathway. Furthermore, the sequential phosphorylation of TSC2 by AMPK and GSK3 reveals a molecular mechanism of signal integration in cell growth regulation.
Latent knowledge can be extracted from the electronic notes that are recorded during patient encounters with the health system. Using these clinical notes to decipher a patient's underlying ...comorbidites, symptom burdens, and treatment courses is an ongoing challenge. Latent topic model as an efficient Bayesian method can be used to model each patient's clinical notes as "documents" and the words in the notes as "tokens". However, standard latent topic models assume that all of the notes follow the same topic distribution, regardless of the type of note or the domain expertise of the author (such as doctors or nurses). We propose a novel application of latent topic modeling, using multi-note topic model (MNTM) to jointly infer distinct topic distributions of notes of different types. We applied our model to clinical notes from the MIMIC-III dataset to infer distinct topic distributions over the physician and nursing note types. Based on manual assessments made by clinicians, we observed a significant improvement in topic interpretability using MNTM modeling over the baseline single-note topic models that ignore the note types. Moreover, our MNTM model led to a significantly higher prediction accuracy for prolonged mechanical ventilation and mortality using only the first 48 hours of patient data. By correlating the patients' topic mixture with hospital mortality and prolonged mechanical ventilation, we identified several diagnostic topics that are associated with poor outcomes. Because of its elegant and intuitive formation, we envision a broad application of our approach in mining multi-modality text-based healthcare information that goes beyond clinical notes. Code available at https://github.com/li-lab-mcgill/heterogeneous_ehr.
Multiple clinical trials support the effectiveness of cardiac resynchronization therapy (CRT); however, optimal patient selection remains challenging due to substantial treatment heterogeneity among ...patients who meet the clinical practice guidelines.
To apply machine learning to create an algorithm that predicts CRT outcome using electronic health record (EHR) data avaible before the procedure.
We applied machine learning and natural language processing to the EHR of 990 patients who received CRT at two academic hospitals between 2004-2015. The primary outcome was reduced CRT benefit, defined as <0% improvement in left ventricular ejection fraction (LVEF) 6-18 months post-procedure or death by 18 months. Data regarding demographics, laboratory values, medications, clinical characteristics, and past health services utilization were extracted from the EHR available before the CRT procedure. Bigrams (i.e., two-word sequences) were also extracted from the clinical notes using natural language processing. Patients accrued on average 75 clinical notes (SD, 29) before the procedure including data not captured anywhere else in the EHR. A machine learning model was built using 80% of the patient sample (training and validation dataset), and tested on a held-out 20% patient sample (test dataset). Among 990 patients receiving CRT the mean age was 71.6 (SD, 11.8), 78.1% were male, 87.2% non-Hispanic white, and the mean baseline LVEF was 24.8% (SD, 7.69). Out of 990 patients, 403 (40.7%) were identified as having a reduced benefit from the CRT device (<0% LVEF improvement in 25.2%, death by 18 months in 15.6%). The final model identified 26% of these patients at a positive predictive value of 79% (model performance: Fβ (β = 0.1): 77%; recall 0.26; precision 0.79; accuracy 0.65).
A machine learning model that leveraged readily available EHR data and clinical notes identified a subset of CRT patients who may not benefit from CRT before the procedure.
Background Among patients with acute respiratory failure requiring prolonged mechanical ventilation, tracheostomies are typically placed after approximately 7 to 10 days. Yet half of patients ...admitted to the intensive care unit receiving tracheostomy will die within a year, often within three months. Existing mortality prediction models for prolonged mechanical ventilation, such as the ProVent Score, have poor sensitivity and are not applied until after 14 days of mechanical ventilation. We developed a model to predict 3-month mortality in patients requiring more than 7 days of mechanical ventilation using deep learning techniques and compared this to existing mortality models. Methods Retrospective cohort study. Setting: The Medical Information Mart for Intensive Care III Database. Patients: All adults requiring greater than or equal to 7 days of mechanical ventilation. Measurements: A neural network model for 3-month mortality was created using process-of-care variables, including demographic, physiologic and clinical data. The area under the receiver operator curve (AUROC) was compared to the ProVent model at predicting 3 and 12-month mortality. Shapley values were used to identify the variables with the greatest contributions to the model. Results There were 4,334 encounters divided into a development cohort (n = 3467) and a testing cohort (n = 867). The final deep learning model included 250 variables and had an AUROC of 0.74 for predicting 3-month mortality at day 7 of mechanical ventilation versus 0.59 for the ProVent model. Older age and elevated Simplified Acute Physiology Score II (SAPS II) Score on intensive care unit admission had the largest contribution to predicting mortality. Discussion We developed a deep learning prediction model for 3-month mortality among patients requiring greater than or equal to 7 days of mechanical ventilation using a neural network approach utilizing readily available clinical variables. The model outperforms the ProVent model for predicting mortality among patients requiring greater than or equal to 7 days of mechanical ventilation. This model requires external validation.
Mitigating bias in AI at the point of care DeCamp, Matthew; Lindvall, Charlotta
Science (American Association for the Advancement of Science),
2023-Jul-14, 2023-07-14, 20230714, Letnik:
381, Številka:
6654
Journal Article
Recenzirano
Promoting equity in AI in health care requires addressing biases at cli nical implementation.
Adverse drug events (ADEs) are common and have serious consequences in older adults. ED visits are opportunities to identify and alter the course of such vulnerable patients. Current practice, ...however, is limited by inaccurate reporting of medication list, time-consuming medication reconciliation, and poor ADE assessment. This manuscript describes a novel approach to predict, detect, and intervene vulnerable older adults at risk of ADE using machine learning. Toxicologists’ expertise in ADE is essential to creating the machine learning algorithm. Leveraging the existing electronic health records to better capture older adults at risk of ADE in the ED may improve their care.