COVID-19 大流行對精神健康的影響：使用42國Google搜尋資料進行時間趨勢分析

Abstract

研究背景： 在COVID-19 大流行期間，疫情對心理健康造成的負面影響受到重視，然而疫情間許多研究採用便利性樣本，受選擇性偏差影響，有代表性的問題，研究對象僅限於特定地區或國家，因此本篇研究透過網路搜尋引擎資料資料進行分析比較，了解本次疫情對不同國家之人群造成的心理健康影響。 研究方法： 本研究收集來自Google Trends資料庫42國「焦慮」、「失眠」、「孤獨」、「憂鬱」及「自殺」五個字詞的搜尋變化，採用季節性差分整合移動平均自迴歸模型（seasonal-ARIMA model），以2016至2020年共217週之資料預測2020年3月至2021年2月共53週之搜尋量，了解疫情期間各字詞之搜尋量與預期值相比之變化；並使用斯皮爾曼等級相關（Spearman Rank Correlation）分析字詞搜尋量增加之週數與各國GDP 成長率、失業率以及吉尼係數之關係；並以隨機效應模型（Random effects model）分析各字詞53週實際值與預測值之差值與COVID-19確診率、死亡率及疫情間人流移動變化之關係。 結果： 本篇研究所納入的五個字詞「焦慮」、「孤獨」、「失眠」、「憂鬱」、「自殺」在42個國家中呈現隨著時間在疫情爆發後的53週呈現不同的變化。「焦慮」一詞之搜尋量增加的平均週數最高(6.86週)，其次為「失眠」（5.05週），「自殺」（1.57週），「孤獨」（1.00週），與「憂鬱」(0.67週)。相較下，「憂鬱」搜尋量下降的平均週數為3.1週。相關分析顯示GDP成長與吉尼係數與「焦慮」和「失眠」之搜尋量增加週數有關。迴歸分析顯示，行動限制與「焦慮」、「孤獨」、「失眠」之搜尋量增加有關，而疫情爆發的嚴重度(以COVID-19確診率或死亡率為指標)與「焦慮」、「失眠」和「憂鬱」的搜尋量增加有關。 結論： 焦慮和失眠症狀可能是COVID-19 疫情第一年常見的心理健康問題。儘管移動限制政策有助於控制疫情，但各國政府仍應警惕其對大眾心理健康潛在的負面影響，而充分的疫情控制有助於保護人們的心理健康。在全球突發的公共衛生事件期間，Google Trends資料有助於即時監測大眾的心理健康。

Introduction: There were concerns that the COVID-19 pandemic would adversely impactpopulation mental health. However, most surveys used non-probability andconvenience samples, which were susceptible to selection bias, and were restricted tos pecific communities or countries. This study aimed to investigate the impact of the COVID-19 outbreak on population mental health across different countries using population-based internet search data using mental health-related keywords. Methods: We extracted Google Trends data from 42 countries using five search terms:“anxiety”, “insomnia”, “loneliness”, “depression” and “suicide”. We used seasonal Autoregressive Integrated Moving Average (ARIMA) model to estimate the expected Google search volume during the COVID-19 pandemic (March 2020 – February 2021) based on pre-pandemic trends (January 2016 – February 2020). We first conducted Spearman correlation analysis to investigate the association between the total number of weeks with a change in Google search volume (i.e., whether the observed search volume was above or below the 95% confidence intervals of the expected volume) and GDP growth rate, unemployment rate, and Gini index across countries. We then applied random-effects models to investigate the associations of the weekly Google search volume difference (i.e., the observed value minus the expected value) with the weekly number of COVID-19 confirmed cases, the weekly number of COVID-19 death, and the weekly residential mobility change across the 42 study countries. Results: Changes in the Google search volume of five mental health related terms showed varying trends across the 42 study countries during the first 53 weeks of the COVID-19 outbreak. Search for “anxiety” had the highest mean number of weeks with increased search volume (6.86), followed by “insomnia” (5.05), “suicide” (1.57), “loneliness” (1.00), and “depression” (0.67). By contrast, search for “depression” also had a mean number of weeks with decreased search volume of 3.10. Correlation analyses showed that GDP growth and Gini index were associated with searches for “anxiety” and “insomnia”. Regression analyses showed that restriction on movement was associated with increased searches for “anxiety”, “loneliness”, and “insomnia” but decreased searches for "depression". The severity of the COVID-19 outbreaks, indicated by rates of confirmed cases or mortality, was associated with increased searches for “anxiety”, “insomnia”, and “depression”. Conclusions: Anxiety and insomnia symptoms might be the common mental health problems during the first year of the COVID-19 pandemic. Although movement restriction policies would help control the outbreak, governments should be alert to the potential negative mental health effects on the population. Adequate virus outbreak control would also protect the population’s mental health. During the worldwide public health emergency, Google Trends data could be useful to monitor population mental health in real time.