Event Prediction Model Research Articles

Abstract 17348: Cardiovascular Risk Prediction in Diabetes From Machine Learning: The ACCORD Study

Background: Atherosclerotic cardiovascular disease (ASCVD) risk prediction in persons with type 2 diabetes (T2DM) using existing calculators is imprecise. We aimed to develop a machine-learning (ML) model for prediction of ASCVD events in adults with T2DM. Methods: We utilized subjects with T2DM from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial without known CVD and calculated their 10-year ASCVD risk using the ACC/AHA pooled cohort risk calculator (PCRC) predicting the composite outcome of myocardial infarction, non-fatal stroke and cardiovascular death using age, gender, race, systolic blood pressure (SBP), antihypertensive medication use, total cholesterol, high-density lipoprotein cholesterol, current smoking status and diabetes mellitus status. We developed an ASCVD risk calculator based on Random Forest (RF) ML algorithms using follow-up data from ACCORD with the same 9 predictors. 5-fold stratified random split was applied as cross-validation strategies. Results: A total of 6581 T2DM participants without baseline ASCVD were included in our final sample with a median follow up of 9.1 years. The performance of PCRC was modest with an AUC=0.604. In contrast, the ML model had much better performance with a RF AUC=0.866. The figure shows the rank of feature importance (%) from random forest modeling (from high to low): age, systolic blood pressure, total cholesterol, HDL-C, female gender, White ethnicity, current smoker, hypertension treatment. Conclusion: The ML ASCVD Risk Calculator outperforms the AHA/ACC PCRC in predicting ASCVD outcomes among those with T2DM from the ACCORD trial. Age, SBP, total cholesterol and HDL-C were the most important features in ASCVD prediction among those with T2DM. Future studies need to validate these and other ML algorithms and to explore their applicability in guidelines.

Abstract 15495: Interleukin 18 Binding Protein Predicts Future Cardiovascular Morbidity and Mortality in Subjects Undergoing Coronary Angiography - the Casablanca Cohort

Introduction: Increased systemic inflammation is one of the key triggers in the initiation and progression of atherosclerosis. We investigated interleukin 18 binding protein (IL-18BP), a new inflammatory marker of the IL-1 cytokine family, on its predictive value for cardiovascular disease in subjects under investigation for stenotic atherosclerotic disease. Methods: We performed a prospective, single-center, investigator-initiated, observational cohort study in 927 subjects who underwent coronary and/or peripheral angiography. The cohort was randomly split into a discovery (n=627) and a validation set (n =267). Predictive value for incident cardiovascular events (composite of cardiovascular death, myocardial infarction and stroke), cardiovascular mortality and all-cause mortality was investigated using Cox proportional hazard models, including concordance index analyses and assessment of net reclassification for 10-year cardiovascular risk categories using Weibull modelling. Results: An one standard deviation increase in log-transformed IL-18BP was strongly associated with cardiovascular events (HR 1.96, 95% CI 1.54-2.51), cardiovascular mortality (HR 1.59, 95% CI 1.04-2.44) and all-cause mortality (HR 1.72, 95 CI 1.43-2.07), independent of the traditional cardiovascular risk factors, including body mass index. C-index of a clinical model for prediction of cardiovascular events improved from 0.69 (95% CI to 0.62-0.75) to 0.73 (95% CI 0.67-0.79) with inclusion of IL-18BP in the model; in doing so, a reclassification of 35.9% (p < 0.001) for events was demonstrated. Conclusions: Among a population of patient undergoing coronary angiography, results of IL18-BP measurement improved cardiovascular risk prediction beyond the traditional risk factors. Kaplan-Meier survival curves for CV events during one-year follow up comparing tertiles IL-18BP (discovery set)

Machine Learning Models for Risk Prediction of Adverse Events in Hyperkalemic Patients

Machine Learning Models for Risk Prediction of Adverse Events in Hyperkalemic Patients

Systematic review on event prediction models

Predicting an event can comes as a huge advantage for a lot of people, especially if it involves profits that a company can make. In this paper, matters that were discussed were the algorithms used for the event prediction systems, the type of algorithms and the comparison between different types of algorithm. Two main categories that were compared was the single algorithm for the prediction model and hybrid algorithms which is a combination of two or more algorithms in the model. Hybrid algorithm shows signs of better efficiency when data used were complex and large but a smaller dataset is more suitable with a single algorithm as the base as it is easier to apply with satisfactory results.

The role of pharmacogenomics in the personalization of Parkinson's disease treatment.

Parkinson's disease (PD)-related phenotypes can vary among patients substantially, including response to dopaminergic treatment in terms of efficacy and occurrence of adverse events. Many pharmacogenetic studies have already been conducted to find genetic markers of response to dopaminergic treatment. Integration of genetic and clinical data hasalready resulted in construction of clinical pharmacogenetic models for prediction of adverse events. However, the results of pharmacogenetic studies are inconsistent. More comprehensive genome-wide approaches are needed to find genetic biomarkers of PD-related phenotypes to better explain the variability in response to treatment. These genetic markers should be integrated with clinical, environmental, imaging, and other omics data to build clinically useful algorithms for personalization of PD management.

Event prediction based on evolutionary event ontology knowledge

The evolution and development of breaking news events usually present regular patterns, leading to the happening of sequential events. Therefore, the analysis of such evolutionary patterns among events and prediction to breaking news events from free text is a valuable capability for decision support systems. Traditional systems tend to focus on contents distribution information but ignore the inherent regularity of evolutionary events. We introduce evolutionary event ontology knowledge (EEOK) structuring the evolutionary patterns in five different event domains, namely Explosion, Conflagration, Geological Hazard, Traffic Accident, Personal Injury. Based on EEOK which provides a representing general-purpose ontology knowledge, we also explore a framework with a pipeline semantic analysis procedure of event extraction, evolutionary event recognition, and event prediction. Since the evolutionary event under each event domain has different evolution patterns, our proposed event prediction model combines the event types to capture the inherent regulation of evolutionary events. Comparative analyses are presented to show the effectiveness of the proposed prediction model compared to other alternative methods.

A Novel Model for Prediction of Thromboembolic and Cardiovascular Events in Patients Without Atrial Fibrillation

Patients without atrial fibrillation (AF) constitute approximately 75% of patients suffering thromboembolism and major adverse cardiovascular events (MACE), but evidence supporting risk stratification in these patients is sparse. We aimed to develop a risk prediction model for identification of patients without AF at high risk of first-time thromboembolic events. We included 72,381 coronary angiography patients without AF and without previous ischemic stroke or transient ischemic attack. The cohort was randomly divided into a derivation cohort (80%, n = 57,680) and a validation cohort (20%, n = 14,701). The primary thromboembolic end point was a composite of ischemic stroke, transient ischemic attack, and systemic embolism. MACE was defined as a composite of cardiac death, myocardial infarction, and ischemic stroke. The final model was compared with 2 validated clinical risk models (CHADS2 and CHA2DS2-VASc). The risk prediction model assigned 1 point to heart failure, hypertension, diabetes mellitus, renal disease, age 65 to 74 years, active smoking, and multivessel obstructive coronary artery disease, and 2 points to age ≥75 years and peripheral artery disease. A C-index of 0.66 (95% CI 0.64 to 0.69) for prediction of the composite thromboembolic end point was found in the validation cohort, which was higher than for CHADS2 (C-index 0.63 [95% CI 0.60 to 0.67]; p < 0.001) and CHA2DS2-VASc (C-index 0.64 [95% CI 0.62 to 0.67]; p = 0.034). The model also predicted MACE (C-index 0.71 [95% CI 0.69 to 0.73]). In conclusion it is possible to identify patients without AF at high risk of first-time thromboembolic events and MACE by use of a simple clinical prediction model.

Hierarchical Gated Recurrent Unit with Semantic Attention for Event Prediction

Event prediction plays an important role in financial risk assessment and disaster warning, which can help government decision-making and economic investment. Previous works are mainly based on time series for event prediction such as statistical language model and recurrent neural network, while ignoring the impact of prior knowledge on event prediction. This makes the direction of event prediction often biased or wrong. In this paper, we propose a hierarchical event prediction model based on time series and prior knowledge. To ensure the accuracy of the event prediction, the model obtains the time-based event information and prior knowledge of events by Gated Recurrent Unit and Associated Link Network respectively. The semantic selective attention mechanism is used to fuse the time-based event information and prior knowledge, and finally generate predicted events. Experimental results on Chinese News datasets demonstrate that our model significantly outperforms the state-of-the-art methods, and increases the accuracy by 2.8%.

Leveraging fine-grained transaction data for customer life event predictions

This real-world study with a large European financial services provider combines aggregated customer data including customer demographics, behavior and contact with the firm, with fine-grained transaction data to predict four different customer life events: moving, birth of a child, new relationship, and end of a relationship. The fine-grained transaction data—approximately 60 million debit transactions involving around 132,000 customers to >1.5 million different counterparties over a one-year period—reveal a pseudo-social network that supports the derivation of behavioral similarity measures. To advance decision support systems literature, this study validates the proposed customer life event prediction model in a real-world setting in the financial services industry; compares models that rely on aggregated data, fine-grained transaction data, and their combination; and extends existing methods to incorporate fine-grained data that preserve recency, frequency, and monetary value information of the transactions. The results show that the proposed model predicts life events significantly better than random guessing, especially with the combination of fine-grained transaction and aggregated data. Incorporating recency, frequency, and monetary value information of fine-grained transaction data also significantly improves performance compared with models based on binary logs. Fine-grained transaction data accounts for the largest part of the total variable importance, for all but one of the life events.

Popularity Prediction for Single Tweet based on Heterogeneous Bass Model

Predicting the popularity of a single tweet is useful for both users and enterprises. However, adopting existing topic or event prediction models cannot obtain satisfactory results. The reason is that one topic or event that consists of multiple tweets, has more features and characteristics than a single tweet. In this article, we propose two variations of Heterogeneous Bass models ( HBass ), originally developed in the field of marketing science, namely Spatial-Temporal Heterogeneous Bass Model ( ST-HBass ) and Feature-Driven Heterogeneous Bass Model ( FD-HBass ), to predict the popularity of a single tweet at the early stage and the stable stage. We further design an Interaction Enhancement to improve the performance, which considers the competition and cooperation from different tweets with the common topic. In addition, it is often difficult to depict popularity quantitatively. We design an experiment to get the weight of favorite, retweet and reply, and apply the linear regression to calculate the popularity. Furthermore, we design a clustering method to bound the popular threshold. Once the weight and popular threshold are determined, the status whether a tweet will be popular or not can be justified. Our model is validated by conducting experiments on real-world Twitter data, and the results show the efficiency and accuracy of our model, with less absolute percent error and the best Precision and F-score . In all, we introduce Bass model into social network single-tweet prediction to show it can achieve excellent performance.

84. Evaluation of the NHSN Standardized Infection Ratio (SIR) Risk Adjustment for HO-CDI in Oncology and ICU Patients in General Acute Care Hospitals

BackgroundThe NHSN healthcare-facility onset Clostridioides difficile infection (CDI) standardized infection ratio (SIR) is used to compare hospital quality and set hospital reimbursement but inadequate risk adjustment could penalize hospitals unnecessarily. We hypothesized that general hospitals with large oncology and/or ICU populations were not fully adjusted in the 2015 NHSN acute care hospital CDI Laboratory-Identified (LabID) event prediction model and SIRs would be affected.MethodsWe validated a negative binomial regression HO-CDI event prediction model identical to the 2015 published model and used FY2016 data from eight general hospitals in California to test our hypothesis. We compared HO-CDI events and SIR values, with and without oncology/hematopoietic stem cell transplant or ICU unit events, patient-days, admissions, bed counts, and adjustment parameters included.ResultsSeven major teaching and one nonteaching general acute care hospitals were included (see Table). Eight had oncology/hematopoietic stem cell transplant units; seven had ≥43 ICU beds (median: 134; interquartile range [IQR]: 84–161). The median facility unmodified FacWideIn SIR was 1.23 [IQR: 1.15, 1.29]. Removal of oncology unit data resulted in a 15% median facility decrease in HO-CDI events (IQR: 14%, 21%) and −8% median facility decrease in SIR (IQR: −2%, −14%). Removal of ICU unit data resulted in a 22% median facility decrease in HO-CDI events (IQR: 16%, 26%) and 97% median facility increase in SIR at each facility (IQR: 78%, 105%).ConclusionThe ICU bed adjustment in the 2015 NHSN SIR is a powerful correction that fully adjusted for ICU HO-CDI events at all hospitals in the study. However, the lack of risk adjustment for oncology/hematopoietic stem cell transplant unit HO-CDI events suggests that the current model unfairly penalizes general acute facilities, many of which also provide specialized oncologic care. Thus, the model needs to be re-adjusted to account for this important specialty care population in general acute care facilities. Disclosures All Authors: No reported Disclosures.

29Prognostic safety of automatic cancellation of rest myocardial perfusion scan by machine learning: a report from multicenter REFINE SPECT registry of new generation SPECT

Abstract Background We aimed to develop a machine learning (ML) computer score derived from stress imaging and clinical data, which indicates if the rest scan could be automatically and safely canceled in the routine stress/rest myocardial perfusion SPECT (MPS). Methods A total of 20414 stress/rest cases from the REFINE SPECT registry collected from 5 sites in 3 countries with Tc-99m-based MPS images, clinical data, and clinical follow-up were included in the study. All images were automatically processed at our Medical Center. The automatically generated myocardial contours were checked by experienced technologists. In total, 93 variables (26 clinical, 17 stress-test, and 50 stress-imaging variables) were used to build a LogitBoost model for prediction of adverse events (AE), including coronary revascularization, death, myocardial infarction, and unstable angina. 10-fold cross-validation was performed to separate test from validation data for the assessment of ML. The overall ML predictive performance was compared to quantitative (stress total perfusion deficit [TPD]) by the area under the receiver operating characteristic curves (AUC). ML cut-off (ML1) to simulate the decision of cancellation of the rest scan was set to result in the same % of normal scans as these determined by the normal clinical reader diagnosis on a 4-point scale in the whole population, or the same % of scans with visual summed stress scores (SSS) = 0 in the subpopulation with available SSS. A second ML cutoff (ML2) was established to achieve a 1% annual risk of AE. The annual risk of AE of the normal ML score was compared with normal clinical diagnosis and with the finding of SSS = 0. Results The mean follow-up interval was 4.7±1.5 years. Overall, 3542 AE were observed (3.7% annual risk). The AUC for AE was higher for ML (0.780±0.005) than for stress TPD (0.698±0.006) (p&lt;0.001). Normal clinical diagnosis was reported in 60% cases. In 70% (14242 scans) with available segmental scores, 53% had SSS=0. ML1 and ML2 thresholds were compared with normal visual diagnosis and with SSS = 0 for AE (Figure). ML1 achieved a lower annual risk (1.5%) than normal clinical diagnosis (2.1%) or SSS = 0 (1.6% versus 2.3%) (p&lt;0.001). The more conservative ML2 threshold with a 1% annual risk of AE resulted in a 40% canceling rate. Figure 1 Conclusion ML could be used to automatically cancel the rest MPS scan with the same proportion as using normal visual MPS reading, but with significantly lower AE rate in stress-only scans. Acknowledgement/Funding R01HL089765 from the National Heart, Lung, and Blood Institute/National Institutes of Health (NHLBI/NIH)

Estimating individual lifetime benefit and bleeding risk of adding rivaroxaban to aspirin for patients with stable cardiovascular disease: results from the COMPASS trial.

Adding rivaroxaban to aspirin in patients with stable atherosclerotic disease reduces the recurrence of cardiovascular disease (CVD) but increases the risk of major bleeding. The aim of this study was to estimate the individual lifetime treatment benefit and harm of adding low-dose rivaroxaban to aspirin in patients with stable cardiovascular disease. Patients with established CVD from the COMPASS trial (n = 27390) and SMART prospective cohort study (n = 8139) were used. Using the pre-existing lifetime SMART-REACH model for recurrent CVD, and a newly developed Fine and Gray competing risk-adjusted lifetime model for major bleeding, individual treatment effects from adding low-dose rivaroxaban to aspirin in patients with stable CVD were estimated, expressed in terms of (i) life-years free of stroke or myocardial infarction (MI) gained; and (ii) life-years free from major bleeding lost. Calibration of the SMART-REACH model for prediction of recurrent CVD events in the COMPASS study was good. The major bleeding risk model as derived in the COMPASS trial showed good external calibration in the SMART cohort. Predicted individual gain in life expectancy free of stroke or MI from added low-dose rivaroxaban had a median of 16 months (range 1-48 months), while predicted individualized lifetime lost in terms of major bleeding had a median of 2 months (range 0-20 months). There is a wide distribution in lifetime gain and harm from adding low-dose rivaroxaban to aspirin in individual patients with stable CVD. Using these lifetime models, benefits and bleeding risk can be weighed for each individual patient, which could facilitate treatment decisions in clinical practice.

Cascading detection model for prediction of apnea-hypopnea events based on nasal flow and arterial blood oxygen saturation

PurposeSleep apnea and hypopnea syndrome (SAHS) seriously affects sleep quality. In recent years, much research has focused on the detection of SAHS using various physiological signals and algorithms. The purpose of this study is to find an efficient model for detection of apnea-hypopnea events based on nasal flow and SpO2 signals.MethodsA 60-s detector and a 10-s detector were cascaded for precise detection of apnea-hypopnea (AH) events. Random forests were adopted for classification of data segments based on morphological features extracted from nasal flow and arterial blood oxygen saturation (SpO2). Then the segments’ classification results were fed into an event detector to locate the start and end time of every AH event and predict the AH index (AHI).ResultsA retrospective study of 24 subjects’ polysomnography recordings was conducted. According to segment analysis, the cascading detection model reached an accuracy of 88.3%. While Pearson’s correlation coefficient between estimated AHI and reference AHI was 0.99, in the diagnosis of SAHS severity, the proposed method exhibited a performance with Cohen’s kappa coefficient of 0.76.ConclusionsThe cascading detection model is able to detect AH events and provide an estimate of AHI. The results indicate that it has the potential to be a useful tool for SAHS diagnosis.

Urban events prediction via convolutional neural networks and Instagram data

In today’s world, it is crucial to be proactive and be prepared for events which are not happening yet. Thus, there is no surprise that in the field of social media analysis the research agenda has moved from the development of event detection methods to a brand new area - event prediction models. This research field is extremely important for all sorts of applications, from natural disasters preparation and criminal activity prevention to urban management and development of smart cities. However, even the leading models have an important disadvantage: they are based on prior knowledge about events being expected. So forecasting systems based on such models are heavily limited by a list of events that can be predicted and all events of other types will be out of systems’ scope. In this work, we try to address this issue and propose a deep learning model, which is able to predict an area of the future event in the urban environment. This model is able to predict the future state of the city - a level of users activity in the location-based social network Instagram - with the average deviation from the ground truth of 1%, and achieves 69% recall when solving the events prediction problem.

POPULARITY PREDICTION FOR SINGLE TWEET BASED ON HETEROGENEOUS BASS MODEL

Predicting the popularity of a single tweet is useful for both users and enterprises. However, adopting existing topic or event prediction models cannot obtain satisfactory results. The reason is that one topic or event that consists of multiple tweets, has more features and characteristics than a single tweet. In this paper, we propose two variations of Heterogeneous Bass models (HBass), originally developed in the field of marketing science, namely Spatial-Temporal Heterogeneous Bass Model (ST-HBass) and Feature-Driven Heterogeneous Bass Model (FDHBass), to predict the popularity of a single tweet at the early stage and the stable stage. We further design an Interaction Enhancement to improve the performance, which considers the competition and cooperation from different tweets with the common topic. In addition, it is often difficult to depict popularity quantitatively. We design an experiment to get the weight of favorite, retweet and reply, and apply the linear regression to calculate the popularity. Furthermore, we design a clustering method to bound the popular threshold. Once the weight and popular threshold are determined, the status whether a tweet will be popular or not can be justified. Our model is validated by conducting experiments on real-world Twitter data, and the results show the efficiency and accuracy of our model, with less absolute percent error and the best Precision and F-score. In all, we introduce Bass model into social network single-tweet prediction to show it can achieve excellent performance.

A biomarker of collagen type I degradation is associated with cardiovascular events and mortality in patients with atherosclerosis.

Atherosclerosis is characterized by accumulation of lipids, cells and extracellular matrix (ECM) proteins in the arterial wall. Collagen type I (COL1), a component of the arterial ECM, is cleaved by matrix metalloproteinases (MMPs) and known to be remodelled in atherosclerosis. We explored whether the MMP-mediated COL1 biomarker, C1M, was associated with cardiovascular events, cardiovascular mortality and all-cause mortality in a large prospective cohort of patients with known atherosclerosis. Serum from 787 patients who underwent a carotid endarterectomy was included. Circulating levels of C1M were measured in serum. A total of 473 patients were followed for 6years after surgery. Associations between C1M and incidence of cardiovascular events, cardiovascular mortality and all-cause mortality were assessed by Kaplan-Meier curves and Cox regression analysis. A total of 101 (21.4%) patients suffered from nonfatal cardiovascular events during the follow-up period, and 64 (13.5%) patients died. Of these, 39 (60.9%) died from cardiovascular diseases. Patients with C1M levels above the median were significantly associated with cardiovascular events, cardiovascular mortality and all-cause mortality (P<0.001, P=0.004 and P<0.001, respectively). C1M was included in the final model for prediction of cardiovascular events (HR 2.15, 95% CI 1.40-3.32, P=0.001), cardiovascular mortality (HR 2.20, 95% CI 1.07-4.51, P=0.031) and all-cause mortality (HR 2.98 95% CI 1.67-5.33, P=<0.001). In patients with atherosclerotic carotid lesions, high levels of C1M predicted cardiovascular events, cardiovascular mortality and all-cause mortality. These findings emphasize the importance of remodelling mechanisms in atherosclerosis that are now becoming more and more explored.

Space-based observations of crustal deflections for drought characterization in Brazil

Widespread environmental impacts of frequent drought episodes in Brazil have resulted in several drought-related diagnostics studies. However, the potential of many “opportunistic sensors”, such as the Global Positioning System (GPS), has not yet been considered in hydrological hazard monitoring in Brazil. In this study, the response of the Earth's crust to Brazil's 2012–2015 drought event in different structural provinces is analyzed by comparing GPS-observed vertical crustal deformations (VCDs) with the terrestrial water storage (TWS) derived from the Gravity Recovery and Climate Experiment (GRACE). The results indicate that there is no spatial correlation between annual amplitudes of the TWS and VCDs in different structural provinces apart from the purely elastic response of the crust to TWS dynamics, at almost all the 39 GPS stations that were analyzed. However, approximately 15% of the monitoring stations show that VCD leads TWS with a phase lag of 2–4 months. Errors associated with VCD and TWS are within the accepted range for space geodetic techniques (i.e., GPS and GRACE) and despite the need for further investigation, the phase lead seems to be associated with rainfall, which impacts the TWS through the hydrographs. Overall, the GPS-based drought index (DIVCD) reflects the water depletion in many regions of Brazil, which agrees with the GRACE-based DITWS in terms of the Spearman correlation coefficient (ranging from 0.4 to 0.9) in the Amazon, Tocantins, La Plata, and São Francisco river basins. This agreement confirms the drought persistence during the study period and that DIVCD can be used to monitor hydrological droughts. In regions in which DITWS sufficiently agrees with DIVCD (48% of the sites), near real-time drought monitoring is feasible. This could be useful in the optimization of models for the forward prediction of drought events in other regions worldwide, where GPS vertical displacements strongly correlate with hydrological GRACE signals.

PREDICTIVE PERFORMANCE AND CLINICAL UTILITY OF A NEW PREDICTIVE MODEL OF CARDIOVASCULAR RISK FOR YOUNG AND MIDDLE-AGED WORKING POPULATION

Objective: Identifying people at high cardiovascular risk is challenging, especially at young age. The main objective of the IberScore study was to derive a mathematical model for risk prediction of fatal and non-fatal cardiovascular events from a relatively young and healthy working population. Design and method: A predictive function for fatal and non-fatal CV events was derived from a cohort of 774,404 workers (70.4% of the target population), free of CV disease at entry, who were followed during 10 years. Workers ages ranged 16–65 years (mean 35.7, SD 10.7). 71.7% were men, which represented the real proportion in the target population. Cardiovascular ageing was estimated (independently by genders) based on the effects of well-known cardiovascular risk factors as: smoking; cholesterol; blood glucose; HDL; systolic blood pressure (SBP); obesity; history of dyslipidemia, diabetes and hypertension. Results: Along the 10-year follow-up we found 3,762 first cardiovascular events (6‰) in derivation cohort. Most of them (80.3%) were non-fatal ischemic events. We derived a logistic flexible parametric model to predict 10-year cardiovascular risk. 82% of those who suffered a cardiovascular event during the follow-up span had been previously classified as “high risk” or “very high risk” using our model, whereas only 12% of them were classified in the same groups using SCORE. The latter also showed a weak discrimination power for risk stratification while IberScore clearly distinguished the four risk categories. IberScore was well calibrated and showed outstanding predictive performance and clinical utility even at young and middle-age workers. Cut-off points were fixed taking on account their discriminatory capacity and the balance between costs and benefits of the treatments that will be prescribed based on them. Conclusions: IberScore worked much better to estimate cardiovascular risk in a relatively young and healthy Spanish working population when compared to other models. Cardiovascular aging, as the result of the effects of risk factors, should be at the core of CV risk estimation. Cut-off points should be set considering the benefit we seek with the treatments we have in mind.

Dynamic Prediction of Motor Diagnosis inHuntington's Disease Using a Joint Modeling Approach.

Prediction of motor diagnosis in Huntington's disease (HD) can be improved by incorporating other phenotypic and biological clinical measures in addition to cytosine-adenine-guanine (CAG) repeat length and age. The objective was to compare various clinical and biomarker trajectories for tracking HD progression and predicting motor conversion. Participants were from the PREDICT-HD study. We constructed a mixed-effect model to describe the change of measures while jointly modeling the process with time to HD diagnosis. The model was then used for subject-specific prediction. We employed the time-dependent receiver operating characteristic (ROC) method to assess the discriminating capability of the measures to identify high and low risk patients. The strongest predictor was used to illustrate the dynamic prediction of the disease risk and future trajectories of biomarkers for three hypothetical patients. 1078 individuals were included in this analysis. Five longitudinal clinical and imaging measures were compared. The putamen volume had the best discrimination performance with area under the curve (AUC) ranging from 0.74 to 0.82 over time. The total motor score showed a comparable discriminative ability with AUC ranging from 0.69 to 0.78 over time. The model showed that decreasing putamen volume was a significant predictor of motor conversion. A web-based calculator was developed for implementing the methods. By jointly modeling longitudinal data with time-to-event outcomes, it is possible to construct an individualized dynamic event prediction model that renews over time with accumulating evidence. If validated, this could be a valuable tool to guide the clinician in predicting age of onset and potentially rate of progression.