The impact of government actions and risk perception on the promotion of self-protective behaviors during the COVID-19 pandemic

ENGLISH VERSION

Full Reference: Alvarez-Galvez, J., Anastasiou, A., Lamnisos, D., Constantinou, M., Nicolaou, C., Papacostas, S., Vasiliou, V. S., McHugh, L., Lubenko, J., Ruiz, F. J., Paez-Blarrina, M., Montesinos, F., Valdivia-Salas, S., Merwin, R. M., Karekla, M., Gloster, A. T., & Kassianos, A. P. (2023). The impact of government actions and risk perception on the promotion of self-protective behaviors during the COVID-19 pandemic. PLoS ONE, 18(4), e0284433. https://doi.org/10.1371/journal.pone.0284433

Study Type: Secondary data analysis using quasi-experimental interrupted time series analysis

Background and theoretical framework

This research is grounded in Protection Motivation Theory (PMT), which posits that fear-inducing health messages trigger cognitive appraisals of threat severity and susceptibility. The theory proposes that individuals' coping responses—whether adaptive (protective behaviors) or maladaptive (avoidance or denial)—depend on the balance between perceived threat and perceived efficacy of recommended actions. During the COVID-19 pandemic, the researchers hypothesized that both government stringency measures and population-level risk perception would influence adherence to self-protective behaviors, though potentially in different ways across geopolitical contexts.

The study tested two primary hypotheses: (1) government responses aimed at controlling pandemic spread can reduce risk perception and consequently modify self-protective behaviors, and (2) during health emergencies, threat perception and resulting fear responses may function as a key motivator for maintaining self-protective adherence. Notably, this is positioned as the first comparative study examining how government actions and risk perception jointly influence protective behaviors across multiple countries with heterogeneous policy responses and pandemic trajectories.

Method

Study Design and Population: The research employed secondary data analysis from the COVID-19 Impact project survey, a population-based cross-sectional study conducted April 7 to June 7, 2020. From an original sample spanning 78 countries, nine countries were selected based on sufficient temporal data points (minimum 30 days of continuous measurement) and adequate sample sizes. Final analyzed populations included: Cyprus (N=957), Germany (N=279), Greece (N=270), Ireland (N=414), Latvia (N=1,285), Spain (N=296), Switzerland (N=550), United Kingdom (N=100), and United States (N=268), totaling 4,419 adult participants (age ≥18 years).

Measurement of Self-Protective Behaviors: Three key behavioral measures were assessed on 0-10 Likert scales: hand washing frequency, social isolation adherence, and social distancing compliance. A composite outcome variable was constructed by averaging these three indicators, yielding a scale from 0 (minimal adherence) to 10 (maximum adherence).

Measurement of Government Response: The Oxford COVID-19 Government Response Tracker (OxCGRT) stringency index was employed, a validated composite measure comprising nine policy indicators (lockdown policies, school and workplace closures, travel restrictions, public event cancellations, public transport restrictions, stay-at-home requirements, restrictions on internal movement, and international travel controls) scored on a 0-100 scale, with 100 representing maximum stringency.

Measurement of Risk Perception: Daily confirmed COVID-19 case counts by country served as a proxy indicator for population-level risk perception and threat salience, reflecting the pandemic's epidemiological trajectory within each jurisdiction.

Statistical Analyses: The primary analytical approach employed Interrupted Time Series Analysis (ITSA), an extension of the difference-in-differences (DiD) design. The ITSA model specified was: Y_t = β₀ + β₁T_t + β₂X_t + β₃X_tT_t + ε_t, where Y represents the behavioral outcome, T denotes time in days from study start, X indicates the timing of an intervention or policy change, and the interaction term (XT) captures slope changes following the interruption. This design enables decomposition of level changes (immediate effect) and slope changes (sustained effect) at specific temporal breakpoints.

Complementing ITSA, Change-Point Detection Analysis (CPDA) using the Isolate-Detect methodology (via R packages IDetect and breakfast) was employed to objectively identify the temporal location of significant behavioral shifts without imposing predetermined intervention dates. One-way ANOVA tested differences in baseline adherence across countries. Ethics approval was obtained from the Cyprus National Bioethics Committee (reference: EEBK ΕΠ 2020.01.60).

Results

Across all nine countries, baseline self-protective behavior adherence was notably high, generally exceeding 8.5 on the 0-10 scale. However, significant between-country variation was documented through one-way ANOVA comparisons (social isolation: F=19.80, p<0.001; social distancing: F=16.81, p<0.001; hand washing: F=12.87, p<0.001; composite self-protective behaviors: F=14.47, p<0.001). This heterogeneity suggested that country-level factors—including cultural context, media representations, trust in institutions, and prior pandemic experience—moderated behavioral responses.

---

Primary Analytical Results

Government Stringency and Behavioral Change: Interrupted time series analysis revealed complex, country-specific relationships between government policy relaxation and self-protective behavior modification. In Cyprus (change point 5/3/20, β=-0.099, p<0.001), Greece (5/4/20, β=-0.158, p=0.029), and the United Kingdom (5/11/20, β=-0.127, p=0.001), relaxation of government stringency measures was followed by statistically significant decreases in self-protective behaviors. This pattern aligns with theoretical expectations that reduced official stress signals reduce population-level threat perception.

Conversely, other countries exhibited resilience in behavioral maintenance despite policy relaxation. Germany (5/2/20) and Ireland (5/17/20) showed no significant behavioral decline associated with policy changes (both p>0.668). Most notably, Spain (5/18/20, β=0.176, p=0.049) and Latvia (5/11/20, β=0.088, p=0.046) demonstrated increased protective behaviors precisely when government measures were relaxed—suggesting that population-level risk perception maintained behavioral adherence independently of policy signals. Switzerland exhibited a more complex temporal pattern, with an initial policy-behavior disconnection (4/26/20, β=0.056, p=0.093) followed by stabilization (5/10/20, p=0.734). The United States (4/29/20) showed no significant relationship (p=0.984).

Daily COVID-19 Cases and Behavioral Responses: When employing daily case counts as a proxy for risk perception, distinct patterns emerged across jurisdictions. Cyprus (4/23/20, β=-0.056, p=0.001) and Greece (4/18/20, β=-0.063, p=0.009) demonstrated significant behavioral relaxation when daily cases declined, suggesting population-level risk perception tracked epidemiological trends. Ireland showed marginal significance (4/16/20, β=-0.012, p=0.057). Germany (5/3/20), Spain (5/4/20), and the United Kingdom (5/3/20) exhibited no significant case-behavior relationships.

Uniquely, Switzerland demonstrated increased protective behaviors when daily case counts decreased (4/12/20, β=0.014, p=0.050), suggesting either persistent risk perception despite declining cases or adherence to alternative protective motivation sources (such as personal vulnerability perception or institutional trust). Latvia and the United States showed no statistically significant temporal variation attributable to case count fluctuations.

---

Interpretation and Theoretical Implications

These findings provide qualified support for both primary hypotheses, though with substantial country-level heterogeneity requiring interpretive nuance. The Protection Motivation Theory framework helps explain both the strong baseline adherence and the differential responses to policy and epidemiological signals. In Cyprus, Greece, and the United Kingdom, government policy changes served as a dominant appraisal cue, effectively signaling reduced threat and thus permitting behavioral adaptation. In Spain, Latvia, and Switzerland, population-level risk perception—whether derived from personal vulnerability assessments, trust in scientific evidence, or international media exposure—appeared to override governmental policy signals, maintaining protective motivation despite official "all-clear" signals.

The authors interpret the Spain and Latvia patterns as evidence that "perceived sense of risk at population level may have greater impact on collective behaviors than government-directed changes." This suggests that during early pandemic phases, when scientific uncertainty remained high and international case trajectories continued accelerating, populations maintained autonomous threat appraisals. Pandemic fatigue, though mentioned as a potential mechanism, likely cannot fully explain the heterogeneous patterns, as fatigue would predict uniform behavioral decline across countries.

Discussion and conclusions

Methodological Contributions

The integration of Change-Point Detection Analysis with Interrupted Time Series Analysis represents a methodological innovation, permitting objective identification of temporal breakpoints without a priori assumptions about policy timing. This approach addresses limitations of conventional ITSA that require predetermined intervention dates, thereby reducing specification error and increasing analytical transparency.

---

Limitations on Inference

The study's temporal scope—restricted to the early pandemic period (April-June 2020)—precludes generalization to subsequent waves, variant emergence, or later pandemic phases characterized by vaccination campaigns or evolved public perception. The reduction from 78 to 9 countries reflects substantial attrition due to insufficient longitudinal data points, potentially introducing selection bias favoring countries with systematic surveillance systems and administrative capacity. The absence of direct individual-level risk perception measures, relying instead on aggregate daily case counts, limits causal inference regarding psychological mechanisms. Finally, the aggregate-level analytical design precludes individual-level inferences, preventing examination of demographic or psychological heterogeneity in responses.

---

Significance and contribution

This study demonstrates that government public health policy and population-level epidemiological threat perception operate as distinct, sometimes competing influences on self-protective behavior maintenance during pandemic emergencies. Effective pandemic communication and policy implementation strategies must account for country-specific institutional trust, media landscapes, and population risk perception trajectories. Similar government interventions may produce divergent behavioral outcomes depending on contextual factors, supporting calls for tailored, context-sensitive public health communication strategies.

The research underscores that top-down policy mandates function optimally when aligned with bottom-up risk perception. Disconnection between official policy signals and population-perceived threat can result in either excessive behavioral maintenance (potentially increasing pandemic fatigue) or insufficient adherence to genuinely beneficial protective actions. Future pandemic preparedness initiatives should integrate real-time population risk perception assessment with policy modification to optimize alignment between official guidance and collective threat appraisals.

---

---

---

VERSIÓN EN ESPAÑOL