Smell and taste changes are early indicators of the COVID-19 pandemic and political decision effectiveness

The study investigates whether self-reported changes in smell and taste can serve as early, specific indicators of SARS-CoV-2 spread and impending strain on healthcare systems, compared to existing governmental indicators based on emergency room consultations. In the context of national lockdowns and phased reopening strategies (e.g., France’s red/green regional labeling), the authors aim to (1) test spatial associations between self-reported chemosensory changes and hospital stress (COVID-19 hospitalizations, critical care admissions, mortality) across French regions, and (2) examine temporal relationships between peaks in chemosensory symptom onset, governmental ER consultation indicators, critical care admissions, and lockdown timing. A secondary aim evaluates how these dynamics vary across countries (France, Italy, UK) with different lockdown stringency levels, to assess the utility of chemosensory changes as early markers of infection spread and policy effectiveness.

Multiple studies across Europe and beyond report high prevalence of olfactory and gustatory dysfunctions in COVID-19, often occurring early and with higher specificity than fever or cough (e.g., Iran, Spain, France, Italy, Germany, UK; refs 2–11). Prior work suggests sudden smell/taste loss can predict COVID-19 cases and clinical course, and that chemosensory symptoms may precede other manifestations. Seasonal patterns of olfactory disturbances have been described in other countries with modest spring increases, but these amplitudes are limited and differ from the pronounced peak observed during COVID-19 surges. Additional corroboration from Google search behavior aligns with symptom reporting during the pandemic period rather than prior years.

Design and data sources: Primary data came from the Global Consortium for Chemosensory Research (GCCR) crowd-sourced online survey (30+ languages), collected April 7–May 14, 2020, approved by Penn State IRB (STUDY00014904). Recruitment occurred via word of mouth and media (social, TV, radio, print). Inclusion criteria: adults (>18) reporting a respiratory disease in the prior two weeks; for analyses here, only participants reporting smell and/or taste change with a quantitative decrease ≥5 (on 0–100 scale) from pre-illness to illness were included. Country of residence was parsed (allowing for variants/typos) to identify respondents in France, Italy, and the UK. For France, geographic mapping used the pre-2016 22-region division and 96 departments via reported city and public datasets (data.gouv.fr), normalizing counts by regional population (INSEE). Sample sizes varied by analysis due to missingness: France n≈1476–4720; Italy n≈264–1241; UK n≈243–750. Demographics were predominantly female (FR 66.38%, IT 69.3%, UK 76.0%); mean ages ~41 years. In the French subset, 15% lab-confirmed COVID-19 positive, 44% clinically diagnosed, 41% undiagnosed but with chemosensory change. Complementary French survey: An independent French online survey (April 8–May 8, 2020; 950 eligible respondents) assessed chemosensory disorders with or without COVID-19 and quality-of-life impacts. Inclusion: first-time, complete responses, adults >18, reporting completion and onset dates. Approved by CNRS ethics. Public health and contextual data: French healthcare stress indicators (hospitalizations, Critical Care Resuscitation Unit [CCRU] admissions, deaths) from Santé Publique France (Géodes). Governmental virus circulation indicator: ratio of ER consultations for suspected COVID-19 to general ER consultations (ICD-10 codes provided). French red/green labels, allergy ER ratios for prior years, and Google Trends queries by region (Feb 1–May 10, 2020) were analyzed (normalized relative scores). Cross-country daily cases/deaths from ECDC (downloaded May 22, 2020). Government response stringency index from Oxford COVID-19 tracker. Temporal analyses: Symptom chronology used survey-reported dates of first symptoms and chemosensory-change onset. Data visualizations included 7-day running averages and normalization to peak. Statistics: Pre-registered analyses (OSF) included Pearson correlations (e.g., between regional chemosensory reports and CCRU admissions, hospitalizations, deaths), Chi-square tests and Biserial correlations for red/green labels, and correlations with Google Trends and the governmental indicator. Robustness checks included department-level correlations, exclusion of extreme regions (Alsace, Île-de-France), sub-analyses limited to non-clinically diagnosed individuals, and bias checks versus regional media coverage. Additional pre-registered tests (Mann-Kendall trend, change-point detection) and parts of Google Trends analyses were noted but not presented.

• Spatial associations in France: Strong correlation between self-reported smell/taste changes and CCRU admissions across regions (R=0.88, p=8.9×10^-08), remaining significant after excluding Alsace and Île-de-France (R=0.72, p<3×10^-04). At department level, correlations remained high (e.g., CCRU admissions: R=0.76, p<5×10^-19). Associations held when restricting to individuals not clinically diagnosed (e.g., CCRU admissions: R=0.83, p=1.65×10^-06). No correlation between regional media coverage and number of responses (R<0.01, p>0.9) mitigates sampling bias concerns.
• Comparison to governmental indicator: Smell/taste changes correlated more strongly with hospital admissions and deaths than the governmental ER consult ratio (admissions: R_smell=0.81, p=6.71×10^-06 vs R_gov=0.60, p=3.35×10^-03; deaths: R_smell=0.75, p=5.62×10^-05 vs R_gov=0.58, p=4.97×10^-03).
• Temporal dynamics in France: Peak onset of smell/taste changes occurred 4 days after lockdown, with first symptoms reported slightly earlier. The governmental ER consult indicator peaked 11 days after lockdown; CCRU admissions peaked 14 days after. Peaks were robust across survey completion dates, replicated in an independent French survey (n=950), and consistent across age and gender. Allergy-season confounds were unlikely (allergy ER ratios peak around week 30, outside study window; low pollen risk during first lockdown week).
• Google Trends: Online searches for early symptoms (fever, cough, aches) peaked around March 18, coinciding with reported first symptoms; searches for “taste loss” and “smell loss” peaked a few days later, aligning with chemosensory-change onset; searches for shortness of breath preceded CCRU admission peaks. Google queries related to smell/taste loss correlated with healthcare stress indicators (e.g., CCRU admissions R≈0.8, p<4×10^-03).
• Cross-country comparisons (France, Italy, UK): Countries with higher, earlier lockdown stringency (France, Italy) exhibited rapid decreases in self-reported chemosensory changes and overall symptoms, followed by declines in confirmed cases (with ~2-week lag). The UK, with less stringent/ later lockdown, showed slower declines and a plateau in confirmed cases. Declines in new-onset chemosensory changes anticipated subsequent trends in confirmed cases.

The findings demonstrate that self-reported smell and taste changes are tightly linked to healthcare system stress (hospitalizations, CCRU admissions, deaths) and precede traditional governmental indicators, making them valuable early markers of SARS-CoV-2 spread. Temporal patterns suggest many individuals were infected just before lockdown, with chemosensory changes arising early in the disease course. Multiple robustness checks (replication in an independent survey, consistency across age/gender, lack of correlation with media coverage, and supportive Google Trends data) argue against major recall or media bias. Seasonal allergy confounds are unlikely given non-overlapping seasonal peaks and pollen risk levels. Cross-country analyses show that rapid reductions in chemosensory-change incidence track the effectiveness of stringent lockdown policies, whereas slower decreases align with less stringent measures and sustained case counts. Thus, monitoring population-level chemosensory changes can inform public health surveillance and policy decisions, particularly where testing is limited.

Self-reported sudden changes in smell and taste provide a minimally invasive, cost-effective, and early indicator of COVID-19 spread and impending hospital burden. They outperform existing governmental ER-based indicators in predicting healthcare stress and reflect the effectiveness of public health measures across countries with different lockdown stringency. The authors recommend large-scale, prospective, real-time monitoring of chemosensory symptoms, integrated with standardized data collection to control for selection biases, and development of statistical models to predict future hospital admissions and guide local outbreak detection and reopening strategies.

• Self-reported, retrospective symptom timing introduces potential recall bias, although analyses (stable onset distributions across completion dates, minimal explicit references to lockdown in narratives, concordance with real-time Google Trends) mitigate this concern.
• Selection biases inherent to online surveys (e.g., over-representation of women, under-representation of seniors, technology access, motivation) may affect generalizability; however, age/gender stratified analyses showed similar temporal peaks, and results replicated in an independent survey with different recruitment/format.
• Media coverage might influence participation and symptom salience, though regional media exposure did not correlate with response counts; implicit biases cannot be fully excluded.
• Correlational design precludes causal inference; real-time prospective monitoring is needed.
• Governmental indicators and case definitions evolved during the study period, potentially affecting comparisons.
• Google Trends provides relative, not absolute, search volumes; interpretations should be cautious.