The highly contagious nature of severe acute respiratory syndrome coronavirus 2, leading to the coronavirus disease 2019 (COVID-19) pandemic, necessitated the enactment of public health actions on an unprecedented scale in US history. Closure of all nonessential businesses along with strict social-distancing measures were instituted to curb disease transmission across the nation. After gradually closing some businesses and schools, the state of New York instituted New York State on PAUSE, a statewide lockdown, effective on March 22, 2020. Social isolation, restrictions of activities, workplace closures, as well as associated financial losses and the fears of COVID-19 may place a considerable psychological burden on people. Literature on the effects of quarantine highlights their negative consequences on mental health.1 To better understand mental health concerns during the New York COVID-19 lockdown, we analyzed trends in internet searches for mental health issues.
This study was conducted in accordance with the Declaration of Helsinki; thus, this analysis of publicly available data with no possibility of identifying individuals did not require ethical approval by an institutional review board. By adopting an approach previously used to monitor suicide search queries,2 we compared observed search volumes of selected mental health issues during the full quarantine lockdown in New York with their expected volumes, based on historical search data. Using Google Trends, we extracted trends for the topics suicide, anxiety, panic attack, insomnia, and depression, specifying New York State as the location of the search behavior. By selecting this specific location in Google Trends, we restricted searches to people in New York State at the time of the search based on a determination by Google; these determinations are made by using their users’ location data (eg, IP address, location history, previous search activity, GPS data). We examined relative search volumes (RSV), which represent absolute search volumes standardized from 0 to 100, and forecasted the weekly RSV from March 22, the beginning of New York State on PAUSE, until May 14, 2020, a day before some New York regions started phased reopening. Weekly RSV forecasts were computed, along with their 95% bootstrapped prediction intervals (PI), using the ARIMA algorithm implemented in the forecast package in R software, version 3.5.3 (R Foundation).
Searches for anxiety significantly increased following March 22, 2020, and remained significantly higher than expected for 3 consecutive weeks, during which searches were on average 18% (95% PI, 5%-29%) greater than expected (Figure 1). Searches for panic attack soared during the first week of the lockdown: search volumes increased 56% (95% PI, 37%-97%). They subsequently remained significantly higher for 5 consecutive weeks. During the entire lockdown, searches for insomnia were on average 21% (95% PI, 1%-55%) higher than expected. Suicide and depression (Figure 2) did not exhibit significantly different RSV from what was expected. There was a spike in searches for suicide during the week of April 26 through May 2. This brief increase occurred following the publicized suicide of a New York physician,3 and search levels returned to normal the following week.
Searches for anxiety, panic attack, and insomnia rose significantly during the lockdown, then eventually reverted to their mean. Assuming this is indicative of significant increases in these symptoms, this suggests the need to provide focused competent online and telemedicine services during quarantines. With the possibility of a second wave of severe acute respiratory syndrome coronavirus 2 looming and a return to lockdowns, we should ensure that individuals searching the internet can rapidly obtain reliable information on strategies to cope with distress and online help from authoritative sources.
Searches for depression did not increase during the lockdown. However, depression could increase if the quarantines return or as a result of the economic turmoil and unemployment that might ensue. Suicide queries did not increase during the lockdown. This is consistent with reports from the New York City crisis line, NYC Well, which despite a 40% increase in calls during this period, experienced no increase in the number of callers with suicidal ideation and no increase in emergency suicide rescues (John Draper, PhD, email communication, June 3, 2020). This might be explained by a pulling-together effect of feeling solidarity with others facing the same collective experience.4,5 Nevertheless, suicide rates and hospitalizations should continue to be closely monitored because they are associated with higher unemployment6 and increased alcohol and firearm sales.5
Google Trends could help monitor population mental health problems in real time. However, trends in searches have several limitations. We do not know if people were searching for themselves or for others. Number of searches may return to normal not because anxiety, panic attacks, or insomnia have decreased, but rather because people had already searched the internet for help. Nonetheless, posting useful resources for persons in distress is a feasible, inexpensive solution to help support those experiencing distress in the pandemic.
Accepted for Publication: June 9, 2020.
Corresponding Author: Stefan Stijelja, BS, Université du Québec à Montréal, Montreal, CP 8888, Succ Centre-Ville, Montreal H3C 3P8, Canada (stijela.stefan.2@uqam.ca).
Published Online: October 5, 2020. doi:10.1001/jamainternmed.2020.3271
Author Contributions: Mr Stijelja had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Stijelja.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Stijelja.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Stijelja.
Administrative, technical, or material support: Mishara.
Supervision: All authors.
Conflict of Interest Disclosures: None reported.