Abstract
Background
Mpox (formerly monkeypox) has an incubation period of 3-17 days, creating a window when infections are present but undetected. Pre-exposure prevention depends on the two-dose Mpox vaccine, which requires about six weeks from the first injection to reach peak protection. Therefore, early and affordable surveillance capable of identifying outbreaks prior to the development of immune protection is critical. Digital traces such as Google search activity may provide early signals of epidemic dynamics, but their utility for Mpox in real-world contexts remains insufficiently characterized.
Methods
We assembled monthly country-level Mpox cases and deaths from the World Health Organization global trends dashboard from 1 May, 2022 to 30 November, 2024 and matched them to country-specific Google Trends relative search volume (RSV) for “Monkeypox.” A total of 96 countries were analyzed, including 38 Organization for Economic Cooperation and Development (OECD) members as high-income comparators. For each country we computed Pearson and Spearman correlations between RSV and (i) concurrent cases and deaths and (ii) cases and deaths one month later to assess short-term predictability.
Results
Concurrent RSV–case associations were predominantly positive, with a third of countries showing statistically significant correlations; several OECD members exhibited strong relationships (e.g., Canada, France, Belgium, United States). Correlations with deaths were uniformly weak, consistent with sparse, delayed mortality. Predictive performance improved when RSV led cases by one month: nearly a half of countries demonstrated significant RSV and cases associations, with OECD countries showing higher correlations.
Conclusions
Google search interest provides an informative and practical adjunct to conventional Mpox surveillance, offering a one-month early warning in most settings, particularly across OECD countries, while mortality forecasting remains unreliable due to low event counts and longer clinical latency. By leveraging Google Trends as a predictive tool, public-health authorities can anticipate rising transmission and implement preventive measures—such as targeted vaccination—before outbreaks escalate.
Introduction
The emergence and re-emergence of infectious diseases pose substantial threats to global public health, emphasizing the need for accurate, real-time surveillance and early outbreak prediction tools.1,2 Mpox, previously known as Monkeypox, has recently seen unprecedented global spread, prompting heightened attention from health authorities worldwide.3,4 In August 2024, the World Health Organization (WHO) designated Mpox a public health emergency of international concern, underscoring the urgency of strengthened surveillance and proactive response. Traditional surveillance methods, relying primarily on clinical reports and laboratory confirmations, often experience delays, resulting in a critical gap between outbreak onset and public awareness. 5 Therefore, alternative data-driven approaches for timely disease tracking and prediction have gained increasing interest.
Among these approaches, analyzing search engine query data—particularly Google Trends—has demonstrated considerable potential in providing real-time insights into public interest, awareness, and health-related behavior.6,7 Previous studies have successfully utilized Google Trends data to monitor and predict outbreaks of infectious diseases such as influenza, 8 COVID-19,9,10 dengue fever, 11 and Zika virus, 12 highlighting a consistent positive correlation between online search activity and subsequent disease incidence. These findings suggest that search query trends may serve as effective proxies for public concern and symptom occurrence, potentially enabling the prediction of future outbreaks.
Beyond its global spread, Mpox transmission has distinct epidemiologic features that shape surveillance needs. The 2022 multi-country outbreak spread predominantly through close, often intimate skin-to-skin contact within sexual networks. 13 Vaccination policy further underscores the importance of targeted, timely detection. Current recommendations from WHO and Centers for Disease Control and Prevention (CDC) do not support routine mass vaccination of the general population; instead, thirdgeneration Modified Vaccinia Ankara-Bavarian Nordic (MVA-BN) vaccination is prioritized for populations at elevated risk (e.g., specific exposure risks, occupational risks, or sexual risk profiles) and for outbreak-response strategies such as ring vaccination.14,15 This targeted approach increases the value of tools that can anticipate local surges early enough to facilitate focused preventive actions.
Motivated by these observations, our study explores the specific case of Mpox, analyzing the correlation between Google Trends search volume for the keyword “Monkeypox” and the actual reported incidence and resulted deaths of Mpox infections. Our preliminary analysis indicates a strong positive correlation between search activity and outbreak occurrence, implying that heightened public interest correlates with reported infection spikes. Furthermore, we demonstrate the potential utility of leveraging this correlation as a predictive indicator, suggesting that Google Trends data may provide early signals for Mpox outbreaks one month in advance.
This study aims to bridge the gap between digital health informatics and epidemiological surveillance, contributing to the advancement of predictive analytics in infectious disease monitoring and outbreak preparedness. Importantly, we show that Google Trends can function not only as a surveillance adjunct but as a predictive tool that enables earlier, preventive interventions—including targeted vaccination—before outbreaks escalate.
Materials and methods
Data
National surveillance data on laboratory-confirmed mpox cases and deaths were obtained from the WHO public dashboard “Global Mpox Trends”. Monthly records spanning May 1, 2022 through November 30, 2024 were downloaded and included the number of cases and deaths reported by each country. For the same period, country-specific relative search volume (RSV) indices for the keyword “Monkeypox” were retrieved from Google Trends under the “all categories” filter and aggregated at weekly resolution, which were later transformed into monthly resolution to ensure temporal synchrony with the WHO surveillance calendar. Countries without access of RSV values were excluded, resulting in a total of 96 nations, of which 38 are current members of the Organisation for Economic Co-operation and Development (OECD) and serve as a proxy for high-income settings.
Measurement
Statistical analyses were performed in Python using the packages pandas, scipy, and statsmodels libraries. For every country, Pearson’s correlation coefficient and Spearman’s rank correlation coefficient were computed between weekly RSV and monthly incident cases, and between RSV and deaths, both contemporaneously and with the one-month lead. Pearson evaluates the strength of a linear relationship between monthly Google Trends search interest and incident Mpox counts and is most appropriate when variables are approximately normally distributed and outliers are limited. Spearman assesses monotonic relationships by correlating ranks, providing robustness to non-normality, skewed count data, and outliers. Reporting both measures offers complementary perspectives, capturing linear dependence while remaining sensitive to non-linear but consistently increasing or decreasing patterns. Results of both estimators with the statistical significance levels of 0.05 and 0.01 were included, as both statistics have been applied in previous study. 16
Ethics
Because all information analyzed is aggregated and publicly accessible, the study was exempt from institutional review board oversight and conforms to the ethical standards of research that does not involve human subjects.
Results
Summary statistics of correlations between Google Trends relative search volume and Mpox cases in all evaluable countries and in OECD member countries, presented for both concurrent correlations and one-month lead correlations.
*Standard Deviation.
Predictive performance for Mpox cases improved when RSV was advanced by a month. In the one-month lead analysis of 96 evaluable countries, Pearson coefficients increased (median 0.326, IQR 0.000–0.592; mean ± SD 0.312 ± 0.320; range −0.224 to 0.875), and Spearman coefficients similarly strengthened (median 0.466, IQR 0.178–0.597; mean ± SD 0.368 ± 0.291; range −0.521 to 0.887). Under two-tailed tests, 44 (45.84%) countries were significant for Pearson and 55 (57.29%) for Spearman at p < 0.05; critically, 34 (35.42%) countries for Pearson and 51 (53.13%) for Spearman remained significant at p < 0.01, supporting a robust and geographically widespread one-month early-warning signal. Lead correlations with deaths, which were evaluated in 33 countries, were still notably weaker than for cases. In total, 12 (36.36%) countries for Pearson and 11 (33.33%) countries for Spearman were significant at p < 0.05, and 7 (21.21%) countries for Pearson and 7 (21.21%) countries for Spearman were significant at p < 0.01. The countries contributing to the mortality signal were those with larger and more temporally structured death series; examples include the United States, Peru, and Brazil, several of which showed strong Pearson or Spearman coefficients.
OECD member states demonstrated stronger and more consistent relationships than the global average. In the concurrent OECD subset of 38 countries, Pearson coefficients centered higher (median 0.342, IQR 0.226–0.544; mean ± SD 0.345 ± 0.246) and Spearman coefficients were markedly elevated (median 0.543, IQR 0.390–0.616; mean ± SD 0.481 ± 0.219). At p < 0.05, 17 (44.74%) and 32 (84.21%) OECD countries were significant for Pearson and Spearman, respectively. At a stricter p < 0.01, 11 (28.95%) for Pearson and 26 (68.42%) for Spearman in OECD countries remained significant, underscoring the stability of associations in high-income settings with extensive media coverage and high internet penetration. The most prominent concurrent Pearson correlations were observed in Canada, Portugal, France, Switzerland, Italy, Spain, and the United States of America (Figure 1). Countries with weaker or negative concurrent correlations included Japan, Costa Rica, New Zealand, and Slovakia, suggesting smaller outbreaks, differing search behavior, or alternative search-engine use. Monthly Mpox case counts and corresponding average Google Trends RSV are shown for three countries with strong Pearson correlations (United States, Canada, and Switzerland) and three countries with weak Pearson correlations (Japan, New Zealand, and Slovakia).
The one-month lead OECD analysis showed further gains, with Pearson coefficients exhibiting a median of 0.573 (IQR 0.335–0.692; mean ± SD 0.479 ± 0.293) and Spearman a median of 0.545 (IQR 0.451–0.643; mean ± SD 0.505 ± 0.222). At p < 0.05, 27 (71.05%) and 31 (81.58%) OECD countries were significant for Pearson and Spearman; importantly, 24 (63.16%) and 28 (73.68%) remained significant at p < 0.01, indicating a pervasive and statistically robust one-month predictive signal across high-income settings. Leading prediction performers included Belgium, Colombia, Chile, Israel, Switzerland, Mexico, Italy, and Germany. Weaker predictive signals were concentrated in Japan, Costa Rica, Slovakia, and Australia. For deaths, OECD data were available for ten countries. Only the United States of America and Mexico reach statistical significance in both Pearson and Spearman, reinforcing the conclusion that mortality forecasting from search data is generally unreliable in this context.
Discussion
Our findings confirm that population-level information seeking, as captured by Google Trends, offers a meaningful proxy for Mpox transmission dynamics across a wide geographical spectrum. In 96 evaluable countries, concurrent search activity was statistically correlated with incident cases in roughly one-third of settings, and when queries were shifted forward by one month, nearly half of all countries—and nearly three-quarters of OECD members—retained statistically significant associations. These results extend earlier single-country studies that reported temporally aligned spikes in Mpox–related searches during the 2022 outbreak in the United States,17,18 and place them in a genuinely global context spanning both high-income and other regions.
Epidemiologically, the 2022 multicountry wave outside Africa was driven predominantly by clade IIb and spread chiefly through close, often intimate, skin-toskin contact within sexual networks. Clinically, cases frequently presented with fever, lymphadenopathy, and painful mucocutaneous rash, including anogenital lesions, with most illness being self-limited in otherwise healthy adults. 19 In contrast, the more recent surge centered in Central Africa (2024–2025) has involved clade I with broader household and community transmission and substantially higher severity, especially among children, with markedly greater shares of global cases and deaths reported from the Democratic Republic of the Congo. 20
Against this backdrop, the one-month lead we observe for the search signal is operationally important because it maps onto the timetable for vaccine-derived protection. Vaccination with third-generation MVA-BN is targeted to people at increased risk and for outbreak response (e.g., ring vaccination), not for routine mass immunization of the general population. Protection builds over time: the standard two-dose schedule is given 28 days apart, and peak protection is reached about two weeks after the second dose—approximately six weeks from the first injection.21,22 Thus, a 4-week early-warning window provides a practical decision period to initiate the vaccination series, open focused clinics, and direct risk communication so that immunity can mature in step with the anticipated rise in cases. In settings with limited vaccine supply, this timing advantage is crucial for efficient, targeted deployment to the highest-risk networks and frontline workers. 23
Correlations were markedly stronger in OECD members, a finding plausibly explained by (i) near-universal internet access and Google’s dominant market share, (ii) more agile reporting systems that compress the lag between infection and notification, and (iii) extensive, multilingual media landscapes that amplify public awareness. Outliers such as Japan, New Zealand, and Slovakia, where predictive signals were weak or negative, illustrate some caveats, however. Japan’s preference for domestic search engines and New Zealand’s and Slovakia’s low case counts may each attenuate measurable search interest. These country-level idiosyncrasies underscore the need for local calibration when deploying digital indicators.
Finally, consistent with the clinical course and lower case-fatality outside the current African epicenter, mortality proved difficult to predict from search data: deaths are sparse, thus weakening contemporaneous or short-lead associations. Search-based surveillance should therefore be treated as a morbidity-sensitive trigger—best suited to initiating faster, more precise public-health interventions—rather than as a tool for forecasting lethality.
In sum, the practical value of our study is that Google Trends can underpin not only earlier detection, but also faster and more precisely targeted Mpox control actions—exactly what is required when vaccines are prioritized for high-risk groups and must be allocated strategically.
Limitations and future work
Several caveats warrant emphasis. First, ecological correlations cannot establish causality; spikes in search activity may reflect media attention rather than true changes in infection pressure. Second, our analysis relied on a single English keyword. Although this aligns with prior Mpox work, multilingual query panels would capture vernacular terms (for example, viruela s´ımica in Latin America) and likely improve sensitivity. Third, our analyses used monthly data, which can blur shorter lead–lag patterns that might be visible with weekly data. Fourth, the WHO’s renaming of “Monkeypox” to “Mpox” in late 2022 created a shift in public vocabulary. Because our Google Trends queries tracked only “Monkeypox,” later searches for “Mpox” may have been missed, which could weaken correlations toward the end of our study period. Finally, we quantified how closely searches and cases move together, but we did not build a model to predict the exact number of future cases; correlation alone does not tell us about forecast accuracy.
Mpox surveillance should move toward hybrid earlywarning systems that combine digital signals with routine clinical and laboratory data. 24 Within such systems, researchers should track both “monkeypox” and “mpox,” along with local-language terms, to avoid losing signal when vocabulary changes. Expanding beyond Google searches to faster-moving platforms (e.g., TikTok, Reddit) may capture earlier shifts in public behavior. Bringing these data sources together in a single modeling pipeline can improve risk estimates, extend lead time, and better support proactive public-health responses.
Conclusions
This study demonstrates that Google Trends provides a robust, actionable indicator of Mpox activity, offering an early-warning window of approximately one month that aligns with the timetable for vaccination. The lead time identified in our analysis enables health authorities to initiate targeted vaccination and other early public-health interventions before the clinical burden peaks. Because Google search data are freely available, continuously updated, and not constrained by laboratory or reporting delays, they serve as a practical complement to conventional surveillance. Looking ahead, incorporating multilingual search terms and integrating signals from multiple digital platforms will further improve timeliness and targeting, strengthening the role of digital surveillance as an effective public-health intervention to reduce transmission during Mpox outbreaks.
Footnotes
Acknowledgments
National surveillance data on laboratory-confirmed mpox cases and deaths were obtained from the WHO public dashboard “Global Mpox Trends”.
Ethical consideration
Because all information analyzed is aggregated and publicly accessible, the study was exempt from institutional review board oversight.
Author contributions
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.