Machine Learning for Sudden Cardiac Death Prediction in the Atherosclerosis Risk in Communities Study

Introduction Sudden cardiac death (SCD) is a devastating consequence often without antecedent expectation. Current risk stratification methods derived from baseline independently modeled risk factors are insufficient. Novel random forest machine learning (ML) approach incorporating time-dependent variables and complex interactions may improve SCD risk prediction. Methods Atherosclerosis Risk in Communities (ARIC) study participants were followed for adjudicated SCD. ML models were compared to standard Poisson regression models for interval data, an approximation to Cox regression, with stepwise variable selection. Eighty-two time-varying variables (demographics, lifestyle factors, clinical characteristics, biomarkers, etc.) collected at four visits over 12 years (1987-98) were used as candidate predictors. Predictive accuracy was assessed by area under the receiver operating characteristic curve (AUC) through out-of-bag prediction for ML models and 5-fold cross validation for the Poisson regression models. Results Over a median follow-up time of 23.5 years, 583 SCD events occurred among 15,661 ARIC participants (mean age 54 years and 55% women). Compared to different Poisson regression models (AUC at 6-year ranges from 0.77-0.83), the ML model improved prediction (AUC at 6-year 0.89). Top predictors identified by ML model included prior coronary heart disease (CHD), which explained 47.9% of the total phenotypic variance, diabetes mellitus, hypertension, and T wave abnormality in any of leads I, aVL, or V6. Using the top ML predictors to select variables, the Poisson regression model AUC at 6-year was 0.77 suggesting that the non-linear dependencies and interactions captured by ML, are the main reasons for its improved prediction performance. Conclusions Applying novel ML approach with time-varying predictors improves the prediction of SCD. Interactions of dynamic clinical characteristics are important for risk-stratifying SCD in the general population.

PERBANDINGAN MODEL REGRESI BINOMIAL NEGATIF BIVARIAT DENGAN MODEL GEOGRAPHICALLY WEIGHTED NEGATIVE BINOMIAL BIVARIAT REGRESSION (GWNBBR) PADA KASUS ANGKA KEMATIAN BAYI DAN KEMATIAN IBU DI JAWA TENGAH

Infant mortality (0-11 months) and maternal mortality (during pregnancy, childbirth, and postpartum) are significant indicators in determining the level of public health. Central Java Province which has 35 regencies/cities is included in the top five regions with the highest number of infant and maternal mortality in Indonesia. The data characteristics of the number of infants and maternal mortality are count data. Therefore, the Poisson Regression method can be used to analyze the factors that influence the number of infants and maternal mortality. In Poisson regression analysis, there must be a fulfilled assumption, called equidispersion. Frequently, the variance of count data is greater than the mean, which is known as the overdispersion. The research, binomial negative bivariate regression is used as a solutions to overcome the problem of overdispersion in poisson regression. This method produce a global model. In reality, the geographical, socio-cultural, and economic conditions of each region will be different. This illustrates the effect of spatial heterogeneity, so it needs to be developed into Geographically Weighted Negative Binomial Bivariate Regression (GWNBBR). The model of GWNBBR provides weighting based on the position or distance from one observation area to another. Significant variables for modeling infant mortality cases included the percentage of obstetric complications treated (X1), the percentage of infants who were exclusively breastfed (X3), and the percentage of poor people (X5). Significant variable for modeling maternal mortality cases is the percentage of poor people (X5). Based on the AIC value, GWNBBR model is better than binomial negatif bivariat regression model because it has a smaller AIC value.

Trends of in-hospital cardiac arrests in a single tertiary hospital with a mature rapid response system

Background Most studies on rapid response system (RRS) have simply focused on its role and effectiveness in reducing in-hospital cardiac arrests (IHCAs) or hospital mortality, regardless of the predictability of IHCA. This study aimed to identify the characteristics of IHCAs including predictability of the IHCAs as our RRS matures for 10 years, to determine the best measure for RRS evaluation. Methods Data on all consecutive adult patients who experienced IHCA and received cardiopulmonary resuscitation in general wards between January 2010 and December 2019 were reviewed. IHCAs were classified into three groups: preventable IHCA (P-IHCA), non-preventable IHCA (NP-IHCA), and inevitable IHCA (I-IHCA). The annual changes of three groups of IHCAs were analyzed with Poisson regression models. Results Of a total of 800 IHCA patients, 149 (18.6%) had P-IHCA, 465 (58.1%) had NP-IHCA, and 186 (23.2%) had I-IHCA. The number of the RRS activations increased significantly from 1,164 in 2010 to 1,560 in 2019 (P = 0.009), and in-hospital mortality rate was significantly decreased from 9.20/1,000 patients in 2010 to 7.23/1000 patients in 2019 (P = 0.009). The trend for the overall IHCA rate was stable, from 0.77/1,000 patients in 2010 to 1.06/1,000 patients in 2019 (P = 0.929). However, while the incidence of NP-IHCA (P = 0.927) and I-IHCA (P = 0.421) was relatively unchanged over time, the incidence of P-IHCA decreased from 0.19/1,000 patients in 2010 to 0.12/1,000 patients in 2019 (P = 0.025). Conclusions The incidence of P-IHCA could be a quality metric to measure the clinical outcomes of RRS implementation and maturation than overall IHCAs.

Legislative provisions for standardised tobacco packaging and changes in public support in 27 European countries

BackgroundGlobal adoption of standardised packaging requirements for tobacco products is a victory for public health, but their proliferation and impacts rely partly on public support. How this is related to legislation remains underassessed. This study explored change over time in public support for standardised packaging in countries with varying degrees of legislative provisions.MethodsWe used data from 27 European countries, collected from 2017 (n=28, 300) and 2020 (n=27, 901) waves of the Eurobarometer survey, to assess self-reported support for standardised packaging regulations among both smokers and non-smokers. Countries were grouped into three categories of policy adoption (policy implemented; policy legislated; no legislation) and changes in support were assessed using multilevel Poisson regression models.ResultsIn 2020, public support for standardised packaging was 71% (95% CI 68% to 74%) in countries that implemented standardised packaging legislation, 57% (55% to 60%) in countries that had legislated but not yet implemented legislation and 41% (40%to 42%) in countries with no legislation. Compared with 2017, this represented a relative change of +8% (1% to 15%), +12% (5% to 21%) and −5% (95% CI −2% to −8%), respectively, in the three country categories. Among smokers, there was no indication of change in support across the three groups. Among non-smokers, support increased in countries with existing legislation (adjusted prevalence ratio [aPR]=1.14, 95% CI 1.06 to 1.23) and decreased in countries with no legislation (aPR=0.93, 0.90 to 0.97).ConclusionsPublic support for standardised packaging regulations increased in countries implementing and legislating for these measures, particularly among non-smokers. An overall increase in support provides reassurance for policymakers defending policy action on tobacco packaging, as well as for those seeking to implement standardised packaging in their own countries.

Improved log-Gaussian approximation for over-dispersed Poisson regression: Application to spatial analysis of COVID-19

In the era of open data, Poisson and other count regression models are increasingly important. Still, conventional Poisson regression has remaining issues in terms of identifiability and computational efficiency. Especially, due to an identification problem, Poisson regression can be unstable for small samples with many zeros. Provided this, we develop a closed-form inference for an over-dispersed Poisson regression including Poisson additive mixed models. The approach is derived via mode-based log-Gaussian approximation. The resulting method is fast, practical, and free from the identification problem. Monte Carlo experiments demonstrate that the estimation error of the proposed method is a considerably smaller estimation error than the closed-form alternatives and as small as the usual Poisson regressions. For counts with many zeros, our approximation has better estimation accuracy than conventional Poisson regression. We obtained similar results in the case of Poisson additive mixed modeling considering spatial or group effects. The developed method was applied for analyzing COVID-19 data in Japan. This result suggests that influences of pedestrian density, age, and other factors on the number of cases change over periods.

A Flexible Mixed Model for Clustered Count Data

Clustered count data are commonly modeled using Poisson regression with random effects to account for the correlation induced by clustering. The Poisson mixed model allows for overdispersion via the nature of the within-cluster correlation, however, departures from equi-dispersion may also exist due to the underlying count process mechanism. We study the cross-sectional COM-Poisson regression model—a generalized regression model for count data in light of data dispersion—together with random effects for analysis of clustered count data. We demonstrate model flexibility of the COM-Poisson random intercept model, including choice of the random effect distribution, via simulated and real data examples. We find that COM-Poisson mixed models provide comparable model fit to well-known mixed models for associated special cases of clustered discrete data, and result in improved model fit for data with intermediate levels of over- or underdispersion in the count mechanism. Accordingly, the proposed models are useful for capturing dispersion not consistent with commonly used statistical models, and also serve as a practical diagnostic tool.

An Epidemiological Study of Kidney Calculi in Western Guangdong

Objective: To investigate the risk factors of kidney calculi in its high prevalence areas of western Guangdong, and provide the proper prevention measures. Methods: A cross-sectional survey was conducted among individuals in Maoming, western Guangdong. Univariate and Poisson regression analyses were applied to investigate the influence of the epidemiology, lifestyle, and environmental factors on renal calculi. Risk ratios with 95% confidence interval were used to estimate the association between the investigated factors and the prevalence of renal calculi. Results: 500 questionnaires were sent out and 481 valid questionnaires were recycled. Among 481 respondents, 84 had renal calculi with a prevalence rate of 17.46%. Univariate regression analysis showed statistically significant differences in the prevalence of renal calculi among different groups of sex, ages, family history of kidney calculi, diet and drinking habit. Poisson regression analysis showed that daily water intake and drinking boiled water were protective factors, while male, family history of kidney calculi, diet high in protein, sugar, vitamin C and calcium were risk factors. Additionally, high sugar diet was not statistically significantly associated with the occurrence of renal calculi. Conclusion: The occurrence of kidney calculi in western Guangdong is closely related to demographic characteristics of individuals, living and dietary habits of the resident populations.

Bayesian and Classical Prediction Models for Categorical and Count Data

In this chapter, we explain, under a Bayesian framework, the fundamentals and practical issues for implementing genomic prediction models for categorical and count traits. First, we derive the Bayesian ordinal model and exemplify it with plant breeding data. These examples were implemented in the library BGLR. We also derive the ordinal logistic regression. The fundamentals and practical issues of penalized multinomial logistic regression and penalized Poisson regression are given including several examples illustrating the use of the glmnet library. All the examples include main effects of environments and genotypes as well as the genotype × environment interaction term.