Analyzing Seasonal Variations in Suicide With Fourier Poisson Time-Series Regression: A Registry-Based Study From Norway, 1969–2007

Abstract

Seasonal variation in the number of suicides has long been acknowledged. It has been suggested that this seasonality has declined in recent years, but studies have generally used statistical methods incapable of confirming this. We examined all suicides occurring in Norway during 1969-2007 (more than 20,000 suicides in total) to establish whether seasonality decreased over time. Fitting of additive Fourier Poisson time-series regression models allowed for formal testing of a possible linear decrease in seasonality, or a reduction at a specific point in time, while adjusting for a possible smooth nonlinear long-term change without having to categorize time into discrete yearly units. The models were compared using Akaike's Information Criterion and analysis of variance. A model with a seasonal pattern was significantly superior to a model without one. There was a reduction in seasonality during the period. Both the model assuming a linear decrease in seasonality and the model assuming a change at a specific point in time were both superior to a model assuming constant seasonality, thus confirming by formal statistical testing that the magnitude of the seasonality in suicides has diminished. The additive Fourier Poisson time-series regression model would also be useful for studying other temporal phenomena with seasonal components.

Chris Cornell, the Black Hole Sun, and the Seasonality of Suicide

“…Goodwin and Jamison [7] reviewed 61 studies in the northern hemisphere and found a striking May peak in suicides and a less common, smaller bimodal October peak. Since the review of Goodwin and Jamison [7], there have been dozens of additional studies that have replicated consistent late spring peaks in suicide in nearly all of the largest datasets, including in the northern hemisphere (in those studies that reported both a monthly peak and a total n for suicides), the USA (May; n = 649,843, the present study), Japan (April; n = 501,950 [29]), Sweden (May; n = 161,182 [30]), Italy (May; n = 141,180 [31]), Finland (May; n = 96,159 [30]), England (January; n = 92,909 [32]), Austria (May; n = 67,741 [33]), Mexico (May; n = 64,298 [34]), Spain (July; n = 42,905 [35]), Denmark (May; n = 37,987 [36]), Switzerland (May/June; n = 37,518 [37]), Norway (May; n = 20,156 [38]), China (May; n = 14,450 [39]), and Ireland (April; n = 7,899 [40]). In the southern hemisphere, reciprocal spring peaks have been found in Australia (November; n = 57,936 [41]), South Africa (October; n = 16,389 [42]), and São Paulo, Brazil (November; n = 6,916 [43]).…”

“…Statistically significant increases in the amplitude of the seasonal suicide rhythm have also been found in Ireland (1980–2000) [40] and Australia (1970–1999) [41]. By contrast, the amplitude of the seasonal suicide rhythm did not change significantly in either Austria (1970–2008) [33] or China (1991–2009) [39], whereas in Norway (1969–2007) [38] and possibly Slovenia (1971–2002) [45] the amplitude decreased significantly. Clearly, the factors that contribute to these disparate temporal variations in the amplitude of the seasonal suicide risk factor are of great interest [8].…”

Early Impact of the Affordable Care Act on Uptake of Long-acting Reversible Contraceptive Methods

“…In addition, because we noted what appeared to be seasonal variation in LARC uptake (with drops in the first quarter of each year), we added dummy variables to adjust for quarter/season. 23 …”

Bayesian Model Search for Nonstationary Periodic Time Series

“…Bayesian inference can, in principle, be achieved in a similar way as described in the paper, namely by iterating between segment and change-point model moves, where the formulation of the acceptance probabilities and some proposal distributions need to be modified accordingly. We believe that such an extension would find use in several ranges of applications, for example in studying population cycles in ecology and epidemiology, where the abundance of species are measured as count variables (White & Bennetts 1996, Bhaskaran et al 2013, Bramness et al 2015.…”