Temporal and seasonal patterns of online interest in ibuprofen across fourteen European countries: a five-year infodemiological study with robust time-series analysis and chronological predictive validation

Abstract

Ibuprofen is widely used as a non-prescription analgesic and antipyretic, yet online information-seeking remains incompletely characterized. We assessed trends, seasonality, adjusted cross-country co-movement and out-of-sample predictability of Google search attention across 14 European countries. Monthly Google Trends relative search volume (RSV; 0-100) for the harmonized generic search term “ibuprofen” was extracted on 20 July using the Web Search property and all categories, each country being queried separately so that every series is normalized within country (July 2021 to June 2026; 60 months per country). Trend inference used calendar-month-adjusted linear regression with Newey-West standard errors and Benjamini-Hochberg false-discovery-rate correction. Seasonality was evaluated using robust STL decomposition and winter-versus-summer contrasts. Cross-country co-movement was assessed on STL remainders. Seasonal ARIMA forecasts were validated by rolling origins against seasonal naive. Random Forest and Gradient Boosting were tested on a final 12-month chronological holdout and compared with harmonic regression and seasonal naive. Across 840 country-month observations, eight countries showed significant increases: Hungary, Poland, Turkey, Sweden, Belgium, the United Kingdom, the Czech Republic and the Netherlands; none showed a significant decline. Seasonal peaks fell between December and February in all 14 countries (December in 10) and troughs between April and July (June in 7); seasonal strength ranged from 0.20 (Norway) to 0.72 (Poland). After removing trend and seasonality, the strongest residual correlations were Germany-Austria (r=0.79) and Poland-Czech Republic (r=0.76). ARIMA outperformed seasonal naive only at 1 month (RMSE 10.56 vs 11.18), but not at 3, 6, or 12 months. On the chronological holdout, harmonic regression performed best (R²=0.242; RMSE=1.221); Random Forest (R²=-0.026; RMSE=1.420) and Gradient Boosting (R²=-0.030; RMSE=1.423) did not outperform seasonal naive (R²=-0.005; RMSE=1.406). Ibuprofen search attention increased in a subset of countries and showed recurrent winter seasonality. Predictive performance was horizon-dependent and modest. Google Trends normalization means findings reflect within-country search attention rather than absolute cross-country interest or medication use.

Keywords

• Ibuprofen
• non-steroidal anti-inflammatory drugs
• infodemiology
• Google Trends
• digital epidemiology
• machine learning

Ethics Approval

Not required as this study used only aggregated, anonymized and publicly available data and did not involve human participants or identifiable information

Data Availability Statement

The source data were derived from the publicly accessible Google Trends platform. Because Google Trends uses sampling and dynamic normalization, values regenerated at a later date may differ slightly from those in the analyzed dataset. The complete query specification required to regenerate the raw series is provided in Supplementary Table S1. The processed country-month dataset and the full analysis code are available from the corresponding author on reasonable request.