Search

This paper provides guidance for researchers with some mathematical background on the conduct of time-to-event analysis in observational studies based on intensity (hazard) models. Discussions of basic concepts like time axis, event definition and censoring are given. Hazard models are introduced, with special emphasis on the Cox proportional hazards regression model. We provide check lists that may be useful both when fitting the model and assessing its goodness of fit and when interpreting the results. Special attention is paid to how to avoid problems with immortal time bias by introducing time-dependent covariates. We discuss prediction based on hazard models and difficulties when attempting to draw proper causal conclusions from such models. Finally, we present a series of examples where the methods and check lists are exemplified. Computational details and implementation using the freely available R software are documented in Supplementary Material. The paper was prepared as part of the STRATOS initiative. [ABSTRACT FROM AUTHOR]

Introduction: The opioid crisis is a pervasive public health threat in the U.S. Simulation modeling approaches that integrate a systems perspective are used to understand the complexity of this crisis and analyze what policy interventions can best address it. However, limitations in currently available data sources can hamper the quantification of these models.Methods: To understand and discuss data needs and challenges for opioid systems modeling, a meeting of federal partners, modeling teams, and data experts was held at the U.S. Food and Drug Administration in April 2019. This paper synthesizes the meeting discussions and interprets them in the context of ongoing simulation modeling work.Results: The current landscape of national-level quantitative data sources of potential use in opioid systems modeling is identified, and significant issues within data sources are discussed. Major recommendations on how to improve data sources are to: maintain close collaboration among modeling teams, enhance data collection to better fit modeling needs, focus on bridging the most crucial information gaps, engage in direct and regular interaction between modelers and data experts, and gain a clearer definition of policymakers' research questions and policy goals.Conclusions: This article provides an important step in identifying and discussing data challenges in opioid research generally and opioid systems modeling specifically. It also identifies opportunities for systems modelers and government agencies to improve opioid systems models. [ABSTRACT FROM AUTHOR]

Clinical prediction models (CPMs) can predict clinically relevant outcomes or events. Typically, prognostic CPMs are derived to predict the risk of a single future outcome. However, there are many medical applications where two or more outcomes are of interest, meaning this should be more widely reflected in CPMs so they can accurately estimate the joint risk of multiple outcomes simultaneously. A potentially naïve approach to multi-outcome risk prediction is to derive a CPM for each outcome separately, then multiply the predicted risks. This approach is only valid if the outcomes are conditionally independent given the covariates, and it fails to exploit the potential relationships between the outcomes. This paper outlines several approaches that could be used to develop CPMs for multiple binary outcomes. We consider four methods, ranging in complexity and conditional independence assumptions: namely, probabilistic classifier chain, multinomial logistic regression, multivariate logistic regression, and a Bayesian probit model. These are compared with methods that rely on conditional independence: separate univariate CPMs and stacked regression. Employing a simulation study and real-world example, we illustrate that CPMs for joint risk prediction of multiple outcomes should only be derived using methods that model the residual correlation between outcomes. In such a situation, our results suggest that probabilistic classification chains, multinomial logistic regression or the Bayesian probit model are all appropriate choices. We call into question the development of CPMs for each outcome in isolation when multiple correlated or structurally related outcomes are of interest and recommend more multivariate approaches to risk prediction. [ABSTRACT FROM AUTHOR]

Background: Even though investigating predictors of intervention success (e.g Cognitive Training, CT) is gaining more and more interest in the light of an individualized medicine, results on specific predictors of intervention success in the overall field are mixed and inconsistent due to different and sometimes inappropriate statistical methods used. Therefore, the present paper gives a guidance on the appropriate use of multiple regression analyses to identify predictors of CT and similar non-pharmacological interventions.Methods: We simulated data based on a predefined true model and ran a series of different analyses to evaluate their performance in retrieving the true model coefficients. The true model consisted of a 2 (between: experimental vs. control group) × 2 (within: pre- vs. post-treatment) design with two continuous predictors, one of which predicted the success in the intervention group and the other did not. In analyzing the data, we considered four commonly used dependent variables (post-test score, absolute change score, relative change score, residual score), five regression models, eight sample sizes, and four levels of reliability.Results: Our results indicated that a regression model including the investigated predictor, Group (experimental vs. control), pre-test score, and the interaction between the investigated predictor and the Group as predictors, and the absolute change score as the dependent variable seemed most convenient for the given experimental design. Although the pre-test score should be included as a predictor in the regression model for reasons of statistical power, its coefficient should not be interpreted because even if there is no true relationship, a negative and statistically significant regression coefficient commonly emerges.Conclusion: Employing simulation methods, theoretical reasoning, and mathematical derivations, we were able to derive recommendations regarding the analysis of data in one of the most prevalent experimental designs in research on CT and external predictors of CT success. These insights can contribute to the application of considered data analyses in future studies and facilitate cumulative knowledge gain. [ABSTRACT FROM AUTHOR]

Causal inference requires theory and prior knowledge to structure analyses, and is not usually thought of as an arena for the application of prediction modelling. However, contemporary causal inference methods, premised on counterfactual or potential outcomes approaches, often include processing steps before the final estimation step. The purposes of this paper are: (i) to overview the recent emergence of prediction underpinning steps in contemporary causal inference methods as a useful perspective on contemporary causal inference methods, and (ii) explore the role of machine learning (as one approach to 'best prediction') in causal inference. Causal inference methods covered include propensity scores, inverse probability of treatment weights (IPTWs), G computation and targeted maximum likelihood estimation (TMLE). Machine learning has been used more for propensity scores and TMLE, and there is potential for increased use in G computation and estimation of IPTWs. [ABSTRACT FROM AUTHOR]

Objective: This paper aims to describe and test novel computational driver models, predicting drivers' brake reaction times (BRTs) to different levels of lead vehicle braking, during driving with cruise control (CC) and during silent failures of adaptive cruise control (ACC).Background: Validated computational models predicting BRTs to silent failures of automation are lacking but are important for assessing the safety benefits of automated driving.Method: Two alternative models of driver response to silent ACC failures are proposed: a looming prediction model, assuming that drivers embody a generative model of ACC, and a lower gain model, assuming that drivers' arousal decreases due to monitoring of the automated system. Predictions of BRTs issued by the models were tested using a driving simulator study.Results: The driving simulator study confirmed the predictions of the models: (a) BRTs were significantly shorter with an increase in kinematic criticality, both during driving with CC and during driving with ACC; (b) BRTs were significantly delayed when driving with ACC compared with driving with CC. However, the predicted BRTs were longer than the ones observed, entailing a fitting of the models to the data from the study.Conclusion: Both the looming prediction model and the lower gain model predict well the BRTs for the ACC driving condition. However, the looming prediction model has the advantage of being able to predict average BRTs using the exact same parameters as the model fitted to the CC driving data.Application: Knowledge resulting from this research can be helpful for assessing the safety benefits of automated driving. [ABSTRACT FROM AUTHOR]

As a threat for transportation system, traffic crashes have a wide range of social consequences for governments. Traffic crashes are increasing in developing countries and Iran as a developing country is not immune from this risk. There are several researches in the literature to predict traffic crash severity based on artificial neural networks (ANNs), support vector machines and decision trees. This paper attempts to investigate the crash injury severity of rural roads by using a hybrid clustering and classification approach to compare the performance of classification algorithms before and after applying the clustering. In this paper, a novel rule-based genetic algorithm (GA) is proposed to predict crash injury severity, which is evaluated by performance criteria in comparison with classification algorithms like ANN. The results obtained from analysis of 13,673 crashes (5600 property damage, 778 fatal crashes, 4690 slight injuries and 2605 severe injuries) on rural roads in Tehran Province of Iran during 2011–2013 revealed that the proposed GA method outperforms other classification algorithms based on classification metrics like precision (86%), recall (88%) and accuracy (87%). Moreover, the proposed GA method has the highest level of interpretation, is easy to understand and provides feedback to analysts. [ABSTRACT FROM PUBLISHER]

Background: Electronic medical records contain a variety of valuable medical information for patients. So, when we are able to recognize and extract risk factors for disease from EMRs of patients with cardiovascular disease (CVD), and are able to use them to predict CVD, we have the ability to automatically process clinical texts, resulting in an improved accuracy of supporting doctors for the clinical diagnosis of CVD. In the case where CVD is becoming more worldwide, predictive CVD based on EMRs has been studied by many researchers to address this important aspect of improving diagnostic efficiency.Methods: This paper proposes an Enhanced Character-level Deep Convolutional Neural Networks (EnDCNN) model for cardiovascular disease prediction.Results: On the manually annotated Chinese EMRs corpus, our risk factor identification extraction model achieved 0.9073 of F-score, our prediction model achieved 0.9516 of F-score, and the prediction result is better than the most previous methods.Conclusions: The character-level model based on text region embedding can well map risk factors and their labels as a unit into a vector, and downsampling plays a crucial role in improving the training efficiency of deep CNN. What's more, the shortcut connections with pre-activation used in our model architecture implements dimension-matching free in training. [ABSTRACT FROM AUTHOR]

Background: The accelerated growth of elderly population is creating a heavy burden to the healthcare system in many developed countries and regions. Electrocardiogram (ECG) analysis has been recognized as effective approach to cardiovascular disease diagnosis and widely utilized for monitoring personalized health conditions.Method: In this study, we present a novel approach to forecasting one-day-forward wellness conditions for community-dwelling elderly by analyzing single lead short ECG signals acquired from a station-based monitoring device. More specifically, exponentially weighted moving-average (EWMA) method is employed to eliminate the high-frequency noise from original signals at first. Then, Fisher-Yates normalization approach is used to adjust the self-evaluated wellness score distribution since the scores among different individuals are skewed. Finally, both deep learning-based and traditional machine learning-based methods are utilized for building wellness forecasting models.Results: The experiment results show that the deep learning-based methods achieve the best fitted forecasting performance, where the forecasting accuracy and F value are 93.21% and 91.98% respectively. The deep learning-based methods, with the merit of non-hand-crafted engineering, have superior wellness forecasting performance towards the competitive traditional machine learning-based methods.Conclusion: The developed approach in this paper is effective in wellness forecasting for community-dwelling elderly, which can provide insights in terms of implementing a cost-effective approach to informing healthcare provider about health conditions of elderly in advance and taking timely interventions to reduce the risk of malignant events. [ABSTRACT FROM AUTHOR]

After the diagnosis of a disease, one major objective is to predict cumulative probabilities of events such as clinical relapse or death from the individual information collected up to a prediction time, usually including biomarker repeated measurements. Several competing estimators have been proposed, mainly from two approaches: joint modelling and landmarking. These approaches differ by the information used, the model assumptions and the complexity of the computational procedures. This paper aims to review the two approaches, precisely define the derived estimators of dynamic predictions and compare their performances notably in case of misspecification. The ultimate goal is to provide key elements for the use of individual dynamic predictions in clinical practice. Prediction of two competing causes of prostate cancer progression from the history of prostate-specific antigen is used as a motivated example. We formally define the quantity to estimate and its estimators, propose techniques to assess the uncertainty around predictions and validate them. We then conduct an in-depth simulation study compare the estimators in terms of prediction error, discriminatory power, efficiency and robustness to model assumptions. We show that prediction tools should be handled with care, in particular by properly specifying models and estimators. [ABSTRACT FROM AUTHOR]

Count outcomes with excessive zeros are common in behavioral and social studies, and zero-inflated count models such as zero-inflated Poisson (ZIP) and zero-inflated Negative Binomial (ZINB) can be applied when such zero-inflated count data are used as response variable. However, when the zero-inflated count data are used as predictors, ignoring the difference of structural and random zeros can result in biased estimates. In this paper, a generalized estimating equation (GEE)-type mixture model is proposed to jointly model the response of interest and the zero-inflated count predictors. Simulation studies show that the proposed method performs well for practical settings and is more robust for model misspecification than the likelihood-based approach. A case study is also provided for illustration. [ABSTRACT FROM AUTHOR]

In the last 50 years Thailand has achieved advanced demographic and health transitions. Many infectious diseases are controlled and infant and maternal mortality rates are among the lowest in the region. Within such a low mortality setting, however, substantial variations in health persist, with education being a major driver. This paper assesses the ongoing mortality transition in Thailand, examining relationships between risk factors and mortality outcomes among a large nationwide cohort of distance-learning Open University students, as well as examining the differential mortality benefit enjoyed by this educated group when compared with the general Thai population. The cohort comprised 87,151 participants, who in 2005 completed a questionnaire collecting detailed information on socio-demographics, health behaviours and health outcomes. Participants were aged 15-87 at baseline and lived in all regions of the country. Cohort members' citizen identification numbers were matched with official death registration records to identify mortality among participants. A total of 1401 deaths were identified up to November 2016. Results show cohort study participants experiencing mortality at approximately one-third of the rate of the general population in the same age and sex groups. The gap between the mortality rates in the two groups widened with increased age. Differential risk factor prevalence among the two populations, particularly lower overweight and obesity prevalence and lower cigarette smoking rates in the cohort, can explain some of this variation. The largely unmeasurable effect of aspiration for life improvement through distance education while embedded in their communities is shown by this study to have a powerful effect on mortality risk. With overall education levels, including higher education, rising in Thailand this growing group of educated aspirational Thais may represent future trends in Thai mortality. Identifying the drivers and characteristics of this mortality variation can help inform policies to provide health services and to help reduce mortality in the whole population. [ABSTRACT FROM AUTHOR]

Background: In the past 20 years, the trends of ischemic heart disease (IHD) mortality in China have been described in divergent claims. This research analyzes mortality trends for IHD by using the data from 102 continuous Disease Surveillance Points (DSP) from 1991 to 2009.Method: The 102 continuous DSP covered 7.3 million people during the period 1991-2000, and then were expanded to a population of 52 million in the same areas for 2004-2009. The data were adjusted by using garbage code redistribution and underreporting rate, mapped from international classification of diseases ICD-9 to ICD-10. The mortality rates for IHD were further adjusted by the crude death proportion multiplied by the total number of deaths in the mortality envelope, which was calculated by using logrt = a + bt. Age-standard death rates (ASDRs) were computed using China's 2010 census population structure. Trend in IHD was calculated from ASDRs by using a joinpoint regression model.Results: The IHD ASDRs increased in total in regions with an average annual percentage change (AAPC) 4.96%, especially for the Southwest (AAPC = 7.97%) and Northeast areas (AAPC = 7.10%), and for male and female subjects (with 5% AAPC) as well. In rural areas, the year 2000 was a cut-off point for mortality rate with annual percentage change increasing from 3.52% in 1991-2000 to 9.02% in 2000-2009, which was much higher than in urban areas (AAPC = 1.05%). And the proportion of deaths increased in older adults, and more male deaths occurred before age 60 compared to female deaths.Conclusion: By observing a wide range of areas across China from 1991 to 2009, this paper concludes that the ASDR trend for IHD increased. These trends reflect changes in the Chinese standard of living and lifestyle with diets higher in fat, higher blood lipids and increased body weight. [ABSTRACT FROM AUTHOR]

Home hemodialysis (HD) was first introduced in the 1960s with a rapid increase in its use due to inability of dialysis units to accommodate patient demand. A sharp decline was subsequently seen with expanding outpatient dialysis facilities and changes in reimbursement policies. In the last decade, with emerging reports of benefits with home HD and more user-friendly equipment, there has been resurgence in home HD. However, home HD remains underutilized with considerable variations between and within countries. This paper will review the history of home HD, elaborate on its established benefits, identify some of the barriers in uptake of this modality and expand on potential strategies to overcome these barriers. [ABSTRACT FROM AUTHOR]

Projections of both the short- and long-term course of the variant Creutzfeldt-Jakob disease (vCJD) epidemic in Great Britain have demonstrated great uncertainty due to the lack of knowledge of key aspects of the biology and epidemiology of this new disease. Such projections are sensitive to assumptions made regarding the pattern of exposure to BSE-infected animals, the effectiveness of control measures introduced in 1989 and 1996 in reducing this exposure, the functional form of the incubation period distribution and patterns of age-dependent susceptibility/exposure. This paper provides short-term projections for vCJD onsets using the time- and age-distributed onset data to the end of 2000, with results that are directly comparable to the other papers in this issue. These results demonstrate the continued uncertainty in the future scale of this disease. [ABSTRACT FROM AUTHOR]

This paper examines the paper of Haux et al. in this issue of this journal. It gives some background on specifics of the German health care system, which underlie the theses and prognoses proposed by Haux et al. In analogy to a forecast of the future of health informatics, which is now 10 years old, I then suggest that these attempts meet two types of challenges:•that of overestimating the positive effects of recent advances, which later are found not to scale up; and•that of blind spots with respect to unforeseen significant advances.The attempt to find indications of such in the projections of Haux et al. leads, among other, to the conclusion that the projections of direct linkages between patients and care providers may be overoptimistic. As to wheather the deviating opinions matter in the end, it is concluded that the technology advances may require less attention that the restructuring of the health care system required to take advantage of the advances of technology. [Copyright &y& Elsevier]

This editorial is a commentary on the review paper by Ramirez Varela et al. entitled "Mapping the historical development of physical activity and health research: a structured literature review and citation network analysis." This editorial highlights the significance and implications of this review, with a particular focus on future research and policy directions. [ABSTRACT FROM AUTHOR]

Patients' experience of stereotype threat in clinical settings and encounters may be one contributor to health care disparities. Stereotype threat occurs when cues in the environment make negative stereotypes associated with an individual's group status salient, triggering physiological and psychological processes that have detrimental consequences for behavior. By recognizing and understanding the factors that can trigger stereotype threat and understanding its consequences in medical settings, providers can prevent it from occurring or ameliorate its consequences for patient behavior and outcomes. In this paper, we discuss the implications of stereotype threat for medical education and trainee performance and offer practical suggestions for how future providers might reduce stereotype threat in their exam rooms and clinics. [ABSTRACT FROM AUTHOR]

Adults experiencing homelessness are more likely to have an alcohol use disorder compared to adults in the general population. Although shelter-based treatments are common, completion rates tend to be poor, suggesting a need for more effective approaches that are tailored to this understudied and underserved population. One barrier to developing more effective treatments is the limited knowledge of the triggers of alcohol use among homeless adults. This paper describes the use of ecological momentary assessment (EMA) to identify predictors of "imminent drinking" (i.e., drinking within the next 4 h), among a sample of adults experiencing homelessness and receiving health services at a homeless shelter. A total of 78 mostly male (84.6%) adults experiencing homelessness (mean age = 46.6) who reported hazardous drinking completed up to five EMAs per day over 4 weeks (a total of 4557 completed EMAs). The study used machine learning techniques to create a drinking risk algorithm that predicted 82% of imminent drinking episodes within 4 h of the first drink of the day, and correctly identified 76% of nondrinking episodes. The algorithm included the following 7 predictors of imminent drinking: urge to drink, having alcohol easily available, feeling confident that alcohol would improve mood, feeling depressed, lower commitment to being alcohol free, not interacting with someone drinking alcohol, and being indoors. The research team used the results to develop intervention content (e.g., brief tailored messages) that will be delivered when imminent drinking is detected in an upcoming intervention phase. Specifically, we created three theoretically grounded message tracks focused on urge/craving, social/availability, and negative affect/mood, which are further tailored to a participant's current drinking goal (i.e., stay sober, drink less, no goal) to support positive change. To our knowledge, this is the first study to develop tailored intervention messages based on likelihood of imminent drinking, current drinking triggers, and drinking goals among adults experiencing homelessness. [ABSTRACT FROM AUTHOR]

Aging is associated with a number of physiologic and functional declines that can contribute to increased disability, frailty, and falls. Contributing factors are the loss of muscle mass and strength as age increases, a phenomenon called sarcopenia. Sarcopenia can result or be exacerbated by certain chronic conditions, and can also increase the burden of chronic disease. Current research has demonstrated that strength-training exercises have the ability to combat weakness and frailty and their debilitating consequences. Done regularly (e.g., 2 to 3 days per week), these exercises build muscle strength and muscle mass and preserve bone density, independence, and vitality with age. In addition, strength training also has the ability to reduce the risk of osteoporosis and the signs and symptoms of numerous chronic diseases such as heart disease, arthritis, and type 2 diabetes, while also improving sleep and reducing depression. This paper reviews the current research on strength training and older adults, evaluating exercise protocols in a variety of populations. It is clear that a variety of strength-training prescriptions from highly controlled laboratory-based to minimally supervised home-based programs have the ability to elicit meaningful health benefits in older adults. The key challenges as this field of exercise science moves forward are to best identify the most appropriate strength-training recommendations for older adults and to greatly increase the access to safe and effective programs in a variety of settings. [ABSTRACT FROM AUTHOR]

Incubation period of the new variant Creutzfeldt-Jakob disease (vCJD) from infection to clinical onset and the eventual impact of the disease remain major concerns. Based on i) epidemiological conceptualization of human exposure to BSE contaminated material, ii) exponentially decreasing susceptibility after 15 years of age, and iii) typical incubation period (IP) distributions for time from infection to onset, we have previously estimated mean incubation period and projected number of vCJD cases. In this paper, we investigate the robustness of these estimates with respect to i-iii using the UK's 113 vCJD cases with clinical onset before December 2000. Mean incubation period was estimated at 16.4 years (95% CI 11.4-24.8), 15.9 years (95% CI 11.4-22.0), 14.1 years (95% CI 10.4-24.2) with the log-normal, Gamma and Weibull distributions respectively. Corresponding predictions for the total size of the epidemic ranged from 183 to 304. Maximal susceptibility to infection between 1.3 and 15.9 years and decreasing by 15% per year of age thereafter yielded the best fit. The shape of the IP distribution did not affect the predictions. In summary, within a set of reasonable assumptions, mean incubation period for vCJD ranged from 15 to 20 years, and the eventual impact of vCJD was a few hundred patients. [ABSTRACT FROM AUTHOR]

Physicians are often asked to make prognostic assessments but often worry that their assessments will prove inaccurate. Prognostic systems were developed to enhance the accuracy of such assessments. This paper describes an approach for evaluating prognostic systems based on the accuracy (calibration and discrimination) and generalizability (reproducibility and transportability) of the system's predictions. Reproducibility is the ability to produce accurate predictions among patients not included in the development of the system but from the same population. Transportability is the ability to produce accurate predictions among patients drawn from a different but plausibly related population. On the basis of the observation that the generalizability of a prognostic system is commonly limited to a single historical period, geographic location, methodologic approach, disease spectrum, or follow-up interval, we describe a working hierarchy of the cumulative generalizability of prognostic systems. This approach is illustrated in a structured review of the Dukes and Jass staging systems for colon and rectal cancer and applied to a young man with colon cancer. Because it treats the development of the system as a "black box" and evaluates only the performance of the predictions, the approach can be applied to any system that generates predicted probabilities. Although the Dukes and Jass staging systems are discrete, the approach can also be applied to systems that generate continuous predictions and, with some modification, to systems that predict over multiple time periods. Like any scientific hypothesis, the generalizability of a prognostic system is established by being tested and being found accurate across increasingly diverse settings. The more numerous and diverse the settings in which the system is tested and found accurate, the more likely it will generalize to an untested setting. [ABSTRACT FROM AUTHOR]

In this paper we argue that predictive processing (PP) theory cannot account for the phenomenon of affect-biased attention - prioritized attention to stimuli that are affectively salient because of their associations with reward or punishment. Specifically, the PP hypothesis that selective attention can be analyzed in terms of the optimization of precision expectations cannot accommodate affect-biased attention; affectively salient stimuli can capture our attention even when precision expectations are low. We review the prospects of three recent attempts to accommodate affect with tools internal to PP theory: Miller and Clark's (2018) embodied inference; Seth's (2013) interoceptive inference; and Joffily and Coricelli's (2013) rate of change of free energy. In each case we argue that the account does not resolve the challenge from affect-biased attention. For this reason, we conclude that prediction error minimization is not sufficient to explain all mental phenomena, contrary to the claim that the PP framework provides a unified theory of all mental phenomena or the brain's cognitive functioning. Nevertheless, we suggest that empirical investigation of the interaction between affective salience and precision expectations should prove helpful in understanding the limits of PP theory, and may provide new directions for the application of a Bayesian perspective to perception. [ABSTRACT FROM AUTHOR]

Patients with advanced cancer are burdened physically and psychologically, so there is an urgent need to pay more attention to their health-related quality of life (HRQOL). With an expected clinical endpoint prediction, over-treatment can be effectively eliminated by the means of palliative care at the right time. This paper develops a deep learning based approach for cancer clinical endpoint prediction based on patient's electronic health records (EHR). Due to the pervasive existence of categorical information in EHR, it brings unavoidably obstacles to the effective numerical learning algorithms. To address this issue, we propose a novel cross-field categorical attributes embedding (CCAE) model to learn a vectorized representation for cancer patients in attribute-level by orders, in which the strong semantic coupling among categorical variables are well exploited. By transforming the order-dependency modeling into a sequence learning task in an ingenious way, recurrent neural network is adopted to capture the semantic relevance among multi-order representations. Experimental results from the SEER-Medicare EHR dataset have illustrated that the proposed model can achieve competitive prediction performance compared with other baselines. [ABSTRACT FROM AUTHOR]