Search

Medical events are often infrequent, thus becomes hard to predict. In this paper, we focus on predictor that forecasts whether a medical event would occur in the next year, and analyzes the impact of event's frequency and data size via predictor's performance. In the experiment, we made 1572 predictors for medical events using Medical Insurance Claims (MICs) data from 800,000 participants and 205.8 million claims over 8 years. The result revealed that (a) forecasting error will be increased when predicting low-frequency events, and (b) increasing the number of training dataset reduces errors. This result suggests that increasing data size is a key to solve low frequency problems. However, we still need additional methods to cope with sparse and imbalanced data. [ABSTRACT FROM AUTHOR]

Accurate identification of the QRS complex is critical to analyse heart rate variability (HRV), which is linked to various adverse outcomes in premature infants. Reliable and accurate extraction of HRV characteristics at a large scale in the neonatal context remains a challenge. In this paper, we investigate the capabilities of 15 state-of-the-art QRS complex detection implementations using two real-world preterm neonatal datasets. As an attempt to improve the accuracy and reliability, we introduce a weighted ensemble-based method as an alternative. Obtained results indicate the superiority of the proposed method over the state of the art on both datasets with an F1-score of 0.966 (95% CI 0.962-0.97) and 0.893 (95% CI 0.892-0.894). This motivates the deployment of ensemble-based methods for any HRV-based analysis to ensure robust and accurate QRS complex detection. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic has caused millions of infections and deaths worldwide in an ongoing pandemic. With the passage of time, several variants of this virus have surfaced. Machine learning methods and algorithms have been very useful in understanding the virus and its implications so far. In this paper, we have studied a set of novelty detection algorithms and applied it to the problem of detecting COVID-19 variants. Our results show accuracies of 79.64% and 82.43% on the B.1.1.7 and B.1.351 variants respectively on ProtVec unaligned COVID-19 spike protein sequences using One Class SVM with fine-tuned parameters. We believe that a system for automated and timely detection of variants will help countries formulate mitigation measures and study remedies in terms of medicines and vaccines that can protect against the new variants. [ABSTRACT FROM AUTHOR]

Coronary artery disease (CAD) has the highest disease burden worldwide. To manage this burden, predictive models are required to screen patients for preventative treatment. A range of variables have been explored for their capacity to predict disease, including phenotypic (age, sex, BMI and smoking status), medical imaging (carotid artery thickness) and genotypic. We use a machine learning models and the UK Biobank cohort to measure the prediction capacity of these 3 variable categories, both in combination and isolation. We demonstrate that phenotypic variables from the Framingham risk score have the best prediction capacity, although a combination of phenotypic, medical imaging and genotypic variables deliver the most specific models. Furthermore, we demonstrate that Variant Spark, a random forest based GWAS platform, performs effective feature selection for SNP-based genotype variables, identifying 115 significantly associated SNPs to the CAD phenotype. [ABSTRACT FROM AUTHOR]

The lack of transparency and explainability hinders the clinical adoption of Machine learning (ML) algorithms. While explainable artificial intelligence (XAI) methods have been proposed, little research has focused on the agreement between these methods and expert clinical knowledge. This study applies current state-of-the-art explainability methods to clinical decision support algorithms developed for Electronic Medical Records (EMR) data to analyse the concordance between these factors and discusses causes for identified discrepancies from a clinical and technical perspective. Important factors for achieving trustworthy XAI solutions for clinical decision support are also discussed. [ABSTRACT FROM AUTHOR]

Telehealth services are becoming more and more popular, leading to an increasing amount of data to be monitored by health professionals. Machine learning can support them in managing these data. Therefore, the right machine learning algorithms need to be applied to the right data. We have implemented and validated different algorithms for selecting optimal time instances from time series data derived from a diabetes telehealth service. Intrinsic, supervised, and unsupervised instance selection algorithms were analysed. Instance selection had a huge impact on the accuracy of our random forest model for dropout prediction. The best results were achieved with a One Class Support Vector Machine, which improved the area under the receiver operating curve of the original algorithm from 69.91 to 75.88 %. We conclude that, although hardly mentioned in telehealth literature so far, instance selection has the potential to significantly improve the accuracy of machine learning algorithms. [ABSTRACT FROM AUTHOR]

The Cascade-HF protocol is a Continuous Remote Patient Monitoring (CRPM) study at a major health system in the United States to reduce Heart Failure (HF)-related hospitalizations and readmissions using wearable biosensors to collect physiological data over a 30-day period to determine decompensation risk among HF patients. The alerts produced, coupled with electronic patient-reported outcomes, are utilized daily by the home health team, and escalated to the heart failure team as needed, for proactive actions. Limited research has examined anticipating the implementation and workflow challenges of such complex CRPM studies such as resource planning and staffing decisions that leverage the recorded data to drive clinical preparedness and operational efficiency. This preliminary analysis applies discrete event simulation modeling to the Cascade-HF protocol using pilot data from a soft launch to assess workload of the clinical team, evaluate escalation patterns and provide decision support recommendations to enable scale-up for all post-discharge patients. [ABSTRACT FROM AUTHOR]

Pain is a common reason for accessing healthcare resources and is a growing area of research, especially in its overlap with mental health. Mental health electronic health records are a good data source to study this overlap. However, much information on pain is held in the free text of these records, where mentions of pain present a unique natural language processing problem due to its ambiguous nature. This project uses data from an anonymised mental health electronic health records database. A machine learning based classification algorithm is trained to classify sentences as discussing patient pain or not. This will facilitate the extraction of relevant pain information from large databases. 1,985 documents were manually triple-annotated for creation of gold standard training data, which was used to train four classification algorithms. The best performing model achieved an F1-score of 0.98 (95% CI 0.98-0.99). [ABSTRACT FROM AUTHOR]

In the United States, more than 12% of the population will experience thyroid dysfunction. Patient symptoms often reported with thyroid dysfunction include fatigue and weight change. However, little is understood about the relationship between these symptoms documented in the outpatient setting and ordering patterns for thyroid testing among various patient groups by age and sex. We developed a natural language processing and deep learning pipeline to identify patientreported outcomes of weight change and fatigue among patients with a thyroid stimulating hormone test. We built upon prior works by comparing 5 open-source, Bidirectional Encoder Representations from Transformers (BERT) to determine which models could accurately identify these symptoms from clinical texts. For both fatigue (f) and weight change (wc), Bio_ClinicalBERT achieved the highest F1-score (f: 0.900; wc: 0.906) compared BERT (f: 0.899; wc: 0.890), DistilBERT (f: 0.852; wc: 0.912), Biomedical RoBERTa (f: 0.864; wc: 0.904), and PubMedBERT (f: 0.882; wc: 0.892). [ABSTRACT FROM AUTHOR]

Clinical narratives recording behaviours and emotions of patients are available from EHRs in a forensic psychiatric centre located in Tasmania. This rich data has not been used in risk prediction. Prior work demonstrates natural language processing can be used to identify patient symptoms in these free-text records and can then be used to predict risk. Four dictionaries containing descriptive words of harm were created using the Diagnostic and Statistical Manual of Mental Disorders, the Unified Medical Language System repository, English negative and positive sentiment words, and high-frequency words from the Corpus of Contemporary American English. However, a model based only on these keywords is limited in predictive power. In this study, we introduce an improved NLP approach with a social interaction component to extract additional information about the behavioural and emotional state of patients. These social interactions are subsequently used in a machine-learning model to enhance risk prediction performance. [ABSTRACT FROM AUTHOR]

Infant mortality is characterized by the death of young children under the age of one, and it is an issue affecting millions of children in the world. The objective of this article is to employ concepts of knowledge discovery in databases, specifically of machine learning in the data mining phase, to characterize infant mortality in two states of Brazil: Santa Catarina, with the lowest infant mortality rate of the country's states, and Amapá, with the highest. The classifiers C4.5, JRip, Random Forest, SVM, and Multilayer Perceptron were used, and a brief comparison of the results obtained by the classifiers in both states is made. In addition, the dataset preprocessing is detailed, which includes attribute selection and class balancing. The results show that the features APGAR5, WEIGHT, and CONGENITAL ANOMALY stood out the most from the rules generated by the tree-based classifiers. [ABSTRACT FROM AUTHOR]

Patient outcome is one of the key information categories in incident reporting. Being able to extract meaningful patient fall outcomes would allow better analysis of the consequences and possible mitigating actions for in-hospital fall incidents. This study aims to automate the extraction of patient outcomes from narrative fall incident reports by decomposing this into three classification subtasks: injured or not, injury types, and the number of injuries. Implementing an existing incident report classification (IRC) framework, the experimental results demonstrated that oversampling and structured features were effective to achieve better overall performances across all three subtasks. The study further validated the application of an IRC framework to deal with imbalanced classification problems found in fall patient outcome classification and advanced the science of automatic patient outcomes extraction. [ABSTRACT FROM AUTHOR]

We describe new functionalities that have been added to an existing and widely used software solution in Polish public healthcare. The system automatically evaluates a number of medical scores. It provides epidemiologic monitoring concerning infectious diseases; classifies a particular patient to a specific risk group; and detects anomalies. Moreover, an analysis of a prescribed drug is performed. [ABSTRACT FROM AUTHOR]