The optimal learner for prediction modeling varies depending on the underlying data-generating distribution. Super Learner (SL) is a generic ensemble learning algorithm that uses cross-validation to select among a "library" of candidate prediction models. The SL is not restricted to a single prediction model, but uses the strengths of a variety of learning algorithms to adapt to different databases. While the SL has been shown to perform well in a number of settings, it has not been thoroughly evaluated in large electronic healthcare databases that are common in pharmacoepidemiology and comparative effectiveness research. In this study, we applied and evaluated the performance of the SL in its ability to predict treatment assignment using three electronic healthcare databases. We considered a library of algorithms that consisted of both nonparametric and parametric models. We also considered a novel strategy for prediction modeling that combines the SL with the high-dimensional propensity score (hdPS) variable selection algorithm. Predictive performance was assessed using three metrics: the negative log-likelihood, area under the curve (AUC), and time complexity. Results showed that the best individual algorithm, in terms of predictive performance, varied across datasets. The SL was able to adapt to the given dataset and optimize predictive performance relative to any individual learner. Combining the SL with the hdPS was the most consistent prediction method and may be promising for PS estimation and prediction modeling in electronic healthcare databases.

Recurrence rate of 76% for patients with stage I ULMS was higher than previously published cohorts. Mitotic counts were associated with increased recurrence and decreased OS. Expressions of ER, EGFR, and Ki-67 were not useful for predicting overall recurrence or survival.

Aim: Differences in children's development and susceptibility to diseases and exposures have been observed by sex, yet human studies of sex differences in miRNAs are limited. Materials & methods: The genome-wide miRNA expression was characterized by sequencing-based EdgeSeq assay in cord blood buffy coats from 89 newborns, and 564 miRNAs were further analyzed. Results: Differential expression of most miRNAs was higher in boys. Neurodevelopment, RNA metabolism and metabolic ontology terms were enriched among miRNA targets. The majority of upregulated miRNAs (86%) validated by nCounter maintained positive-fold change values; however, only 21% reached statistical significance by false discovery rate. Conclusion: Accounting for host factors like sex may improve the sensitivity of epigenetic analyses for epidemiological studies in early childhood.

BackgroundIn recent decades, suicide and fatal overdose rates have increased in the US, particularly for working-age adults with no college education. The coincident decline in manufacturing has limited stable employment options for this population. Erosion of the Michigan automobile industry provides a striking case study.MethodsWe used individual-level data from a retrospective cohort study of 26 804 autoworkers in the United Autoworkers-General Motors cohort, using employment records from 1970 to 1994 and mortality follow-up from 1970 to 2015. We estimated HRs for suicide or fatal overdose in relation to leaving work, measured as active or inactive employment status and age at worker exit.ResultsThere were 257 deaths due to either suicide (n=202) or overdose (n=55); all but 21 events occurred after leaving work. The hazard rate for suicide was 16.1 times higher for inactive versus active workers (95% CI 9.8 to 26.5). HRs for suicide were elevated for all younger age groups relative to those leaving work after age 55. Those 30–39 years old at exit had the highest HR for suicide, 1.9 (95% CI 1.2 to 3.0). When overdose was included, the rate increased by twofold for both 19- to 29-year-olds and 30- to 39-year-olds at exit. Risks remained elevated when follow-up was restricted to 5 years after exit.ConclusionsAutoworkers who left work had a higher risk of suicide or overdose than active employees. Those who left before retirement age had higher rates than those who left after, suggesting that leaving work early may increase the risk.