Ineffectiveness of Paramedic Naloxone Administration as a Standalone Metric for Community Opioid Overdoses and the Increasing Use of Naloxone by Community Members.

INTRODUCTION: With Canada's growing opioid crisis, many communities are attempting to monitor cases in real-time. Paramedic Naloxone Administration (PNA) has become a common metric for monitoring overdoses. We evaluate whether the use of naloxone administration counts represents an effective monitoring tool for community opioid overdoses. METHODS: The electronic ambulance call report database of Peterborough Paramedics (Ontario, Canada) was examined. De-identified records from 2016-2019 with problem codes of "Opioid Overdose", along with all patients documented as receiving naloxone were extracted. Chi-square and Bonferroni-adjusted post hoc proportion tests were used for comparison of counts. RESULTS: 558 opioid overdoses were identified, 124 (22%) of which had PNA documented, 181(32%) had naloxone prior to arrival documented and 264 (47%) received no naloxone. Over the three years, the annual number of overdose cases increased, while the proportion of patients receiving PNA decreased significantly each year. PNA was also associated with calls in a residence. Naloxone was administered by a non-paramedic in 262 cases, with 181 of these identified as opioid overdoses and was more common in later years and in cases occurring in public places. CONCLUSION: PNA calls did not account for a significant percentage of opioid overdoses attended to by paramedics. The strong association between PNA and call location being a residence, along with increasing use of community naloxone kits, may cause certain populations to be under-represent if PNA is used as a standalone metric. The decreasing association with time may also lead to a falsely improving metric further reducing its effectiveness. Thus, PNA when used alone may no longer be a suitable metric for opioid overdose tracking.

A dependent Bayesian Dirichlet process model for source apportionment of particle number size distribution.

The relationship between particle exposure and health risks has been well established in recent years. Particulate matter (PM) is made up of different components coming from several sources, which might have different level of toxicity. Hence, identifying these sources is an important task in order to implement effective policies to improve air quality and population health. The problem of identifying sources of particulate pollution has already been studied in the literature. However, current methods require an a priori specification of the number of sources and do not include information on covariates in the source allocations. Here, we propose a novel Bayesian nonparametric approach to overcome these limitations. In particular, we model source contribution using a Dirichlet process as a prior for source profiles, which allows us to estimate the number of components that contribute to particle concentration rather than fixing this number beforehand. To better characterize them we also include meteorological variables (wind speed and direction) as covariates within the allocation process via a flexible Gaussian kernel. We apply the model to apportion particle number size distribution measured near London Gatwick Airport (UK) in 2019. When analyzing this data, we are able to identify the most common PM sources, as well as new sources that have not been identified with the commonly used methods.

An illustration of model agnostic explainability methods applied to environmental data.

Historically, two primary criticisms statisticians have of machine learning and deep neural models is their lack of uncertainty quantification and the inability to do inference (i.e., to explain what inputs are important). Explainable AI has developed in the last few years as a sub-discipline of computer science and machine learning to mitigate these concerns (as well as concerns of fairness and transparency in deep modeling). In this article, our focus is on explaining which inputs are important in models for predicting environmental data. In particular, we focus on three general methods for explainability that are model agnostic and thus applicable across a breadth of models without internal explainability: "feature shuffling", "interpretable local surrogates", and "occlusion analysis". We describe particular implementations of each of these and illustrate their use with a variety of models, all applied to the problem of long-lead forecasting monthly soil moisture in the North American corn belt given sea surface temperature anomalies in the Pacific Ocean.

Air Health Trend Indicator: Association between Short-Term Exposure to Ground Ozone and Circulatory Hospitalizations in Canada for 17 Years, 1996⁻2012.

The Air Health Trend Indicator is designed to estimate the public health risk related to short-term exposure to air pollution and to detect trends in the annual health risks. Daily ozone, circulatory hospitalizations and weather data for 24 cities (about 54% of Canadians) for 17 years (1996⁻2012) were used. This study examined three circulatory causes: ischemic heart disease (IHD, 40% of cases), other heart disease (OHD, 31%) and cerebrovascular disease (CEV, 14%). A Bayesian hierarchical model using a 7-year estimator was employed to find trends in the annual national associations by season, lag of effect, sex and age group (≤65 vs. >65). Warm season 1-day lagged ozone returned higher national risk per 10 ppb: 0.4% (95% credible interval, -0.3⁻1.1%) for IHD, 0.4% (-0.2⁻1.0%) for OHD, and 0.2% (-0.8⁻1.2%) for CEV. Overall mixed trends in annual associations were observed for IHD and CEV, but a decreasing trend for OHD. While little age effect was identified, some sex-specific difference was detected, with males seemingly more vulnerable to ozone for CEV, although this finding needs further investigation. The study findings could reduce a knowledge gap by identifying trends in risk over time as well as sub-populations susceptible to ozone by age and sex.

The Oakville Oil Refinery Closure and Its Influence on Local Hospitalizations: A Natural Experiment on Sulfur Dioxide.

Background: An oil refinery in Oakville, Canada, closed over 2004⁻2005, providing an opportunity for a natural experiment to examine the effects on oil refinery-related air pollution and residents' health. Methods: Environmental and health data were collected for the 16 years around the refinery closure. Toronto (2.5 million persons) and the Greater Toronto Area (GTA, 6.3 million persons) were used as control and reference populations, respectively, for Oakville (160,000 persons). We compared sulfur dioxide and age- and season-standardized hospitalizations, considering potential factors such as changes in demographics, socio-economics, drug prescriptions, and environmental variables. Results: The closure of the refinery eliminated 6000 tons/year of SO2 emissions, with an observed reduction of 20% in wind direction-adjusted ambient concentrations in Oakville. After accounting for trends, a decrease in cold-season peak-centered respiratory hospitalizations was observed for Oakville (reduction of 2.2 cases/1000 persons per year, p = 0.0006 ) but not in Toronto (p = 0.856) and the GTA (p = 0.334). The reduction of respiratory hospitalizations in Oakville post closure appeared to have no observed link to known confounders or effect modifiers. Conclusion: The refinery closure allowed an assessment of the change in community health. This natural experiment provides evidence that a reduction in emissions was associated with improvements in population health. This study design addresses the impact of a removed source of air pollution.

The use of chained two-point clusters for the examination of associations of air pollution with health conditions.

OBJECTIVES: There are a few accepted and intensively applied statistical methods used to study associations of ambient air pollution with health conditions. Among the most popular methods applied to assess short term air health effects are case-crossover (using events) and time-series methodologies (using counts). A few other techniques for studying counts of events have been proposed, including the Generalized Linear Mixed Models (GLMM). One suggested GLMM technique uses cluster structures based on natural embedded hierarchies: days are nested in the days of a week (dow), which, in turn, are nested in months and months in years (< dow, month, years >). MATERIAL AND METHODS: In this study the authors considered clusters with hierarchical structures in a form of < dow, 14-days, year >, where the 14-days hierarchy determines 7 clusters composed of 2 days (the same days) of a week (2 Mondays, 2 Tuesdays, etc.), in 1 year. In this work the authors proposed hierarchical chained clusters in which 2 days of a week are grouped as follows: (first, second), (second, third), (third, fourth) and so on. Such an approach allows determination of an additional series of the slopes on the clusters (second, third), (fourth, fifth), etc., i.e., estimation of the coefficients for other configurations of air pollutant levels. The authors considered a series of 2 point chained clusters covering a year. In such a construction each cluster has one common data point (day) with another one. RESULTS: The authors estimated coefficients (slopes) related to the ambient ozone exposure (mortality) and to 3 selected air pollutants (particulate matter, nitrogen dioxide and ozone) combined into index and considered as health risk exposure (emergency department (ED) visits). The generated results were compared to the estimations obtained from the time-series method and the time-stratified case-crossover method applied to the same data. CONCLUSIONS: The proposed statistical method, based on the chained hierarchical clusters (< dow, 14-days, year >), generated results with shorter confidence intervals than the other methods.