Do Machine Learning Methods Improve Prediction of Ambient Air Pollutants with High Spatial Contrast? A Systematic Review.

BACKGROUND & OBJECTIVE: The use of machine learning for air pollution modelling is rapidly increasing. We conducted a systematic review of studies comparing statistical and machine learning models predicting the spatiotemporal variation of ambient nitrogen dioxide (NO2), ultrafine particles (UFPs) and black carbon (BC) to determine whether and in which scenarios machine learning generates more accurate predictions. METHODS: Web of Science and Scopus were searched up to June 13, 2024. All records were screened by two independent reviewers. Differences in the coefficient of determination (R2) and Root Mean Square Error (RMSE) between best statistical and machine learning methods were compared across categories of methodological elements. RESULTS: A total of 38 studies with 46 model comparisons (30 for NO2, 8 for UFPs and 8 for BC) were included. Linear non-regularized methods and Random Forest were most frequently used. Machine learning outperformed statistical models in 34 comparisons. Mean differences (95% confidence intervals) in R2 and RMSE between best machine learning and statistical models were 0.12 (0.08, 0.17) and 20% (11%, 29%) respectively. Tree-based methods performed best in 12 of 17 multi-model comparisons. Nonlinear or regularization regression methods were used in only 12 comparisons and provided similar performance to machine learning methods. CONCLUSION: This systematic review suggests that machine learning methods, especially tree-based methods, may be superior to linear non-regularized methods for predicting ambient concentrations of NO2, UFPs and BC. Additional comparison studies using nonlinear, regularized and a wider array of machine learning methods are needed to confirm their relative performance. Future air pollution studies would also benefit from more explicit and standardized reporting of methodologies and results.

Exposure to ultrafine particles and the incidence of asthma in children: A population-based cohort study in Montreal, Canada.

Asthma is the most prevalent chronic respiratory disease in children. The role of ultrafine particles (UFPs) in the development of the disease remains unclear. We used a population-based birth cohort to evaluate the association between prenatal and childhood exposure to low levels of ambient UFPs and childhood-onset asthma. Methods: The cohort included all children born and residing in Montreal, Canada, between 2000 and 2015. Children were followed for asthma onset from birth until <13 years of age. Spatially resolved annual mean concentrations of ambient UFPs were estimated from a land use regression model. We assigned prenatal exposure according to the residential postal code at birth. We also considered current exposure during childhood accounting for time-varying residence location. We estimated hazard ratios (HRs) using Cox proportional hazards models adjusted for age, sex, neighborhood material and social deprivation, calendar year, and coexposure to ambient nitrogen dioxide (NO2) and fine particles (PM2.5). Results: The cohort included 352,966 children, with 30,825 children developing asthma during follow-up. Mean prenatal and childhood UFP exposure were 24,706 particles/cm3 (interquartile range [IQR] = 3,785 particles/cm3) and 24,525 particles/cm3 (IQR = 3,427 particles/cm3), respectively. Both prenatal and childhood UFP exposure were not associated with childhood asthma onset in single pollutant models (HR per IQR increase of 0.99 [95% CI = 0.98, 1.00]). Estimates of association remained similar when adjusting for coexposure to ambient NO2 and PM2.5. Conclusion: In this population-based birth cohort, childhood asthma onset was not associated with prenatal or childhood exposure to low concentrations of UFPs.

Opioid prescribing and dispensing: Experiences and perspectives from a survey of community pharmacists practising in the province of Quebec.

BACKGROUND: Canada leads in opioid prescription and consumption rates, and this has resulted in high levels of opioid-related morbidity and mortality. Pharmacists' input could contribute significantly to understanding the disadvantages of opioid prescribing and dispensing and improving the service. This study aimed to examine the experiences of community pharmacists in relation to opioid prescribing and dispensing, with a focus on optimizing collaboration and communication. METHODS: An online survey was performed among pharmacists from the province of Quebec, Canada, in 2016. Pharmacists were eligible if registered and working in community pharmacies. RESULTS: In all, 542 questionnaires were analyzed (participation rate of 8.1%). Pharmacotherapy-related problems were reported in at least 50% of opioid prescriptions: additional drug(s) required (reported by 30% of pharmacists), interaction(s) between opioid(s) and other drug(s) (16%), physician did not meet the general issuing standards for opioid prescriptions (26%) and patient had mild to moderate pain that was easily managed by a nonopioid analgesic (20%). Half of the patients were reported as requesting anticipated refills, possibly indicating abuse or poor pain control. Most pharmacists (89.6%) reported needing to contact physicians in 1 to 3 out of 10 opioid prescriptions, but many pharmacists (71.8%, often or very often) reported difficulties communicating with physicians. CONCLUSIONS: Pharmacists' observations of pharmacotherapy-related problems and patients' unusual behaviours reveal a significant number of issues related to opioid prescribing and dispensing in an outpatient setting. Improved collaboration between physicians and pharmacists appears mandatory to address the issues reported in this study.

Availability, Quality, and Relevance of Toxicogenomics Data for Human Health Risk Assessment: A Scoping Review of the Literature on Trihalomethanes.

Human health risk assessment (HHRA) must be adapted to the challenges of the 21st century, and the use of toxicogenomics data in HHRA is among the changes that regulatory agencies worldwide are trying to implement. However, the use of toxicogenomics data in HHRA is still limited. The purpose of this study was to explore the availability, quality, and relevance to HHRA of toxicogenomics publications as potential barriers to their use in HHRA. We conducted a scoping review of available toxicogenomics literature, using trihalomethanes as a case study. Four bibliographic databases (including the Comparative Toxicogenomics Database) were assessed. An evaluation table was developed to characterize quality and relevance of studies included on the basis of criteria proposed in the literature. Studies were selected and analyzed by 2 independent reviewers. Only 9 studies, published between 1997 and 2015, were included in the analysis. Based on the selected criteria, critical methodological details were often missing; in fact, only 3 out of 9 studies were considered to be of adequate quality for HHRA. No studies met >3 (out of 7) criteria of relevance to HHRA (eg, adequate number of doses and sample size). This first scoping review of toxicogenomics publications on trihalomethanes shows that low availability, quality, and relevance to HHRA of toxicogenomics publications presents potential barriers to their use in HHRA. Improved reporting of methodological details and study design is needed in the future so that toxicogenomics studies can be appropriately assessed regarding their quality and value for HHRA.

Barriers to the use of toxicogenomics data in human health risk assessment: A survey of Canadian risk assessors.

Regulatory agencies worldwide need to modernize human health risk assessment (HHRA) to meet challenges of the 21st century. Toxicogenomics is at the core of this improvement. Today, however, the use of toxicogenomics data in HHRA is very limited. The purpose of this survey was to identify barriers to the application of toxicogenomics data in HHRA by human health risk assessors. An online survey targeting Canadian risk assessors gathered information on their knowledge and perception of toxicogenomics, their current and future inclusion of toxicogenomics data in HHRA, and barriers to the use of such data. Twenty-nine (29) participants completed a questionnaire after 2 months of solicitation. The results show that the application of toxicogenomics data in Canada is marginal, with 85% of respondents reporting that they never or rarely used such data. Knowledge of toxicogenomics by Canadian risk assessors is also limited: about two-thirds of respondents (68%) were not at all or only slightly familiar with the concept. Lack of guidelines for toxicogenomics data interpretation, data quality assessment and on their use in HHRA, were found to be major barriers. In conclusion, there is a need for interventions aimed at facilitating the use of toxicogenomics data in HHRA, when available.