Authors: Erin N. Hulland, Marie-Laure Charpignon, Ghinwa Y. El Hayek, Angel N. Desai, Maimuna S. Majumder
Categories: Article, Google Search Trends, LGBTQ+ advocacy, disease nomenclature, misinformation, monkeypox, mpox, stigma
Source: medRxiv
Historically, many diseases have been named after the species or location of discovery, the discovering scientists, or the most impacted population. However, species-specific disease names often misrepresent the true reservoir; location-based disease names are frequently targeted with xenophobia; some of the discovering scientists have darker histories; and impacted populations have been stigmatized for this association. Acknowledging these concerns, the World Health Organization now proposes naming diseases after their causative pathogen or symptomatology. Recently, this guidance has been retrospectively applied to a disease at the center of an outbreak rife with stigmatization and mpox (f.k.a. ‘monkeypox’). This disease, historically endemic to west and central Africa, has prompted racist remarks as it spread globally in 2022 in an epidemic ongoing today. Moreover, its elevated prevalence among men who have sex with men has yielded increased stigma against the LGBTQ+ community. To address these prejudicial associations, ‘monkeypox‘ was renamed ‘mpox‘ in November 2022.
We used publicly available data from Google Search Trends to determine which countries were quicker to adopt this name change—and understand factors that limit or facilitate its use. Specifically, we built regression models to quantify the relationship between ’mpox’ search intensity in a given country and the country’s type of political regime, robustness of sociopolitical and health systems, level of pandemic preparedness, extent of gender and educational inequalities, and temporal evolution of mpox cases through December 2023. Our results suggest that, when compared to ‘monkeypox’ search intensity, ’mpox’ search intensity was significantly higher in countries with any history of mpox outbreaks or higher levels of LGBTQ+ acceptance; meanwhile, ‘mpox’ search intensity was significantly lower in countries governed by leaders who had recently propagated infectious disease misinformation.
Among infectious diseases with stigmatizing names, mpox is among the first to be revised retrospectively. While the adoption of a given disease name will be context-specific—depending in part on its origins and the affected subpopulations—our study provides generalizable insights, applicable to future changes in disease nomenclature.
Keywords: mpox, monkeypox, Google Search Trends, misinformation, stigma, LGBTQ+ advocacy, disease nomenclature
Historically, diseases have often been named after the animal that first presented with the pathogen (e.g., monkeypox, swine flu, bird flu); the location of first diagnosis or report (e.g., Spanish flu, Marburg virus, Ebola Zaire virus); or the discovering scientist or affected patient’s name (e.g., Hodgkin’s lymphoma, Alzheimer’s disease, Chagas disease). Since the release of specific guidelines in 2015, the World Health Organization (WHO) recommends avoiding such historical practices and instead favors focusing on symptomatology or the causative pathogen when naming new diseases.^1^ This recommendation is responsive to rising xenophobia and avoidance related to location-centric names (e.g., anti-Mexican sentiment following the "Mexican swine flu" moniker for the 2009 H1N1 pandemic^2^), culling of animal populations wrongly believed to be reservoirs for a disease (e.g., pig culling in Egypt due to the 2009 H1N1 "swine flu" pandemic^3^), and occasionally, the prejudicial past of a discovering scientist (e.g., Rett syndrome, named after an Austrian neurologist who was a member of the Nazi party in his youth^4^). In some cases, a disease name may not even accurately reflect its origins. This latter scenario is best illustrated by the Spanish flu, which likely originated in North America, but was first reported in Spain due to limited communication during World War I.^5^
The magnitude of the COVID-19 pandemic shined a bright light on pervasive xenophobia and racism tied to the origins of its causative pathogen—SARS-CoV-2—and on the importance of appropriate risk communication. While authors of scientific publications generally followed the new WHO guidance in naming the disease caused by the novel coronavirus (SARS-CoV-2), prominent figures in politics and the media have not similarly followed suit, often emphasizing the origins of COVID-19 by using inflammatory language to refer to the virus. One such example is former President Trump’s reference to SARS-CoV-2 as the "Chinese virus" in 2020.^6^ Given the use of such stigmatizing language in tweets and press briefings, both online anti-Asian sentiment and physical violence against Asians and Asian Americans rose dramatically.^6-12^
While the COVID-19 pandemic raged on, a simultaneous mpox (f.k.a. ‘monkeypox’) outbreak spread to over 50 countries in 2022.^13^ Previously, mpox was endemic to west and central Africa, with only a few sporadic outbreaks recorded in other countries up until 2022—during which an unprecedented epidemic, comprised of multiple country-level outbreaks, unfolded.^14-16^ While previous mpox outbreaks in endemic regions did not exhibit sexual transmission, this mode of transmission was observed for the first time during the 2022 epidemic.^17^ Given the prevalence of sexual transmission during this epidemic, similarities with the early HIV/AIDS epidemic were drawn, leading to increased stigmatization of gay men in particular.^18-20^ Because of the epidemic’s unprecedented global reach, the presentation and naming of the disease were more visible than ever before. The name ‘monkeypox’ presented not only a misnomer, as monkeys are highly unlikely to be the reservoir for mpox, but also perpetuated an offensive stereotype of African populations, among whom the bulk of mpox infections had previously occurred.^21-23^ Following a call by the WHO to rename the disease in August 2022, the name ‘monkeypox’ was formally changed to ‘mpox’ on November 28, 2022, allowing for one year of simultaneous use of the two names.^24,25^
The transition from ‘monkeypox’ to ‘mpox’ is the first known example of a formal name change for an infectious disease in the Internet Age, designed to explicitly address prejudice and misinformation. It echoes former experiences with AIDS in the early 1980s, when the disease was referred to as "Gay-Related Immune Deficiency" or GRID.^26,27^ Similar to mpox, which was named after its first identification in a captive cynomolgus monkey,^28^ this ‘GRID’ was named after its first presentation in gay men in Los Angeles in 1980,^29^ resulting in tremendous stigmatization of the LGBTQ+ community. Ultimately, in August 1982, the Centers for Disease Control introduced a non-stigmatizing name for this novel Acquired Immune Deficiency Syndrome (AIDS). Nevertheless, the damage associated with the initial name had long-term repercussions—and this stigma persists today.^30^ As such, the 2015 WHO recommendation endeavors to avoid the use of stigmatizing language for newly-named diseases, yet no formal and sufficiently broad guidance has been provided to date for retroactively changing existing disease names en masse, despite countless examples of existing disease names with prejudicial connotations that would require such intervention.^1^ In the absence of formal retrospective naming guidelines, this event not only provides an interesting case study but also creates a precedent for the WHO and its national counterparts to build upon.
Despite this formal guidance, the global adoption of the new term has been inconsistent. As such, our goal was to understand which factors influenced whether a country was a timely adopter of the name change from ‘monkeypox’ to ‘mpox’. In this study, we leveraged Google Search Trends (GST) to examine country-level search intensity of the terms ‘mpox’ and ‘monkeypox’ between November 2022 and December 2023, the time period over which the WHO recommended the dual use of both ‘mpox’ and ‘monkeypox’. Previously, GST has been used to study infectious diseases such as seasonal influenza, SARS-CoV-2, and measles, among others—from predicting and monitoring their spread, to understanding spatiotemporal variation in health-seeking behaviors, discovering new disease symptoms via search trends, developing tailored risk communication plans, and more.^31-40^
Our main analytic goal was to understand the drivers of adoption of the term ‘mpox’ at the country-level. To that end, we developed two key (1) a binary variable reflecting whether search intensity was higher for ‘mpox’ or ‘monkeypox’ during our study period; and (2) a continuous variable defined as the fraction of the search intensity for ‘mpox’ and ‘monkeypox’ attributable to ‘mpox’. We constructed regression models to associate these outcomes with static country-level factors such as type of political regime, robustness of sociopolitical and health systems, level of pandemic preparedness, extent of gender and educational inequalities, temporal evolution of confirmed local mpox cases during the 2022 epidemic, and distribution of mpox cases since discovery. Through this study, we hope to provide insights not only into country-level factors that were most conducive to the public’s adoption of the destigmatized name—as reflected by the intensity of Google search queries—but also into improving adoptions during future disease name changes.
We obtained the trajectory of Google search intensity for ‘mpox‘ and ‘monkeypox‘ over time through Google Search Trends (GST), a publicly-available online platform that provides anonymized, categorized, and aggregated data. We included 184 countries with available GST data on either term throughout the study period, out of 249 distinct countries and territories captured by the platform overall. Given that English is currently the lingua franca in science, thus serving as the language of disease communication and nomenclature, we relied on Google Search queries made in English in our primary analysis. However, we acknowledge the inherent power imbalance of Anglocentrism.^41-46^ Therefore, in an effort to also understand the dynamics of public interest in ‘mpox’ versus ‘monkeypox’ in countries where English is not the primary language, we considered the five other official WHO / United Nations (UN) Arabic, Chinese, French, Russian, and Spanish. We report the results of this sensitivity analysis, aimed at further understanding the intensity of searches that users made in their native language, in the Appendix.
GST returns the proportion of Internet search activity accounted for by each search term in a given location (e.g., metropolitan area, state, country) at a given point in time (e.g., day, week). This quantity, ranging between 0% and 100%, is defined as the number of term-specific searches in that location at that particular point in time divided by the total number of searches over that same region and timespan.^47,48^ By collecting data for both search terms simultaneously (i.e., ‘mpox’ and ‘monkeypox’), we were able to compare their respective search intensity on the same spatiotemporal scale for the period ranging from November 1, 2022 to December 1, 2023. In doing so, our intent was to capture one full year of Google search intensity data following the WHO announcement on November 28, 2022 and thus study the progressive adoption of the term ‘mpox’ by the general public.
We derived two outcome variables defined at the country level. Both variables were based on the average search intensity for the terms ‘mpox‘ and ‘monkeypox‘ in a given country over the study period. The first outcome, a binary variable, captured whether a country had greater average search intensity for ‘mpox‘ than for ‘monkeypox‘ (1 if yes, 0 otherwise), allowing us to distinguish ‘mpox‘ adopters from non-adopters among the 184 countries considered in our study. The second outcome, a continuous variable, was defined as the fraction of search intensity attributed to ‘mpox‘ over the study period (i.e., the ratio of the search intensity for ‘mpox‘ over the total search intensity for ‘mpox‘ and ‘monkeypox‘ combined, expressed as a percentage value); we refer to this variable hereafter as the ‘mpox’ search proportion. While we defined two distinct outcomes to test the sensitivity of our findings to the use of metrics with different granularity (i.e., binary vs continuous), our overall analytic goal was to identify the consistent drivers of ‘mpox’ adoption; thus, our results are presented and interpreted jointly.
In this study, we sought to identify country-level factors contributing to adoption of the term ‘mpox’ around the globe. To achieve this goal, we compiled a suite of 18 publicly-available country-level covariates to model their associations with ‘mpox‘ search intensity (Table 1). Given high LGBTQ+ stigma during the 2022 mpox epidemic, we included a country-level measure of LGBTQ+ acceptance, called the Global Acceptance Index (GAI).^49^ The GAI is a continuous variable ranging from 0 to 10. Since pandemics have become increasingly politicized, as exemplified by COVID-19 in the United States and Brazil,^50,51^ we accounted for various measures of a country’s current political context. Specifically, we relied on variables previously used to understand the COVID-19 pandemic,^52,53^ including the type of political regime of the country (four closed autocracy, electoral autocracy, electoral democracy, or liberal democracy);^54^ the Electoral Democracy Index, a continuous variable ranging from 0 to 1 that acts as a proxy for free and fair elections as well as voting rights;^54^ the Liberal Democracy Index, a continuous variable ranging from 0 to 1 that explores election freedom, civil liberties, and limits to executive power;^54^ and the Corruption Perceptions Index, a continuous variable ranging from 0 to 100 that measures the perceived levels of corruption in the public sector.^55^
Because the health system and its practitioners are often the first point of contact for disease presentation and reporting, we were also interested in understanding whether healthcare access and quality of care in a given country influenced the public’s perception and knowledge of a disease, and ultimately, its nomenclature. To this end, we included the Healthcare Access and Quality Index, a continuous variable ranging from 0 to 100, as a covariate.^56^ Moreover, we capitalized on the 2021 Global Health Security Index (GHSI)—a composite index comprising 171 variables across 37 indicators in total—to measure a country's preparedness to detect, prevent, and respond to future infectious disease outbreaks. Given the rise in emerging and re-emerging infections, we deemed it relevant to characterize each country’s approach to health communication, their strategies to mitigate disease spread, and their targeted efforts to protect at-risk populations. To this end, we focused on three GHSI components related to risk communications.^57^ In particular, we considered the GHSI aggregate risk communications score, as well as two related—but more specific—measures: the inclusiveness of a country’s risk communication plans and the prior use of infectious disease misinformation by a country’s government. The first was captured by GHSI question 3.5.1b, which focuses on an explicit intention to communicate with hard-to-reach populations based on language, rurality, and media access. The second was captured by GHSI question 3.5.2b, which characterizes whether a country’s senior leaders (i.e., presidents or prime ministers) had shared misinformation or disinformation about infectious diseases in the previous two years. Additionally, we leveraged four GHSI components related to inequalities in mobile phone and Internet access, both between genders within a given country and among countries. In the first case, the ratio of access to mobile phones and Internet between males and females in a given country provides an indicator of gender-based disparities; in the second case, country-level indicators of mobile phone and Internet access capture Internet and technology penetration in a country. In both cases, limited access to mobile phones and the Internet ultimately influence who is able to use Internet search platforms like Google Search.
We also accounted for disparities in country wealth, reasoning that the average income level is associated with the average level of English proficiency of the population and thus their likelihood of searching for English terms.^58^ Additionally, given that the 2022 mpox epidemic was heavily concentrated in high- and upper-middle-income countries, we wanted to control for the distribution of country wealth in our analyses.^59^ For these reasons, we included the gross domestic product (GDP) per capita as a covariate.^60^ Reasoning that a country’s average educational attainment would be associated with the public’s propensity to search for ‘mpox’, we similarly explored between-country educational disparities using the mean number of years of schooling among women as a proxy; we specifically chose education levels among women rather than among men to capture a greater heterogeneity among countries due to inequitable access to education for women around the world. Last, we considered the impact that the recent 2022 mpox epidemic may have had on the adoption of ‘mpox’ by the public. More specifically, to assess the relative timing of the epidemic in a given country, we calculated the number of days between the first confirmed mpox case worldwide (in the United Kingdom on May 6, 2022)^61^ and that country’s first confirmed case. In addition, we created a binary indicator of whether a country had ever had a confirmed case of mpox in humans or mammals since it was first discovered in humans in 1970, in order to gauge previous levels of exposure to the disease and its name.^59,62,63^
Given varying patterns of missingness in the covariates across the 184 countries included in the analysis (Appendix Figure 2), we used a multiple imputation approach with predictive mean matching over ten imputations. We considered three different combinations of data selection and imputation (1) selecting countries with no missing data and performing a complete-case analysis with these resulting 154 countries; (2) selecting countries that either had complete data or were missing only the GAI and imputing this variable where necessary, yielding a sample size of 166 countries; and (3) imputing all variables with missing data, resulting in a total of 184 countries. Results presented in the main text capitalize on the fully imputed covariate dataset (i.e., Approach 3); results relying on the two other imputation strategies (i.e., Approaches 1 and 2) are presented in the Appendix. Transformations made to the covariate data in pursuit of normalizing the distribution are available in the Appendix Section 3.1.
Following data extraction and pre-processing, we built univariable and multivariable regression models separately for each of our two outcomes. The binary outcome—whether a country was an adopter (1) or not (0)—was modeled using logistic generalized linear models with time-invariant covariates. The continuous outcome—’mpox’ search proportion—was modeled using generalized linear regression models with the same time-invariant covariates. First, we evaluated the strength of the simple association between each outcome of interest and each of the 18 covariates using univariable models. We retained all variables significant at p<0.1, removed those that were highly collinear, and subsequently implemented multivariable models adjusted for the remaining subset (Table 1). We accounted for uncertainty over the ten iterations of the imputation process by drawing from a multivariable normal distribution; then, we summarized across all ten draws using the median and the 2.5th and 97.5th percentiles to obtain a 95% uncertainty interval.
In addition to our secondary analyses based on Google searches made in languages other than English (Appendix Figures 5 and 6), we conducted six sensitivity analyses to assess the robustness of our results to data and modeling decisions. First, we compared our main results with those emanating from the two alternative imputation strategies described above, namely the complete case analysis (Approach 1- Appendix Figures 7 and 8) and the imputation of only the GAI variable (Approach 2 - Appendix Figures 9 and 10); descriptive statistics for these imputation approaches are also visible in Table 1. Second, we evaluated the impact of restricting our dataset to only the countries that ever had a confirmed case of mpox in order to understand the role of exposure on the adoption of the term ‘mpox’ by the public (Appendix Figures 11 and 12). Third, we considered the impact of using a more stringent significance threshold, at p<0.05 rather than p<0.1, for covariate inclusion into our multivariable regression models (Appendix Figure 13). Fourth, we used a Bonferroni correction to account for post-hoc testing of multiple hypotheses at once (Appendix Figures 14 and 15). Finally, we implemented a Tobit model, well-suited for zero-inflated data, to address a large proportion of zeros in our continuous outcome measuring the ‘mpox’ search proportion (Appendix Figure 16).
All analyses were conducted using R version 4.3.1^64^. This research was supported in part by the National Institute of General Medical Sciences, National Institutes of Health (R35GM146974); the National Science Foundation (SES2200228, SES2230083, & IIS2229881); and the MIT-Harvard Broad Institute Eric & Wendy Schmidt Center. The funding sources had no involvement in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.
In analyzing the intensity of Google searches for ‘mpox’ and ‘monkeypox’ across over 180 countries worldwide, we identified three consistent country-level factors that were significantly associated with adoption of the term ‘mpox’: any history of mpox in the country, greater LGBTQ+ acceptance, and senior leaders who did not recently propagate infectious disease misinformation; our results were robust to the imputation strategy chosen. We present key summary statistics (e.g., median and interquartile range or IQR) for all measured covariates, both pre- and post-imputation, in Table 1. Additionally, we examine a cross-comparison of the three imputation strategies described in the Methods section to assess their relative impact on our model covariates and outcomes (Table 1).
In Figure 1a, nine variables were significantly associated with our binary outcome in our univariable model; meanwhile, when considering our continuous outcome—’mpox’ search proportion—there were 16 significant variables in the univariable model (Figure 2a). When we included only those variables with univariate significance p<0.1 in the two multivariable models, we observed three consistent significant relationships across our two (1) a positive relationship with countries who had ever reported an mpox case, either as part of the 2022 epidemic or in a prior outbreak; (2) a positive relationship with LGBTQ+ GAI score; and (3) a negative relationship with countries with senior leaders who had recently employed misinformation about any infectious disease (Figures 1b and 2b).
Figure 1 - Country-level factors associated with ‘mpox’ adopter countries (binary outcome) in univariable (a) and multivariable (b) modelsVariables that were statistically significantly associated with the binary outcome of interest at alpha = 0.1 in the corresponding univariable models were included in the multivariable model. Highly collinear variables were removed before fitting the multivariable model. Triangle markers in panel (a) denote covariates removed due to high multicollinearity; square markers indicate those covariates removed due to non-statistical significance at alpha = 0.1 in their corresponding univariable model; circle marker indicate covariates included in the multivariable model.
Figure 2 - Country-level factors associated with ‘mpox’ search proportion (continuous outcome) in univariable (a) and multivariable (b) modelsVariables that were statistically significantly associated with the continuous outcome of interest at alpha = 0.1 in the corresponding univariable models were included in the multivariable model. Highly collinear variables were removed before fitting the multivariable model. Triangle markers in panel (a) denote covariates removed due to high multicollinearity; square markers indicate those covariates removed due to non-statistical significance at alpha = 0.1 in their corresponding univariable model; circle marker indicate covariates included in the multivariable model.
Our study suggests that countries with higher LGBTQ+ acceptance, as measured by the GAI, were more likely to adopt the term ‘mpox’ over the term ‘monkeypox’ in the year following the WHO recommendation of the name change. This result suggests that countries where the public is more accepting of queer communities were also more likely to demonstrate a more thoughtful use of destigmatizing language when conducting Google searches. Importantly, this finding was consistent across all sensitivity analyses we conducted, emphasizing its robustness to the choice of imputation strategy and model type. In contrast to historic mpox outbreaks that predominantly affected west and central Africa and had no reported sexual transmission, the 2022 multi-country epidemic of mpox has largely been categorized as a disease among the gay, bisexual, and other men who have sex with men (GBMSM) community.^66,67^ As a result, the GBMSM community has faced backlash and stigma—in addition to the fear of mpox itself.^19,20,68,69^ Therefore, our study underscores the need to increase awareness and acceptance of marginalized populations. Indeed, as exemplified in previous outbreaks—most notably HIV/AIDS— stigma associated with sexual transmission not only impacts quality of life, but can also limit a person’s access to care and reduce their propensity to accurately report their behaviors. These negative repercussions can in turn lead to delayed treatment, poorer health outcomes, and sustained disease propagation.^70-73^ Thus, in order to increase awareness and acceptance of marginalized populations going forward, strong risk communications combating bigotry, prejudice, and stigma are critical early in infectious disease crises.^74^
Another strong correlate of ‘mpox’ search intensity was countries that had ever had a confirmed case of mpox—in humans and other mammals alike. We hypothesize that people living in places with previous cases had increased exposure to news and social media about mpox, increasing awareness of the new name. Previous research about the impact of news media on infectious diseases—including mpox, H1N1, Ebola, and COVID-19—has demonstrated that increased exposure to quality information through news and social media can increase vaccine uptake,^75,76^ reduce disease transmission,^77^ and encourage protective behaviors.^78-80^ Our work reflects the need for awareness campaigns during infectious disease crises, emphasizing proper nomenclature as well as providing high-quality information on disease spread and populations most at-risk. Future research evaluating media communications regarding the mpox name change may provide valuable insights into how the news media referred to the disease over time; whether there were any political, geographical, or temporal trends; and what messaging at what time point corresponded to higher uptake of ‘mpox’. In doing so, more specific guidance can be generated for future media campaigns on infectious disease name changes.
We identified a third strong correlate of ‘mpox’ in countries where senior leaders had recently propagated infectious disease misinformation, adoption was significantly lower. Such misinformation has been increasingly identified as a barrier to vaccines,^81-84^ seeking timely care,^81,85,86^ and positive behavior change; ^81,87^ has eroded trust in governments and authorities;^88-90^ and has ultimately worsened health outcomes.^86,89,91^ During the COVID-19 pandemic, misinformation was employed by politicians in the US and abroad to promote the use of untested medicine such as Ivermectin, Azithromycin, Chloroquine and Hydroxychloroquine;^92-94^ the avoidance of many public health guidelines like vaccination or mask wearing;^95-97^ and dangerous practices like ingesting or injecting bleach,^98,99^ resulting in surges in product purchases, price hikes, and in some cases, serious injury or death.^92,98-101^ More recently, a sentiment analysis of tweets suggested that mpox misinformation contributed to heightened stigma of LGBTQ+ populations, resulting in increased violence, harassment and isolation.^74^ Several strategies have been employed to combat the growing threat of emphasizing quality sources of information; capitalizing on trusted leaders, officials, and scientists to convey accurate information; correcting and calling out misinformation; and increasing frequency of both proactive, accurate messaging and correcting inaccurate information.^102^ Currently, consequences incurred by leaders or politicians who spread misinformation vary by country and region. First, they may face financial and legal consequences, including penalties such as fines and, in some cases, imprisonment. Second, social media platforms themselves may take action and either restrict user privileges (e.g., retweeting, sharing, or posting) or ban such leaders outright; yet many countries still struggle to walk the line between action and censorship.^103,104^ While elected officials who propagate misinformation should undoubtedly be sanctioned, no one course of action is likely to suit every country. In the future, researchers should — in collaboration with dedicated counter-misinformation oversight committees, when possible — seek to understand how effective each strategy is at reducing the spread of misinformation in a given country, providing evidence to guide local decision making.
There are several limitations to this study. First, our main analysis relied on English as the lingua franca of scientific communication and our sensitivity analysis considered the five additional UN languages. These five languages—while widely used—account for only a small fraction of all the languages spoken across the world; as such, we are likely missing search terms for ‘monkeypox’ expressed in other languages, resulting in an incomplete capture of Google search interest in the disease globally. While we acknowledge the diversity of languages worldwide and its importance in public health, around 4 billion people speak at least one of the 6 UN languages as a primary or secondary language, representing about half of the world’s population.^105,106^
Second, consistent with the WHO nomenclature announcement, we only searched the single terms ‘mpox’ and ‘monkeypox’ rather than the composite terms ‘mpox virus’ and ‘monkeypox virus’, which remains the official taxonomic name. However, given the boolean methodology used by GST, data collected for the term ‘monkeypox’ or ‘mpox’ constitute a superset that would also account for the terms ‘monkeypox virus’ and ‘mpox virus’.^107^
Third, at the time of our analysis, foreign language data for two countries—Ecuador and Madagascar—were unavailable. To address this limitation, we attempted a counterfactual sensitivity analysis for the binary outcome, generating all permutations of the possible second-language findings and observed largely similar results to our main analyses—suggesting the robustness of our key findings (Appendix Section 3.7 and Figure 17).
Fourth, this study employs Internet searches conducted by users via Google Search alone, thus excluding those searches attributable to other search engines. Future research examining searches for ‘mpox’ and ‘monkeypox’ on other search engines (e.g., Bing, DuckDuckGo, Yahoo) would thus provide valuable comparative insights. Importantly, over 90% of Internet users across the globe use Google as their primary search engine—with the exception of China and Russia, suggesting GST are capturing the majority of global Internet searches.^108,109^
Last, we employed a suite of 18 covariates in our regression models, with varying degrees of missingness. We used multiple imputation with as many complete variables as possible, but acknowledge differences in resulting sample sizes and covariate data distributions between the complete-case analysis (Approach 1, n=154), the GAI-only imputation analysis (Approach 2, n=166), and the full imputation analysis (Approach 3, n=184). The most notable difference observed was GDP. Since this covariate was available for all 184 included countries, we could contrast its overall distribution to that in the analyses restricted to 154 (Approach 1) and 166 (Approach 2) countries, respectively. Specifically, we observed a lower median GDP in both the GAI-only imputation analysis and the complete-case analysis. This result suggests that health system quality, pandemic preparedness, gender equitability, and social acceptance, which are all correlated with GDP, were likely not missing at random. Nonetheless, all of our sensitivity analyses of imputation strategies yielded similar main results, reinforcing the robustness of our conclusions.
We have seen a surge of zoonotic disease transmission, outbreaks, and pandemics in this century so far; this trend is likely to continue with increasing population connectivity and mobility, deforestation and development encroaching on zoonotic reservoirs, and climate change influencing transmission. As such, the potential for stigmatizing language and xenophobia, as well as their downstream impacts and outcomes, are becoming increasingly visible. We need to plan now for ways to counter this stigma, and learning from those countries most successful in instituting ‘mpox‘ in lieu of ‘monkeypox‘ is a good way to start.
We would like to thank Paula Rodriguez Diaz and Robyn Correll for their assistance in understanding ‘mpox’ trends in South America and the translation of ‘monkeypox’ in Spanish; Nilufar Qahorova for her assistance in translating ‘monkeypox’ to Russian; and Zhanzhan Zhao in her assistance in translating ‘monkeypox’ to Chinese.
All analyses were conducted using R version 4.3.1^64^. This research was supported in part by the National Institute of General Medical Sciences, National Institutes of Health (R35GM146974); the National Science Foundation (SES2200228, SES2230083, & IIS2229881); and the MIT-Harvard Broad Institute Eric & Wendy Schmidt Center. The funding sources had no involvement in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.