Search trends preceding increases in suicide: A cross-correlation study of monthly Google search volume and suicide rate using transfer function models

Abstract

BackgroundVarious associations between monthly Google search volumes (MGSVs) and monthly suicide rates (MSRs) have been reported. However, these studies often analyzed a limited number of search terms using suboptimal statistical methods. While controlling for spurious associations, this study examined a wide array of suicide-related search terms to elucidate if their MGSVs correlated with future MSRs. MethodsMGSVs of 111 candidate suicide-related terms were calculated by averaging 10 time-series data per term obtained from Google Trends. Box-Jenkins transfer function modeling was applied to time-series data of MGSV and MSR among the total, male, and female populations of the United States between 2004 and 2017. Cross-correlation coefficients between MGSVs and MSRs were calculated at lags -3, -2, and -1. Sensitivity analysis identified cross-correlations whose direction and significance (p<0.05) persisted in two other time spans: 126 and 84 months. ResultsEighty-nine terms were analyzed. MGSVs of 31 terms significantly correlated with MSRs in the total, male, or female population. In the sensitivity analysis, three terms stably exhibited significant positive correlation: “generalized anxiety disorder” (total; lag -3), “anxiety disorder” (total and male; lag -3), and “laid off” (total, male, and female; lag -2). The term sleep problem (total and female; lag -1) consistently showed significant negative correlations. LimitationsSex- or age-specific search-volume data, lags of less than a month, and potential confounding factors of MGSV and MSR were not explored. Conclusionstrends in MGSV of four terms tend to precede changes in MSR. These terms may enable more accurate forecasting of future suicides.