Healthcare Datasets Research Articles

Unit-Power Half-Normal Distribution Including Quantile Regression with Applications to Medical Data

In this paper, we present the unit-power half-normal distribution, derived from the power half-normal distribution, for data analysis in the open unit interval. The statistical properties of the unit-power half-normal model are described in detail. Simulation studies are carried out to evaluate the performance of the parameter estimators. Additionally, we implement the quantile regression for this model, which is applied to two real healthcare data sets. Our findings suggest that the unit power half-normal distribution provides a robust and flexible alternative for existing models for proportion data.

Improving class probability estimates in asymmetric health data classification: An experimental comparison of novel calibration methods

In the context of health data classification, imbalanced and asymmetric class distributions can significantly impact the performance of machine learning models. One critical aspect affected by these issues is the reliability of class probability estimates, which are crucial for informed decision-making in healthcare applications. Instead of predicting class values directly for a classification problem, it can be more convenient to predict the probability of an observation belonging to each possible class. This research aims to address the challenges posed by imbalanced and asymmetric responses in health data classification by evaluating the effectiveness of recent calibration methods in improving class probability estimates. We propose Beta calibration techniques and the Stratified Brier score and Jaccard's Score as novel calibration methods and evaluation metrics respectively. The experimental comparison involves implementing and assessing various calibration techniques to determine their impact on model performance and calibration accuracy of simulated and healthcare datasets with varying imbalance ratios. Our results show that the Beta calibration method consistently improved the classifiers' predictive ability. The findings of this study provide valuable insights into selecting the most suitable calibration method for enhancing class probability estimates in healthcare-related machine learning tasks.

Increase in major osteoporotic fractures after therapy with immune checkpoint inhibitors

BackgroundImmune checkpoint inhibitors (ICIs) can cause severe and sometimes long-standing immune-related adverse events (irAEs). Enhanced immune activation from ICI can theoretically result in osteoclast activation, bone loss and fracture. The...

Benchmarking commercial healthcare claims data.

Commercial healthcare claims datasets represent a sample of the US population that is biased along socioeconomic/demographic lines; depending on the target population of interest, results derived from these datasets may not generalize. Rigorous comparisons of claims-derived results to ground-truth data that quantify this bias are lacking. (1) To quantify the extent and variation of the bias associated with commercial healthcare claims data with respect to different target populations; (2) To evaluate how socioeconomic/demographic factors may explain the magnitude of the bias. This is a retrospective observational study. Healthcare claims data come from the Merative™ MarketScan® Commercial Database; reference data for comparison come from the State Inpatient Databases (SID) and the US Census. We considered three target populations, aged 18-64 years: (1) all Americans; (2) Americans with health insurance; (3) Americans with commercial health insurance. We analyzed inpatient discharge records of patients aged 18-64 years, occurring between 01/01/2019 to 12/31/2019 in five states: California, Iowa, Maryland, Massachusetts, and New Jersey. We estimated rates of the 250 most common inpatient procedures, using claims data and using reference data for each target population, and we compared the two estimates. The average rate of inpatient discharges per 100 person-years was 5.39 in the claims data (95% CI: [5.37, 5.40]) and 7.003 (95% CI: [7.002, 7.004]) in the reference data for all Americans, corresponding to a 23.1% underestimate from claims. We found large variation in the extent of relative bias across inpatient procedures, including 22.8% of procedures that were underestimated by more than a factor of 2. There was a significant relationship between socioeconomic/demographic factors and the magnitude of bias: procedures that disproportionately occur in disadvantaged neighborhoods were more underestimated in claims data (R 2 51.6%, p < 0.001). When the target population was restricted to commercially insured Americans, the bias decreased substantially (3.2% of procedures were biased by more than factor of 2), but some variation across procedures remained. Naïve use of healthcare claims data to derive estimates for the underlying US population can be severely biased. The extent of bias is at least partially explained by neighborhood-level socioeconomic factors.

The Multicollinearity Effect on the Performance of Machine Learning Algorithms: Case Examples in Healthcare Modelling

Background: The data extracted from various fields inherently consists of extremely correlated measurements in parallel with the exponential increase in the size of the data that need to be interpreted owing to the technological advances. This problem, called the multicollinearity, influences the performance of both statistical and machine learning algorithms. Statistical models proposed as a potential remedy to this problem have not been sufficiently evaluated in the literature. Therefore, a comprehensive comparison of statistical and machine learning models is required for addressing the multicollinearity problem. Methods: Statistical models (including Ridge, Liu, Lasso and Elastic Net regression) and the eight most important machine learning algorithms (including Cart, Knn, Mlp, MARS, Cubist, Svm, Bagging and XGBoost) are comprehensively compared by using two different healthcare datasets (including Body Fat and Cancer) having multicollinearity problem. The performance of the models is assessed through cross validation methods via root mean square error, mean absolute error and r-squared criteria. Results: The results of the study revealed that statistical models outperformed machine learning models in terms of root mean square error, mean absolute error and r-squared criteria in both training and testing performance. Particularly the Liu regression often achieved better relative performance (up to 7.60% to 46.08% for Body Fat data set and up to 1.55% to 21.53% for Cancer data set on training performance and up to 1.56% to 38.08% for Body Fat data set and up to 3.50% to 23.29% for Cancer data set on testing performance) among regression methods as well as compared to machine algorithms. Conclusions: Liu regression is mostly disregarded in the machine learning literature, but since it outperforms the most powerful and widely used machine learning algorithms, it appears to be a promising tool in almost all fields, especially for regression-based studies including data with multicollinearity problem.

Age-stratified predictions of suicide attempts using machine learning in middle and late adolescence

BackgroundPrevalence of suicidal behaviour increases rapidly in middle to late adolescence. Predicting suicide attempts across different ages would enhance our understanding of how suicidal behaviour manifests in this period of rapid development. This study aimed to develop separate models to predict suicide attempts within a cohort at middle and late adolescence. It also sought to examine differences between the models derived across both developmental stages. MethodsThis study used data from the nationally representative Longitudinal Study of Australian Children (N = 2266). We selected over 700 potential suicide attempt predictors measured via self-report questionnaires, and linked healthcare and education administrative datasets. Logistic regression, random forests, and gradient boosting algorithms were developed to predict suicide attempts across two stages (mid-adolescence: 14–15 years; late adolescence: 18–19 years) using predictors sampled two years prior (mid-adolescence: 12–13 years; late adolescence: 16–17 years). ResultsThe late adolescence models (AUROC = 0.77–0.88, F1-score = 0.22–0.28, Sensitivity = 0.54–0.64) performed better than the mid-adolescence models (AUROC = 0.70–0.76, F1-score = 0.12–0.19, Sensitivity = 0.40–0.64). The most important features for predicting suicide attempts in mid-adolescence were mostly school-related, while the most important features in late adolescence included measures of prior suicidality, psychosocial health, and future plans. ConclusionsTo date, this is the first study to use machine learning models to predict suicide attempts at different ages. Our findings suggest that the optimal suicide risk prediction model differs by stage of adolescence. Future research and interventions should consider that risk presentations can change rapidly during adolescence.

JmBIG: enhancing dynamic risk prediction and personalized medicine through joint modeling of longitudinal and survival data in big routinely collected data

We have introduced the R package jmBIG to facilitate the analysis of large healthcare datasets and the development of predictive models. This package provides a comprehensive set of tools and functions specifically designed for the joint modelling of longitudinal and survival data in the context of big data analytics. The jmBIG package offers efficient and scalable implementations of joint modelling algorithms, allowing for integrating large-scale healthcare datasets.By utilizing the capabilities of jmBIG, researchers and analysts can effectively handle the challenges associated with big healthcare data, such as high dimensionality and complex relationships between multiple outcomes.With the support of jmBIG, analysts can seamlessly fit Bayesian joint models, generate predictions, and evaluate the performance of the models. The package incorporates cutting-edge methodologies and harnesses the computational capabilities of parallel computing to accelerate the analysis of large-scale healthcare datasets significantly. In summary, jmBIG empowers researchers to gain deeper insights into disease progression and treatment response, fostering evidence-based decision-making and paving the way for personalized healthcare interventions that can positively impact patient outcomes on a larger scale.

Exploring The Efficiency of Metaheuristics in Optimal Hyperparameter Tuning for Ensemble Models on Varied Data Modalities

Effective disease detection systems play an important role in healthcare by supporting diagnosis and treatment. This study provides a comparison of hyperparameter tuning methods for disease detection systems using four health datasets; kidney disease, diabetes detection, heart disease and breast cancer detection. The main objective of this research is to prepare datasets by normalizing the input and testing machine learning models such as Naive Bayes Support Vector Machine (SVM), Logistic Regression and k Nearest Neighbor (kNN). to identify effective models for each data set. After implementing the models, we apply three hyperparameter tuning techniques: Grid search, random search, and particle ensemble optimization (PSO). These methods are used to tune the model parameters. Improve overall performance metrics. The evaluation focuses on accuracy measurements to compare model performance before and after hyperparameter tuning. The results of this study illustrate how different tuning techniques can improve the performance of disease detection systems across a range of healthcare datasets. By conducting testing and analysis, we determine the appropriate tuning method for each data set, yielding valuable insights, to develop an accurate and effective disease detection system .These discoveries serve to advance the field of healthcare analytics and machine learning to deliver outcomes for patients and healthcare services.

On responsible machine learning datasets emphasizing fairness, privacy and regulatory norms with examples in biometrics and healthcare

Artificial Intelligence (AI) has seamlessly integrated into numerous scientific domains, catalysing unparalleled enhancements across a broad spectrum of tasks; however, its integrity and trustworthiness have emerged as notable concerns. The scientific community has focused on the development of trustworthy AI algorithms; however, machine learning and deep learning algorithms, popular in the AI community today, intrinsically rely on the quality of their training data. These algorithms are designed to detect patterns within the data, thereby learning the intended behavioural objectives. Any inadequacy in the data has the potential to translate directly into algorithms. In this study we discuss the importance of responsible machine learning datasets through the lens of fairness, privacy and regulatory compliance, and present a large audit of computer vision datasets. Despite the ubiquity of fairness and privacy challenges across diverse data domains, current regulatory frameworks primarily address human-centric data concerns. We therefore focus our discussion on biometric and healthcare datasets, although the principles we outline are broadly applicable across various domains. The audit is conducted through evaluation of the proposed responsible rubric. After surveying over 100 datasets, our detailed analysis of 60 distinct datasets highlights a universal susceptibility to fairness, privacy and regulatory compliance issues. This finding emphasizes the urgent need for revising dataset creation methodologies within the scientific community, especially in light of global advancements in data protection legislation. We assert that our study is critically relevant in the contemporary AI context, offering insights and recommendations that are both timely and essential for the ongoing evolution of AI technologies.

A novel approach for e-health recommender systems

The increasing use of the internet for health information brings challenges due to the complexity and abundance of data, leading to information overload. This highlights the necessity of implementing recommender systems (RSs) within the healthcare domain, with the aim of facilitating more effective and precise healthcare-related decisions for both healthcare providers and users. Health recommendation systems can suggest suitable healthcare items or services based on users' health conditions and needs, including medications, diagnoses, hospitals, doctors, and healthcare services. Despite their potential benefits, RSs encounter significant limitations, including data sparsity, which can lead to recommendations that are unreliable and misleading. Considering the increasing significance of health recommendation systems and the challenge of sparse data, we propose an effective approach to improve precision and coverage in recommending healthcare items or services. This aims to assist users and healthcare practitioners in making informed decisions tailored to their unique needs and health conditions. Empirical testing on two healthcare rating datasets, including sparse datasets, illustrate that our proposed approach outperforms baseline recommendation methods. It excels in improving both the precision and coverage of health-related recommendations, demonstrating effective handling of extremely sparse datasets.

Implementing and evaluating simple resampling techniques in federated learning for imbalanced data

Federated Learning (FL) has emerged as a promising paradigm for privacy-preserving collaborative machine learning. However, the challenge of data imbalance, exacerbated by the non-IID nature of distributed datasets, significantly impacts model performance and fairness in FL systems. This paper investigates the implementation and evaluation of simple resampling techniques to address data imbalance within the FL framework. Using the MIMIC-III healthcare dataset, a simulated FL environment with ten virtual clients was created to test various resampling methods: SMOTE, random undersampling, and a hybrid approach. The study employed logistic regression models and evaluated performance using common and novel FL-specific metrics. Experimental results demonstrate that the hybrid resampling technique significantly outperforms other methods, improving the F1-score by 13.1% and reducing communication rounds by 25.3%. Statistical analyses, including repeated measures ANOVA and hierarchical linear modeling, confirm the robustness of these findings across varied client data distributions. This research provides a replicable framework for addressing data imbalance in FL, contributing to enhanced model fairness and efficiency in privacy-sensitive applications.

Report on the Fifth International Workshop on Health Data Management in the Era of AI (HeDAI 2023)

Artificial intelligence (AI) technologies have the potential to drastically improve healthcare delivery globally. However, healthcare data are diverse, complex, and very large in size. Better information management techniques are needed to deal with the volume and heterogeneity of healthcare data. Better AI techniques are needed to fully harness healthcare datasets to improve healthcare delivery and quality of care for patients. Our workshop aimed to bring together diverse group of researchers and practitioners interested in developing next-generation solutions for challenging problems in healthcare systems. A wide range of topics ranging from federated approaches for data sharing to machine learning for predictive analytics to medical question answering were discussed during the workshop.

Communicating exploratory unsupervised machine learning analysis in age clustering for paediatric disease

BackgroundDespite the increasing availability of electronic healthcare record (EHR) data and wide availability of plug-and-play machine learning (ML) Application Programming Interfaces, the adoption of data-driven decision-making within routine hospital workflows...

Optimization of machine learning models through quantization and data bit reduction in healthcare datasets

This study focuses on enhancing complex machine learning models through quantization and data bit reduction. The primary goal is to reduce processing time while maintaining model performance, which is particularly relevant for intricate models with prolonged execution times. The research employs two medical datasets, namely Heart Disease Prediction and Breast Cancer Detection, and applies optimization techniques to K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) machine-learning models. To achieve optimization, the study employs effective quantization and data bit reduction techniques such as QuantileTransformer, Numpy.round, and KBinsDiscretizer functions. These techniques are utilized to convert input data from float64 to float32 and int32, resulting in a streamlined data representation. The trade-off between processing time and model accuracy is explored, acknowledging that some compromise in accuracy might occur after optimization. The experimentation reveals that there is a noticeable reduction in time complexity after optimization, with a marginal impact on model accuracy. Interestingly, the study concludes that the outcome and efficiency of optimization techniques are influenced not only by the specific technique used but also by the nature of the dataset and machine learning model under consideration. This comprehensive research showcases the applicability of optimization techniques, specifically quantization and data bit reduction, in complex machine learning models. By conducting experiments on medical datasets and analyzing KNN and SVM models, the study underscores the delicate balance between processing time and model accuracy. The findings emphasize that the success of optimization strategies is context-dependent, relying not only on the chosen technique but also on the interplay between the technique, model, and dataset.

Accessible Ecosystem for Clinical Research (Federated Learning for Everyone): Development and Usability Study.

The integrity and reliability of clinical research outcomes rely heavily on access to vast amounts of data. However, the fragmented distribution of these data across multiple institutions, along with ethical and regulatory barriers, presents significant challenges to accessing relevant data. While federated learning offers a promising solution to leverage insights from fragmented data sets, its adoption faces hurdles due to implementation complexities, scalability issues, and inclusivity challenges. This paper introduces Federated Learning for Everyone (FL4E), an accessible framework facilitating multistakeholder collaboration in clinical research. It focuses on simplifying federated learning through an innovative ecosystem-based approach. The "degree of federation" is a fundamental concept of FL4E, allowing for flexible integration of federated and centralized learning models. This feature provides a customizable solution by enabling users to choose the level of data decentralization based on specific health care settings or project needs, making federated learning more adaptable and efficient. By using an ecosystem-based collaborative learning strategy, FL4E encourages a comprehensive platform for managing real-world data, enhancing collaboration and knowledge sharing among its stakeholders. Evaluating FL4E's effectiveness using real-world health care data sets has highlighted its ecosystem-oriented and inclusive design. By applying hybrid models to 2 distinct analytical tasks-classification and survival analysis-within real-world settings, we have effectively measured the "degree of federation" across various contexts. These evaluations show that FL4E's hybrid models not only match the performance of fully federated models but also avoid the substantial overhead usually linked with these models. Achieving this balance greatly enhances collaborative initiatives and broadens the scope of analytical possibilities within the ecosystem. FL4E represents a significant step forward in collaborative clinical research by merging the benefits of centralized and federated learning. Its modular ecosystem-based design and the "degree of federation" feature make it an inclusive, customizable framework suitable for a wide array of clinical research scenarios, promising to revolutionize the field through improved collaboration and data use. Detailed implementation and analyses are available on the associated GitHub repository.

Postpartum haemorrhage and risk of cardiovascular disease in later life: A population-based record linkage cohort study.

To investigate the association between postpartum haemorrhage (PPH) and subsequent cardiovascular disease. Population-based retrospective cohort study, using record linkage between Aberdeen Maternity and Neonatal Databank (AMND) and Scottish healthcare data sets. Grampian region, Scotland. A cohort of 70 904 women who gave birth after 24 weeks of gestation in the period 1986-2016. We used extended Cox regression models to investigate the association between having had one or more occurrences of PPH in any (first or subsequent) births (exposure) and subsequent cardiovascular disease, adjusted for sociodemographic, medical, and pregnancy and birth-related factors. Cardiovascular disease identified from the prescription of selected cardiovascular medications, hospital discharge records or death from cardiovascular disease. In our cohort of 70 904 women (with 124 795 birth records), 25 177 women (36%) had at least one PPH. Compared with not having a PPH, having at least one PPH was associated with an increased risk of developing cardiovascular disease, as defined above, in the first year after birth (adjusted hazard ratio, aHR 1.96; 95% confidence interval, 95% CI 1.51-2.53; p < 0.001). The association was attenuated over time, but strong evidence of increased risk remained at 2-5 years (aHR 1.19, 95% CI 1.11-1.30, P < 0.001) and at 6-15 years after giving birth (aHR 1.17, 95% CI 1.05-1.30, p = 0.005). Compared with women who have never had a PPH, women who have had at least one episode of PPH are twice as likely to develop cardiovascular disease in the first year after birth, and some increased risk persists for up to 15 years.

RWD112 Going Beyond Claims: Unleashing the Power of Diverse Healthcare Datasets in Clinical and HEOR Assessments

RWD112 Going Beyond Claims: Unleashing the Power of Diverse Healthcare Datasets in Clinical and HEOR Assessments

Risk of drug-related death associated with co-prescribing of gabapentinoids and Z-drugs among people receiving opioid-agonist treatment: A national retrospective cohort study

BackgroundPrescribing of gabapentinoids and Z-drug-hypnotics has increased in the population and among people receiving opioid-agonist treatment (OAT) for opioid dependence. Evidence is mixed on whether co-prescribing of sedatives such as gabapentinoids and Z-drugs during OAT increases risk of drug-related death (DRD). MethodsWe conducted a retrospective cohort study of individuals prescribed OAT between 2011 and 2020 in Scotland. Prescribing records were linked to mortality data and other healthcare datasets (sociodemographic, comorbidity). We identified episodes of treatment with gabapentinoids/Z-drugs and used multivariable quasi-Poisson regression to model associations between co-prescription and DRD risk. ResultsAmong 46,602 individuals with 304,783 person-years of follow-up, we found that co-prescription was common, with 25 % and 34 % ever being co-prescribed gabapentinoids and Z-drugs, respectively. Gabapentinoid exposure was strongly associated (adjusted hazard ratio (aHR)=2·18, 95 % CI=1·92, 2·46) and Z-drug exposure moderately associated (aHR=1·39, 95 % CI=1·15, 1·66) with elevated risk of DRD. Gabapentinoid exposure was associated with DRD risk on and off OAT; Z-drug exposure was less strongly associated with DRD risk when on OAT. ConclusionsCo-prescription of gabapentinoids and Z-drugs is common among OAT patients. However, co-prescription is associated with increased risk of DRD. Alternatives to prescribing sedative medications to OAT patients and/or greater monitoring – if prescribed – are needed.

Real-life effectiveness of sacubitril/valsartan in older Belgians with heart failure, reduced ejection fraction and most severe symptoms

We assessed the real-world effectiveness of sacubitril/valsartan in patients with chronic heart failure (HF) and reduced ejection fraction (HFrEF) with an emphasis on those with older age (≥ 75 years) or with New York Heart Association (NYHA) class IV, for whom greater uncertainty existed regarding clinical outcomes. We conducted a retrospective cohort study based on patient-level linkage of electronic healthcare datasets. Data from all adults with HFrEF in Belgium receiving a prescription for sacubitril/valsartan between 01-November-2016 and 31-December-2018 were collected, with a follow-up of > 6 years. The total study population comprised 5446 patients, older than the PARADIGM-HF trial participants, and with higher NYHA class (all P < 0.0001). NYHA class improved following sacubitril/valsartan initiation (P < 0.0001 baseline vs. reassessment). Most concomitant medications were reduced. Remarkably, the risk of hospitalization for a cardiovascular reason and for HF was reduced by > 26% in the overall cohort, and in subgroups of patients ≥ 75 years, with NYHA class III/IV (all P < 0.0001) or with NYHA class IV (P < 0.05), vs. baseline. All-cause mortality did not increase in real-world patients with NYHA class III/IV. The results support the long-term beneficial effects of sacubitril/valsartan in older patients and in those experiencing the most severe symptoms.

The Evolving Role of Artificial Intelligence in Radiotherapy Treatment Planning—A Literature Review

This paper examines the integration of artificial intelligence (AI) in radiotherapy for cancer treatment. The importance of radiotherapy in cancer management and its time-intensive planning process make AI adoption appealing especially with the escalating demand for radiotherapy.This review highlights the efficacy of AI across medical domains, where it surpasses human capabilities in areas such as cardiology and dermatology. Focusing on radiotherapy, the paper details AI's applications in target segmentation, dose optimization, and outcome prediction. It discusses adaptive radiotherapy's benefits and AI's potential to enhance patient outcomes with much improved treatment accuracy. The paper explores ethical concerns, including data privacy and bias, stressing the need for robust guidelines. Educating healthcare professionals and patients about AI's role is crucial as it acknowledges potential job-role changes and concerns about patients' trust in the use of AI.Overall, the integration of AI in radiotherapy holds transformative potential in streamlining processes, improving outcomes, and reducing costs. AI's potential to reduce healthcare costs underscores its significance with impactful change globally. However, successful implementation hinges on addressing ethical and logistical challenges and fostering collaboration among healthcare professionals and patient population data sets for its optimal utilization. Rigorous education, collaborative efforts, and global data sharing will be the compass guiding its' success in radiotherapy and healthcare.