Time series analysis in environmental epidemiology: challenges and considerations.

In environmental epidemiology, time series analyses represent a widely used statistical tool. However, though being commonly used, there is soften confusion regarding the specific requirements, such as which link function might be most appropriate, when or how to control for seasonality or how to account for lags. The present overview draws from experiences in other disciplines and discusses the proper execution of time series analyses based on considerations that are relevant in environmental epidemiology. Time series analysis in environmental epidemiology focuses on acute events caused by short-term changes in exposure. These exposures should be fairly wide-spread affecting a large number of persons, usually all inhabitants of a political entity. Pollutants in air or drinking water as well as meteorological factors serve as typical examples. Despite the many time series analyses performed world-wide, some health effects that would lend themselves to that approach are still under-explored. This would include also some neurological and psychiatric endpoints. Int J Occup Med Environ Health. 2023;36(6):704-16.

Temporal patterns of weekly births and conceptions predicted by meteorology, seasonal variation, and lunar phases.

BACKGROUND: Data reporting the number of births given by women resident in Vienna for each week between 1999 and 2019, and the duration (in weeks) of their pregnancies, were used to estimate the week of conception. When preliminary examinations detected cyclical patterns for births and conceptions, a series of Poisson regressions for births and conceptions were performed to assess whether these cycles could be associated with lunar or solar phases. METHODS: To that end 2 sine-cosine functions, 1 with a wavelength of 1 year (365.25 days) and 1 with a wavelength of 29.529 days, the latter resembling the average length of a lunar cycle, were constructed. In addition, also average weekly temperatures were included in the models. RESULTS: Same week temperature was a strong non-linear predictor for both births and conceptions. Extreme temperatures, especially hot temperatures, increased the number of births while the numbers of successful conceptions were decreased by extreme temperatures. Regarding annual variation in conceptions, the lowest values were found in May and the highest in late autumn and early winter. Therefore, the highest numbers of births are observed in the summer. As per lunar variations, the highest numbers of conceptions occurred at the full moon and the highest numbers of births at the waxing crescent to first quarter moon. DISCUSSION: The latter results should be treated with caution, as they are not hypothesis driven. Furthermore, literature reports on this topic are conflicting. Nevertheless, these findings will be useful in further analyses examining air pollution effects.