Introduction

Adherence is defined as the degree to which a patient follows the instructions given by their healthcare professional, including the use of non-pharmacological measures, such as changes in lifestyle habits and preventive behaviors, as well as the use of medications1. Inadequate adherence is intrinsically related to a negative impact on health outcomes, such as increased mortality and readmissions. For Marcum et al. (2013)2, improving adherence to health measures has a greater impact on the general health of the population when compared to the discovery and development of specific pharmacological treatments. van Dooren et al. (2013)3 highlights that personality traits can be decisive in the adherence and prognosis of many diseases.

However, adherence to COVID-19 preventive measures was not homogeneous and varied according to the characteristics of the community and the individual. These consisted of non-pharmacological measures, which were recommended by World Health Organization (WHO) and included social distancing, hand hygiene, use of masks, ventilation of closed environments, closing of schools, suspension of non-essential work activities, and sheltering-in-place4 According to studies prior to COVID-19, for example, the general rate of therapeutic adherence in developed countries was found to not exceed 50%, and in developing countries adherence was even lower5. In countries where the role of local health authorities in recommending restrictions and preventive measures was reduced or absent, such as Sweden6 and Brazil, a phenomenon emerged in which individual characteristics were the most determining factors in adherence to preventive measures of COVID-19. Among the individual characteristics that stand out in the literature, in their relationship with adherence in the COVID-19 pandemic, are age, gender, depressive symptoms, anxious symptoms and personality traits. The relationship between personality traits and psychological responses during the pandemic has been increasingly explored7,8,9 but few studies have focused specifically on adherence behavior.

The term personality represents patterns of behavior and attitudes that are typical of a given individual. Personality traits make this subject different from others, and are relatively constant and stable in each person10. Personality traits can be used to summarize, predict, and explain an individual’s behaviors, which suggests that the explanation for a person’s attitude can be explained by them, and not by the situation11. Different ways of assessing personality exist, with the Big Three12 and Five13 being the ones that have consolidated their recognition, applicability, and external validity14. The Big Three assesses non-pathological, temperamental, and biologically based traits15,16,17,18 which are more suitable for general population-based studies and are less influenced by situational variability. In contrast, the Big Five focuses on pathological traits and was developed primarily for the clinical screening of personality disorders, as defined in the DSM-5 dimensional model19. Despite differences in theoretical and taxonomic basis, both models are correlated20.

The Big Three, also known as Eysenck’s personality inventory, is subdivided into three major traits: Psychoticism (P), Extraversion (E), and Neuroticism (N). Meanwhile, the Big Five is subdivided into five traits: Conscientiousness, Agreeableness, Extraversion, Neuroticism, and Openness. Conscientiousness and Agreeableness, when evaluated together, are inversely equivalent to P20. Openness does not find psychometric equivalence to Eysenck’s model20. In trait E, people are sociable and uninhibited in one of the poles, and shy and withdrawn in the other, also called introversion. In trait N, the neurotic and emotionally unstable personality is opposite to the emotionally stable and secure pole when under pressure; individuals with high N scores are typically anxious, depressed, experience excessive feelings of guilt, and have low self-esteem. The P trait is characterized by impulsivity on the one hand, and impulse control on the other. High P scores are related to hostility, cruelty, lack of empathy, and nonconformity21. In research, the Big Three model is measured using the 90-item Eysenck Personality Questionnaire (EPQ)21 or its abbreviated version, the 24-item EPQR-A22. In contrast, the Big Five is assessed using the 220-item Personality Inventory for DSM-5 (PID-5)19,23 or its 25-item Brief Form (PID-5-BF)24.

However, the articles that focused on the relationships between preventive measures to COVID-19 and personality traits were restricted to cross-sectional studies and did not evaluate personality using the Eysenck model. The associations demonstrated are sometimes contradictory and show limitations in their clinical applicability. Kaspar and Nordmeyer (2022)25, Aschwanden et al. (2021)26, and Carvalho et al. (2020)27 demonstrated that conscientiousness (the opposite to P) was associated with better adherence to care measures, mainly social distancing and hand hygiene. Regarding the traits E and N, the literature shows no consistency in the results of the association between these two traits and preventive measures to COVID-19. Trait E was negatively associated with physical contact25 and general compliance with preventive measures28 but other studies reported a positive relationship between extraversion and the use of face masks and personal hygiene26. As for N, positive correlations were found with social distancing29 and social isolation30 while other studies reported an association between N and less precaution to COVID-19 or even no association with any non-pharmacological measures evaluated25. Openness has also been evaluated25,26,28,30 mostly without significant associations with adherence to COVID-19 preventive measures.

As previously mentioned, the role of other individual characteristics has also been identified in the literature. Several authors reported significant positive correlations between non-pharmacological preventive measures to COVID-19 and older age and female gender31,32,33,34,35. The effect of anxious symptoms is not clear in the literature: Pengpid et al. (2022)36 identified that the presence of anxious symptoms related to COVID-19 increased adherence to social distancing, while patients with diagnosis of Generalized Anxiety were related to lower adherence to social distancing and hand hygiene; Wong et al. (2020)37 pointed out that higher levels of anxiety scores were related to greater adherence to the use of masks, hand hygiene, social distancing and sheltering-in-place. The correlations between adherence to these measures and depressive symptoms were identified as negative38 or even as statistically non-significant36.

There is a variety of literature on the biological and genetic aspects of individualizing medical practices, but little has been studied on the psychological aspects of personalizing medical care. Among individual characteristics, personality stands out as a personal signature that is stable over time. Therefore, understanding personality traits allows healthcare professionals to identify and predict which patients may have difficulty following, and even accepting, a treatment plan, providing early interventions and more effective communication strategies in the doctor-patient relationship. Before focusing on these possible personalized interventions, it is necessary to consolidate the theoretical framework of the relationship between personality traits and therapeutic adherence. Given the gap in the medical literature on the topic, limited only to cross-sectional studies, the longitudinal impact of personality on adherence to non-pharmacological preventive measures will be assessed in the context of the COVID-19 pandemic.

Our conceptual hypothesis, based on a review of the available scientific literature, is that individuals with a high degree of the personality trait P, equivalent to a low degree of Big Five’s Conscientiousness, are related to lower adherence to COVID-19 preventive measures, and the effect converse would also be true. We also hypothesize that individuals with low levels of the trait N and high levels of the trait E may be related to less adherence to these preventive measures, and the opposite effect would also be true.

Objectives

To elucidate the impact of personality traits (P, E, and N) on adherence to preventive measures to COVID-19 (Wearing masks, Social distancing, Sheltering-in-place, Hand hygiene, Working remotely, Desire to protect yourself, Desire to protect others). Secondarily, to evaluate the mediating effect of Lie subscale, depressive and anxious symptoms, the moderating effect of gender, and adjusting effect of age.

Methods

Data source

This is an umbrella project initially developed by Schmitt et al. (2021)39, in which the agreeing subjects were evaluated since May 2020. Recruitment was voluntary, carried out by a convenience sample using the snowball inclusion method. Informed consent (Supplementary Material II) was obtained from all subjects. The research was widely publicized in the University’s virtual environments, as well as in the virtual media most used by the general population, such as Facebook, Instagram, and WhatsApp (Meta Platforms Inc., Menlo Park, CA, USA), in which links were posted to access the online research protocol, developed in Google Forms (Alphabet Inc., Mountain View, CA, USA), SurveyMonkey (SurveyMonkey Inc., San Mateo, CA, USA), and REDCap (Vanderbilt University, Nashville, TN, USA). The follow-up evaluation was conducted with participants who, at the end of the questionnaire, left their contact details (email or telephone) with an interest in continuing to participate in the research. The study had prior approval from the hospital Research Ethics Committee (number: 2020 − 0141) and its methods were performed in accordance with the relevant guidelines and regulations, especially the Declaration of Helsinki.

Design and setting

This is a longitudinal and prospective study. Data collection took place in different contexts of the pandemic. In May 2021, the peak period of the pandemic in Brazil, the first collection was conducted, with 619 subjects included (Supplementary Figure S1). In September 2022, 337 individuals responded to the first follow-up. In the last stage, September 2023, in which the end of the pandemic had already been declared by the WHO, 217 subjects responded to the questionnaires. Each assessment gave rise to a set of data, to be analyzed separately. Between times, the response rate varied from 34.4 to 54.5%.

Instruments

The assessment of personality traits was performed using the Eysenck Personality Questionnaire Revised and Abbreviated (EPQR-A)22. It consists of 24 questions, six for each trait (P, E, and N) and six additional questions for the Lie subscale. Lie subscale assesses the degree of sincerity of individuals when answering this questionnaire itself, in addition to evaluating the tendency of individuals to attribute or reject attitudes and behaviors with values socially desirable or undesirable40. The questions have dichotomous answers, and the sum of the items in each trait defines the score. It presented excellent validity of its psychometric properties in a Brazilian sample in its version translated into Portuguese41. EPQR-A was chosen for its excellent external validity, its self-applicability, and, mainly, its shorter mean time needed to be completed without compromising the quality of its responses.

The measurement of adherence to preventive measures was defined by seven affirmative questions, such as “I maintain a distance of two meters from other people”. The questions could be answered on a 5-point Likert scale, from “Never” to “Always”. Each question evaluated a preventive measure: Working Remotely (to work from home when possible), Sheltering-in-place (to restrict essential tasks outside the home to a minimum), Social Distancing (to stay 1.5 m away from other individuals), Hand Hygiene (to wash with soap or alcohol gel), Wearing Masks (to use face mask outside home that covers your nose and mouth), along with the desires to Protect Others, Protect Yourself.

In addition, depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9)42 in its validated version in Brazilian Portuguese43 and anxious symptoms using the Generalized Anxiety Disorder-2 (GAD-2) scale44 validated in Brazil by Moreno et al. (2016)45. Other sociodemographic data were also collected, including age, gender, and income.

Statistical analysis

The proposed statistic model is composed of preventive measures to COVID-19 as outcome variables and personality traits as predictor variables (Supplementary Figure S2). The variables Lie subscale, depressive and anxious symptoms were included in the model as mediators, gender was included as moderator, and Age as control variable, thus avoiding the loss of sampling power. No outliers were detected. Individuals under 18 years old, those with significant incomplete responses (missing values above 10%) or who did not accept the informed consent form were excluded.

Although personality traits are conceptually latent variables, analyzing the sum of six items for each trait is not statistically appropriate. Structural Equation Modeling (SEM) was chosen for its ability to model latent psychological traits and multiple interrelated outcomes while accounting for measurement error. While covariance-based SEM (CB-SEM) is typically used for theory testing and model fit evaluation, our study had a predictive and exploratory focus. Therefore, we used Partial Least Squares SEM (PLS-SEM), which is more suitable for complex models, smaller samples, and longitudinal data. PLS-SEM also allows trait scores to be estimated from item-level data without relying on simple summations46,47,48 and provides path coefficients (β) as interpretable results, considered significant when p < 0.05. In addition, it accommodates sample size variation across time points and is robust in comparing group-level patterns across assessments46. The software used was SmartPLS 4.0 (SmartPLS GmbH, Boenningstedt, Germany).

The sampling power calculated was 0.95, when the recommended minimum is 0.80. It was calculated with multiple linear regression (fixed model, R2 deviation from zero), from the F test family, via post hoc analysis49,50,51. Significant differences in the means of variables between Times were checked using the Student’s t-test, Analysis of Variance (ANOVA), and Chi-square test, in the software SPSS version 29.0 (IBM Corporation, Armonk, NY, USA). Other information about the suitability for the statistical model used (e.g. convergent validity, discriminant validity, adjustment for each time), which presented good levels of statistical quality in our model, is provided elsewhere (Supplementary Tables S1, S2).

Results

Sample characteristics

The sociodemographic characteristics of the samples from each time are shown in Table 1. There was a predominance of females (82–86%), white (91–93%), with a mean age of 46–51 years (SD 13 years), married or with a steady partner (60–67%), with a postgraduate degree in progress or completed (59–67%). There was stability over time for the variables gender (p = 0.392), ethnicity (p = 0.848), marital status (p = 0.508) and education (p = 0.054); meanwhile, age (p < 0.001) showed differences between Times. Anxious (M = 3.54 − 4.28, SD = 1.675 − 1.846) and depressive (M = 14.33 − 15.46, SD = 5.359 − 6.421) symptoms were present at all Times analyzed, with there being stability of depression over time (p = 0.100) and a statistically significant reduction in anxiety over time (p = 0.002).

Table 1 Sociodemographic characteristics of an online sample, in a two-year follow-up. Note. April (Apr); September (Sep); Standard Deviation (SD); Mean (M); p < 0.05 in bold.
Full size table

The samples presented similar levels of personality traits (Supplementary Table S3). Trait P showed a non-significant difference between Times, from 0.081 to − 0.0121 points (p > 0.05). Trait E showed a non-significant difference between Times, from 0.05 to 0.148 points (p > 0.05). Trait N was the only exception, as, between Times 1 and 2, there was a small reduction of 0.36 points (p < 0.05) on the Eysenck scale. This reduction has irrelevant clinical significance and can be explained by the negative variation of depressive symptoms between Time 1 and 2.

Impact of personality traits on adherence

Trait P had the greatest impact on adherence to preventive measures (Table 2) predicting lower adherence to all measures evaluated, except Working remotely and Wearing Masks. At Time 1, there was small negative effects of P on preventive measures Desire to protect yourself (β= −0.127, CI95%= −0.225–(–0.034), p = 0.01), Desire to protect others (β= −0.132, CI95%= −0.261–(–0.011), p = 0.032). P showed small negative effects at Time 2 on Social Distancing (β= −0.149, CI95%= −0.256–(–0.045), p = 0.009), Hand Hygiene (β= −0.167, CI95%= − 0.288–(–0.061), p = 0.005), Sheltering-in-place (β= −0.139, CI95%= − 0.246–(–0.034), p = 0.011), and Desire to protect yourself (β= −0.132, CI95%= − 0.261–(–0.011), p = 0.012). At Time 3, effects of P on these measures ceased to be significant.

Table 2 Effects of personality traits and adherence to COVID-19 preventive measures, in a two-year follow-up. Note. *Adherence to wearing masks in Time 1 was 99,9%, therefore was not included in the model; Confidence Interval (CI); p < 0.05 in bold.
Full size table

As for the other traits, E showed a small positive effect on Desire to protect others (β = 0.089, CI95%= 0.009–0.167, p = 0.026) at Time 1, and Desire to protect yourself (β = 0.150, CI95%= − 0.004–0.287, p = 0.05) at Time 3. There were no significant effects of E on preventive measures Social distancing, Hand hygiene, Sheltering-in-place, Wearing masks and Remote work, at any of the times evaluated. Trait N had no significant effect on any of the preventive measures evaluated and at any of the times evaluated.

Mediation and moderation analysis

There were moderating effects of gender in the impact of personality traits and COVID-19 preventive measures (Table 3). P impacted Social distancing among both genders, with males exhibiting a worse adherence (β= −0.296, CI95%= −0.545−(− 0.037), p = 0.022) at Time 2 compared to females (β= −0.115, CI95%= −0.222 − 0.001, p = 0.046). Also at Time 2, the effect of P on Wearing masks was more evident among males (β= −0.302, CI95%= −0.601−(− 0.015), p = 0.040). Among females, the effect of P was more present on Hand hygiene (β= −0.130, CI95%= −0.236−(− 0.021), p = 0.019), Sheltering-in-place (β= −0.095, CI95%= −0.185−(− 0.002), p = 0.047), Desire to protect yourself (β= −0.150−(− 0.126), CI95%= −0.258−(− 0.003), p = 0.004 − 0.050), and Desire to protect others (β= −0.157, CI95%= −0.265−(− 0.053), p = 0.004).

Table 3 Moderation effects of gender between personality traits and COVID-19 preventive measures, in a two-year follow-up. Note. *Adherence to Wearing masks in Time 1 was 99,9%, therefore was not included in the model; Preventive measures not presented were not affected by the moderating effect of Gender; Confidence Interval (CI); p < 0.05 in bold.
Full size table

Additionally, the positive impacts of E on the preventive measures Desire to protect yourself at Time 3 (β = 0.152, CI95%= 0.008 − 0.308, p = 0.049) and Desire to protect others at Times 1 and 3 (β = 0.094 − 0.177, CI95%= 0.003 − 0.330, p = 0.020 − 0.038) were more prominent among women, while the Hand hygiene measure was more prominent among men at Time 1 (β = 0.312, CI95%= 0.126 − 0.501, p = 0.001). Meanwhile, among men, a negative effect of N on adherence to Working remotely at time 2 became significant (β= −0.409, CI95%= −0.837−(− 0.059), p = 0.038).

It was not identified any mediating effects of depressive or anxious symptoms in the significant relationships between personality traits and COVID-19 preventive measures (Table 4). Nonetheless, regarding the direct effects of depressive symptoms on preventive measures, there were a small and positive significant correlation with Social Distancing at Time 2 (β = 0.185, CI95%= −0.208 − 0.066, p = 0.041), a small-moderate and positive correlations with Sheltering-in-place at Time 2 (β = 0.157, CI95%= −0.047 − 0.236, p = 0.006) and 3 (β = 0.304, CI95%= 0.093 − 0.488, p = 0.04), a moderate and positive correlation with Wearing masks at Time 2 (β = 0.244, CI95%= 0.033 − 0.438, p = 0.012) and a moderate and positive correlation with Desire to protect yourself at Time 3 (β = 0.302, CI95%= −0.048 − 0.251, p = 0.048). Regarding the direct effects of anxious symptoms on preventive measures, there was small and positive correlation of anxious symptoms with Hand Hygiene at Time 1 (β = 0.142, CI95%= 0.007 − 0.255, p = 0.031).

There were no mediating effects of the Lie subscale on any of the preventive measures (Table 4). However, there was a direct effect of Lie on the preventive measures Hand hygiene (β= −0.152, CI95%= −0.837−(− 0.059), p < 0.001) and Desire to protect yourself (β= −0.106, CI95%= −0.059 − 0.106, p = 0.013) only at Time 1. Regarding the effects of the control variable age (Table 4), there were positive correlations of age in the three times evaluated with the preventive measures Social distancing (β = 0.234 − 0.307, CI95%= 0.113 − 0.406, p < 0.001) and Sheltering-in-place (β = 0.188 − 0.325, CI95%= 0.089 − 0.435, p = 0.000 − 0.002), at Times 1 and 2 with the Desire to protect yourself measure (β = 0.179 − 0.182, CI95%= −0.154 − 0.255), p = 0.000 − 0.004), and at Times 2 and 3 with Wearing masks (β = 0.216 − 0.243, CI95%= 0.068 − 0.365, p = 0.000 − 0.006).

Table 4 Mediation and control effects of Depression, Anxiety, Lie subscale, Age with adherence to COVID-19 preventive measures, in a two-year follow-up. Note. *Adherence to Wearing masks in Time 1 was 99,9%, therefore was not included in the model; There was not any significant mediating effect of Depression, Anxiety, Age, Liar in the relationship between personality traits and COVID-19 preventive measures; Confidence Interval (CI); p < 0.05 in bold.
Full size table

Discussion

This study, to the best of our knowledge, is the first to longitudinal and prospectively evaluate the impact of personality traits on adherence to COVID-19 preventive measures. Regarding the primary objective, trait P showed a negative impact on Social Distancing, Hand Hygiene, Sheltering-in-place, Desire to protect yourself and Desire to protect others in Time 1, and Desire to protect others in Time 2, corroborating results found by Kaspar and Nordmeyer (2022)25, Aschwanden et al. (2021)26, and Carvalho et al. (2020)27. Trait E, which is not unanimous in the literature, with some authors relating it to better adherence26 and others to worse adherence25,28 presented only in Time 1 a significant and positive impact on adherence to measures Desire to protect oneself, and Desire to protect others. Trait N had no impact on adherence to COVID-19 preventive measures, a similar result to those demonstrated by Kaspar and Nordmeyer (2022)25 and in disagreement with the findings of Abdelrahman (2022)29, Götz et al. (2021)30, who pointed out positive correlations between N and Social distancing, Sheltering-in-place. Thus, the adherence impact at Time 3 (September 2023) was minimal. This finding may be explained by the statistical limitations associated with little behavioral variance to detect, known as floor effect52,53 due to the gradual resolution of the COVID-19 pandemic and the population-wide reduction in adherence to preventive measures.

Moderation by gender was shown to be statistically significant between personality traits and COVID-19 preventive measures. Among men, there was poorer adherence to Wearing masks, Hand hygiene and Social distancing. However, among women, there was worse adherence to Social distancing, Hand hygiene, Sheltering-in-place, Desire to protect yourself and Desire to protect others. Different from what was expected and reported in the literature31,32,33,34,35 better adherence to preventive measures among females was not uniform and depended on the specific measure evaluated.

Mediation by anxious symptoms and depressive symptoms also lacked statistical significance, the latter being in line with Pengbid et al. (2022)36 and contrary to Wong et al. (2020)37. Anxiety and depression variables were specially included in the statistical model to reduce a potential confounding effect on personality results, considering the difference between a trait, which is stable and independent of the context, and a state, which is variable and present in a specific context. Mediation by the Lie subscale was not statistically significant, reducing social desirability bias. The control variable age positively affected Social distancing, Sheltering-in-place, Wearing masks, and Desire to protect yourself. Given the participants’ mean age, the conservative bias in our results on the effects of personality traits on preventive measures must be considered. In addition, the statistically significant difference in age between Times in the descriptive analysis was overcome with the inclusion of age as a control variable in the statistical model.

Among the strengths of this study is obtaining potentially more accurate results compared to existing literature, as (i) this was the first longitudinal study to evaluate the relationship between personality traits and COVID-19 preventive measures; (ii) the use of Structural Equation Modeling with Partial Least Squares (PLS-SEM) provided a robust framework for analyzing latent variables, covariables, and addressing sample size reduction and temporal differences; (iii) loss of follow-up did not affect the internal validity, as the main variable, personality traits, remained stable across Times 1, 2 and 3, and the model’s sampling power was calculated at 0.95, exceeding the recommended minimum of 0.80; (iv) the sample consisted of Brazilians in a unique context where individual characteristics predominantly influenced adherence to preventive measures due to the lack of government authority and collective health recommendations. While this setting was ideal for testing our hypotheses, it may limit the generalizability of the findings to countries with stronger public health enforcement.

There were some factors that can be identified as limiting our results: (i) the research protocol was defined in the first year of the pandemic, necessitating remote data collection, which might have introduced digital access and response patterns biases; (ii) the measurement of adherence to preventive measures was carried out using a self-report questionnaire without prior validation and with little detail, a limitation that occurred in all studies reviewed on the topic, given the unprecedented nature of the pandemic, which highlights the need for future psychometric validation of these measures to improve reliability and accuracy; (iii) given the logistical constraints, such as shelter-in-place orders, governmental restrictions on movement54 a non-probabilistic convenience sampling was carried out, which allowed rapid recruitment of a large number of participants but introduced potential selection bias, thereby limiting the generalizability of the findings to the broader population; (iv) although social desirability (Lie subscale) was controlled for in the statistical analysis, the high educational level of the sample may have introduced an education-related response bias, as greater familiarity with psychological constructs could influence the way participants interpret and self-report their answers; (v) the sample’s gender imbalance, with a predominance of females, may have biased the moderation analysis by reducing the statistical power for the male subgroup, thus limiting the generalizability of the findings; (vi) the cultural and political context of Brazil during the pandemic, characterized by reduced governmental enforcement of preventive measures, may limit the applicability of our findings to other sociocultural contexts.

Therefore, personality traits, particularly P, may play a more determining role in adherence behavior, especially in the absence of strong external constraints. This aligns with Eysenck’s theoretical model, which conceptualizes personality as a product of both temperament and environmental reactivity. Beyond the pandemic context, in situations with weaker external regulation, such as self-managed chronic illness (e.g. hypertension, HIV infection, bipolar disorder) or long-term lifestyle changes (e.g. diet, physical exercise), individuals’ traits may amplify or buffer responses to situational demands. For example, individuals with low P are more likely to maintain treatment routines when external support is limited, whereas those high P may struggle with adherence and would require more structured guidance and reinforcement. These dynamics underscore the need to align interventions with both personality traits and context.

The identification of psychological risk factors for therapeutic non-adherence can result in the development of screening methods, risk indicators, and methods of psychoeducation and specific support. This is a potential strategy to increase effectiveness, efficacy, and health equity55. Ways to improve therapeutic adherence were reviewed in a meta-analysis carried out by Conn & Ruppar (2017)56 in which the best interventions found were those focused on behavior (e.g. goal setting, rewards) and habits (e.g. attachment of adherence to pre-existing habits), carried out face-to-face between patient and healthcare professional. Interventions focused on cognition, such as proposals to change knowledge, beliefs and ways of thinking, did not demonstrate relevant clinical effects. Still, there is a lack of studies on the topic, due to the difficulty of including and following up poorly compliant patients57 and the present study addresses this gap by illustrating how personality traits can inform intervention strategies across various health domains. Specifically, recognizing these individual differences allows for the development of personalized public health messages and clinical interventions, from vaccination campaigns to chronic disease management, that target the psychological drivers of behavior. Such tailored approaches hold the potential to improve therapeutic outcomes, especially for clinically resistant or non-adherent cases, thereby enhancing the overall efficacy of public health initiatives1,58,59,60.

Conclusion

This two-year online follow-up study has appointed the impact of personality traits, specifically those outlined by Eysenck (1976)12(Psychoticism, Extraversion, Neuroticism), on the adherence to COVID-19 non-pharmacological preventive measures. Our findings demonstrated that P significantly impacted adherence to almost all preventive measures evaluated, and this adherence varied over time. Trait E had some impacts on adherence to specific preventive measures, such as Hand hygiene, and N showed no significant effects on adherence. Gender had a significant moderating effect, while mediating effects of Lie subscale, depressive symptoms and anxious symptoms were not substantiated in our analysis.