1,720,965 research outputs found
Interpretability of time-series deep learning models: A study in cardiovascular patients admitted to Intensive care unit
Full text linkInterpretability is fundamental in healthcare problems and the lack of it in deep learning models is currently the major barrier in the usage of such powerful algorithms in the field. The study describes the implementation of an attention layer for Long Short-Term Memory (LSTM) neural network that provides a useful picture on the influence of the several input variables included in the model.
A cohort of 10,616 patients with cardiovascular diseases is selected from the MIMIC III dataset, an openly available database of electronic health records (EHRs) including all patients admitted to an ICU at Boston’s Medical Centre. For each patient, we consider a 10-length sequence of 1-hour windows in which 48 clinical parameters are extracted to predict the occurrence of death in the next 7 days. Inspired from the recent developments in the field of attention mechanisms for sequential data, we implement a recurrent neural network with LSTM cells incorporating an attention mechanism to identify features driving model’s decisions over time.
The performance of the LSTM model, measured in terms of AUC, is 0.790 (SD = 0.015). Regard our primary objective, i.e. model interpretability, we investigate the role of attention weights. We find good correspondence with driving predictors of a transparent model (r = 0.611, 95% CI [0.395, 0.763]). Moreover, most influential features identified at the cohort-level emerge as known risk factors in the clinical context.
Despite the limitations of study dataset, this work brings further evidence of the potential of attention mechanisms in making deep learning model more interpretable and suggests the application of this strategy for the sequential analysis of EHRs
Wavelet‐Mixed Landmark Survival Models for the Effect of Short‐Term Changes of Potassium in Heart Failure Patients
Full text linkStatistical methods to study the association between a longitudinal biomarker and the risk of death are very relevant for the long-term care of subjects affected by chronic illnesses, such as potassium in heart failure patients. Particularly in the presence of comorbidities or pharmacological treatments, sudden crises can cause potassium to undergo very abrupt yet transient changes. In the context of the monitoring of potassium, there is a need for a dynamic model that can be used in clinical practice to assess the risk of death related to an observed patient's potassium trajectory. We considered different landmark survival approaches, starting from the simple approach considering the most recent measurement. We then propose a novel method based on wavelet filtering and landmarking to retrieve the prognostic role of past short-term potassium shifts. We argue that while taking into account the smooth changes in the biomarker, short-term changes cannot be overlooked. State-of-the-art dynamic survival models are prone to give more importance to the smooth component of the potassium profiles. However, our findings suggest that it is essential to also take into account recent potassium instability to capture all the relevant prognostic information. The data used comes from over 2000 subjects, with a total of over 80,000 repeated potassium measurements collected through administrative health records. The proposed wavelet landmark method revealed the prognostic role of past short-term changes in potassium. We also performed a simulation study to assess how and when to apply the proposed wavelet-mixed landmark model
HF progression among outpatients with HF in a community setting
No full textBackground: Incidence and prognostic impact of heart failure (HF) progression has been not well addressed.
Methods: From 2009 until 2015, consecutive ambulatory HF patientswere recruited. HF progressionwas defined by the presence of at least two of the following criteria: step up of ≥1 New York Heart Association (NYHA) class;
decrease LVEF ≥ 10 points; association of diuretics or increase ≥ 50% of furosemide dosage, or HF hospitalization.
Results: 2528 met study criteria (mean age 76; 42% women). Of these, 48% had ischemic heart disease, 18% patients with LVEF ≤ 35%. During a median follow-up of 2.4 years, overall mortality was 31% (95% CI: 29%–33%),
whereas rate of HF progression or death was 57% (95% CI: 55%–59%). The 4-year incidence of HF progression was 39% (95% CI: 37%–41%) whereas the competing mortality rate was 18% (95% CI: 16%–19%). Rates of HF progression and death were higher in HF patients with LVEF ≤ 35% vs N35% (HF progression: 42% vs 38%, p=0.012; death as a competing risk: 22% vs 17%, p = 0.002). HF progression identified HF patients with a worse survival (HR = 3.16, 95% CI: 2.75–3.72). In cause-specific Cox models, age, previous HF hospitalization, chronic obstructive pulmonary disease, chronic kidney disease, anemia, sex, LVEF ≤ 35% emerged as prognostic factors
of HF progression.
Conclusions: Among outpatients with HF, at 4 years 39% presented a HF progression, while 18% died before any sign of HF progression. This trend was higher in patients with LVEF ≤ 35%. These findings may have implications
for healthcare planning and resource allocation
Flexible approaches based on multi-state models and microsimulation to perform real-world cost-effectiveness analyses: an application to pcsk9-inhibitors
Full text linkObjectives: This study aims to show the application of flexible statistical methods in real-world cost-effectiveness analyses applied in the cardiovascular field, focusing specifically on the use of PCSK9 inhibitors for hyperlipidaemia. Methods: The proposed method allowed us to use an electronic health database to emulate a target trial for cost-effectiveness analysis using multi-state modelling and microsimulation. We formally established the study design and provided precise definitions of the causal measures of interest, while also outlining the assumptions necessary for accurately estimating these measures using the available data. Additionally, we thoroughly considered goodness-of-fit assessments and sensitivity analyses of the decision model, which are crucial to capture the complexity of individuals' healthcare pathway and to enhance the validity of this type of health economic models. Results: In the disease model, the Markov assumption was found to be inadequate, and a "time-reset" timescale was implemented together with the use of a time-dependent variable to incorporate past hospitalization history. Furthermore, the microsimulation decision model demonstrated a satisfying goodness-of-fit, as evidenced by the consistent results obtained in the short-term horizon compared to a non-model-based approach. Notably, only in the long-term follow-up PCSK9 inhibitors revealed their favorable cost-effectiveness, with a minimum willingness-to-pay of 39,000 Euro/LY gained. Conclusions: The approach demonstrated its significant utility in several ways. Unlike non-model based or alternative model-based methods, it enabled to 1) investigate long-term cost-effectiveness comprehensively, 2) employ an appropriate disease model that aligns with the specific problem under study, and 3) conduct subgroup-specific cost-effectiveness analyses to gain more targeted insights
Comparison of discrimination and calibration performance of ECG-based machine learning models for prediction of new-onset atrial fibrillation
Full text linkAbstract Background Machine learning (ML) methods to build prediction models starting from electrocardiogram (ECG) signals are an emerging research field. The aim of the present study is to investigate the performances of two ML approaches based on ECGs for the prediction of new-onset atrial fibrillation (AF), in terms of discrimination, calibration and sample size dependence. Methods We trained two models to predict new-onset AF: a convolutional neural network (CNN), that takes as input the raw ECG signals, and an eXtreme Gradient Boosting model (XGB), that uses the signal’s extracted features. A penalized logistic regression model (LR) was used as a benchmark. Discrimination was evaluated with the area under the ROC curve, while calibration with the integrated calibration index. We investigated the dependence of models’ performances on the sample size and on class imbalance corrections introduced with random under-sampling. Results CNN's discrimination was the most affected by the sample size, outperforming XGB and LR only around n = 10.000 observations. Calibration showed only a small dependence on the sample size for all the models considered. Balancing the training set with random undersampling did not improve discrimination in any of the models. Instead, the main effect of imbalance corrections was to worsen the models’ calibration (for CNN, integrated calibration index from 0.014 [0.01, 0.018] to 0.17 [0.16, 0.19]). The sample size emerged as a fundamental point for developing the CNN model, especially in terms of discrimination (AUC = 0.75 [0.73, 0.77] when n = 10.000, AUC = 0.80 [0.79, 0.81] when n = 150.000). The effect of the sample size on the other two models was weaker. Imbalance corrections led to poorly calibrated models, for all the approaches considered, reducing the clinical utility of the models. Conclusions Our results suggest that the choice of approach in the analysis of ECG should be based on the amount of data available, preferring more standard models for small datasets. Moreover, imbalance correction methods should be avoided when developing clinical prediction models, where calibration is crucial
Sex-related differences in chronic heart failure: a community-based study
No full textAIMS: To evaluate sex-related differences among real-life outpatients with chronic heart failure across the ejection fraction spectrum and to evaluate whether these differences might impact therapy and outcomes.METHODS: A total of 2528 heart failure patients were examined between 2009 and 2015 [mean age 76, 42% females; 59% with heart failure with preserved ejection fraction (HFpEF), 17% with heart failure with mid-range ejection fraction (HFmrEF) and 24% with heart failure with reduced ejection fraction (HFrEF)]. Females showed a higher prevalence of HFpEF than males.RESULTS: Females were older, less obese and with less ischaemic heart disease. They have renal failure and anaemia more frequently than males. There were no differences in terms of heart failure therapy in the HFrEF group, but a lower prescription rate of angiotensin-converting enzyme-I/AT1 blockers in HFmrEF and HFpEF and a higher prescription of mineralocorticoid receptor antagonists in the female group with HFpEF were observed. Crude rate mortality and composite outcome (death/heart failure progression) run similarly across sexes regardless of the ejection fraction categories. After adjustment, risk of mortality was significantly lower in females than males in the HFmrEF and HFpEF groups, whereas similar risk was confirmed across sexes in the HFrEF group. Considering prognostic risk factors, noncardiac comorbidities emerged in the HFpEF group.CONCLUSION: In a community-based heart failure cohort, females were differently distributed within heart failure phenotypes and they presented some different characteristics across ejection fraction categories. Although in an unadjusted model there was no significant difference for adverse outcomes, in an adjusted model females showed a lower risk of mortality in HFpEF and HFmrEF. Concerning sex-related prognostic risk factors, noncardiac comorbidities significantly affected adverse prognosis in females with HFpEF
Deep-learning-based prognostic modeling for incident heart failure in patients with diabetes using electronic health records: A retrospective cohort study.
Full text linkPatients with type 2 diabetes mellitus (T2DM) have more than twice the risk of developing heart failure (HF) compared to patients without diabetes. The present study is aimed to build an artificial intelligence (AI) prognostic model that takes in account a large and heterogeneous set of clinical factors and investigates the risk of developing HF in diabetic patients. We carried out an electronic health records- (EHR-) based retrospective cohort study that included patients with cardiological clinical evaluation and no previous diagnosis of HF. Information consists of features extracted from clinical and administrative data obtained as part of routine medical care. The primary endpoint was diagnosis of HF (during out-of-hospital clinical examination or hospitalization). We developed two prognostic models using (1) elastic net regularization for Cox proportional hazard model (COX) and (2) a deep neural network survival method (PHNN), in which a neural network was used to represent a non-linear hazard function and explainability strategies are applied to estimate the influence of predictors on the risk function. Over a median follow-up of 65 months, 17.3% of the 10,614 patients developed HF. The PHNN model outperformed COX both in terms of discrimination (c-index 0.768 vs 0.734) and calibration (2-year integrated calibration index 0.008 vs 0.018). The AI approach led to the identification of 20 predictors of different domains (age, body mass index, echocardiographic and electrocardiographic features, laboratory measurements, comorbidities, therapies) whose relationship with the predicted risk correspond to known trends in the clinical practice. Our results suggest that prognostic models for HF in diabetic patients may improve using EHRs in combination with AI techniques for survival analysis, which provide high flexibility and better performance with respect to standard approaches
Machine learning applications in cardiology
Full text linkCardiovascular diseases remain the leading cause of death globally and impose significant economic burdens. The growing prevalence of cardiovascular diseases underscores the need for advanced prevention and management strategies. Artificial intelligence, specifically with machine learning and deep learning, offers transformative potential in cardiology for a wide range of tasks. This thesis explores the application of artificial intelligence in cardiovascular care, focusing on clinical prediction models, integration of multimodal data, and the development of algorithms for specific cardiovascular conditions. Additionally, it addresses the challenges of model validation and real-world applicability, proposing rigorous methodologies for improving
artificial intelligence’s role in cardiology care
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
