Machine Learning for Drug Overdose Surveillance

ABSTRACT

We describe two recently proposed machine learning approaches for discovering emerging trends in fatal accidental drug overdoses. The Gaussian Process Subset Scan (Herlands, McFowland, Wilson, & Neill, 2017 Neill, D. B. (2017). Multidimensional tensor scan for drug overdose surveillance. Journal of Public Health Informatics, 9(1), e20. doi:10.5210/ojphi.v9i1.7598[Crossref] , [Google Scholar]) enables early detection of emerging patterns in spatio-temporal data, accounting for both the complex, correlated nature of the data and the fact that detecting subtle patterns requires integration of information across multiple spatial areas and multiple time steps. We apply this approach to 17 years of county-aggregated data for monthly opioid overdose deaths in the New York City metropolitan area, showing clear advantages in the utility of discovered patterns as compared to typical anomaly detection approaches. To detect and characterize emerging overdose patterns that differentially affect a subpopulation of the data, including geographic, demographic, and behavioral patterns (e.g., which combinations of drugs are involved), we apply the Multidimensional Tensor Scan (Neill, 2017 Neill, D. B. (2017). Multidimensional tensor scan for drug overdose surveillance. Journal of Public Health Informatics, 9(1), e20. doi:10.5210/ojphi.v9i1.7598[Crossref] , [Google Scholar]) to 8 years of case-level overdose data from Allegheny County, Pennsylvania. We discover previously unidentified overdose patterns which reveal unusual demographic clusters, show impacts of drug legislation, and demonstrate potential for early detection and targeted intervention. These approaches to early detection of overdose patterns can inform prevention and response efforts, as well as understanding the effects of policy changes.     

Tracking of sport and exercise types from midlife to old age: a 20-year cohort study of British men

Background

Previous physical activity (PA) tracking studies have examined the stability of overall PA and/or PA types, but few have investigated how specific types of sport/exercise track over the life course. The aim of this study was to determine how specific sports/exercises in midlife track and predict future sport/exercise and PA in men transitioning to old age.

Methods

Seven thousand seven hundred thirty-five men (aged 40–59?years) recruited in 1978–80 were followed up after 12, 16 and 20?years. At each wave men self-reported participation in sport/exercise. Frequent sport/exercise participants (>?1/month) reported the types of sport/exercise they engaged in. Men also reported total PA, health status, lifestyle behaviours and socio-demographic characteristics. Stability of each sport/exercise was assessed using kappa statistics and intraclass correlation coefficients. Logistic regression estimated the odds of participating in sport/exercise and being active at 20-year follow up according to specific types of sport/exercise in midlife.

Results

Three thousand three hundred eighty-four men with complete data at all waves were included in analyses. Tracking of specific sports/exercises ranged from fair to substantial, with golf being the most common and most stable. Bowls was the most frequently adopted. Odds of participating in sport/exercise and being active in old age varied according to sport/exercise types in midlife. Golf and bowls in midlife were the strongest predictors of sport/exercise participation in old age. Golf, cricket and running/jogging in midlife were among the strongest predictors of being active in old age. Compared to participating in just one sport/exercise in midlife, sampling multiple sports/exercises was more strongly associated with sport/exercise participation and being active in old age.

Conclusion

The stability of sport/exercise participation from midlife to old age varies by type. Specific sports/exercises in midlife may be more likely to predict future PA than others. However, participating in a range of sports/exercises may be optimal for preserving PA into old age.

     

Occupational closure and immigrant entrepreneurship: Russian Jews in Israel

The recent wave of Jews emigrating from countries of the former Soviet Union to Israel has created structural conditions for immigrant entrepreneurship. The large size of the migrant population and its spatial concentration create a demand for services and products that can be provided by immigrants. This paper investigates the factors associated with intentions to open a small business in a sample of Jewish immigrants from the countries of the former Soviet Union. It was found that intentions to open a small business are related to previous business experience, education, income, and length of residence in the country. Immigrants become interested in entrepreneurship after learning that their prospects of finding a job in their profession are meager and explain their motivation to open a small business as being to increase their income. The findings seem to support the disadvantage theory that conceptualizes entrepreneurship as an adaptive mechanism to structural barriers in the primary labor market that create an occupational closure for immigrants possessing low and middle levels of education and income.     

The Relation of Negative Career Thoughts to Depression and Hopelessness

Although some research literature focuses on the integration of mental health and career counseling, there has been little that examines both areas in relation to depression and hopelessness. This study investigated the relationship among dysfunctional career thinking, depression, and hopelessness in a sample of 139 undergraduate and graduate students seeking drop‐in or individual career counseling services at a university career center. The authors found that two aspects of dysfunctional career thinking, decision‐making confusion and commitment anxiety, accounted for a significant amount of variance in depression. Decision‐making confusion also accounted for a significant amount of variance in hopelessness. Implications for counseling practice include the need for more careful screening of career clients who present with high levels of anxiety and negative thinking. Future research could involve more diverse client populations, such as unemployed adults, and explore the use of additional screening measures to assess the intersection of career and mental health issues.     

Sexual orientation differences in HIV testing motivation among college men

Objective: To investigate sexual orientation differences in college men's motivations for HIV testing. Participants: 665 male college students in the Southeastern United States from 2006 to 2014. Methods: Students completed a survey on HIV risk factors and testing motivations. Logistic regressions were conducted to determine the differences between heterosexual men (HM) and sexual minority men (SMM). Results: SMM were more motivated to get tested by concern over past condomless sex, while HM were more often cited supporting the testing program “on principle” and wanting a free t-shirt. SMM and HM differed in behaviors that impact HIV risk and other demographics. However, differences in testing motivation by concern over past condomless sex or wanting a free t-shirt persisted when controlling for these demographic and behavioral differences. Conclusions: Appropriately designed HIV prevention interventions on college campuses should target SMM's distinct concern over past condomless sex as a testing motivation.     

Ignorability conditions for frequentist non parametric analysis of conditional distributions with incomplete data

Rubin (1976 Rubin, D.B. (1976). Inference and missing data. Biometrika 63(3):581–592.[Crossref], [Web of Science ®] , [Google Scholar]) derived general conditions under which inferences that ignore missing data are valid. These conditions are sufficient but not generally necessary, and therefore may be relaxed in some special cases. We consider here the case of frequentist estimation of a conditional cdf subject to missing outcomes. We partition a set of data into outcome, conditioning, and latent variables, all of which potentially affect the probability of a missing response. We describe sufficient conditions under which a complete-case estimate of the conditional cdf of the outcome given the conditioning variable is unbiased. We use simulations on a renal transplant data set (Dienemann et al.) to illustrate the implications of these results.