Effects of Cognition and Affect on Preventive Behaviors During the COVID-19 Pandemic: A Cross-Sectional Study in China

Background: The global outbreak of COVID-19 has become an international public health crisis. Specic antiviral treatments for COVID-19 are not yet available, and prevention is of particular importance to ghting the virus. This study tends to explore and compare the roles of cognitive and affective factors in predicting preventive behavior adoption during the COVID-19 pandemic in China. Methods: An online survey using a quota sampling method to collect responses from 3,000 Chinese adults was conducted from March 2, 2020 to March 23, 2020. Questions included sociodemographic features, coronavirus knowledge, negative emotion, risk perception, and behavioral responses. Multiple regression analyses were conducted to examine the predictors of behavioral responses toward COVID-19. Results: On average, respondents had low levels of knowledge about COVID-19 (the overall correct response rate was 7.5 percent). Most respondents reported moderate to strong negative emotions towards the virus (3.47 out of 5). The average reported perceived chance of infection was 23.89%. For behavioral responses, respondents reported low frequencies of going out for activities (1.98 out of 4) and high frequencies of taking preventive measures (3.22 out of 4). Behavioral responses toward COVID-19 were found to be determined by cognitive and affective variables. Knowledge was negatively related to frequency of going out for activities (β=-0.11, p<.001). Negative emotion (β=0.34, p<.001), and risk perception (β=0.05, p=.007) were positively associated with going out for activities. The explanatory power of affective variables (ΔR 2 =12.1%) was greater than cognitive variables (ΔR 2 =1.0%). For preventive behaviors, knowledge was positively associated with preventive behaviors (β=0.22, p<.001). Negative emotion (β=-0.28, p<.001) and risk perception (β=-0.05, p=.002) were all negatively associated with preventive measures. Affective variables still showed stronger explanatory power (ΔR 2 =8%) than cognitive variables (ΔR 2 =4.4%) in predicting preventive behaviors. Conclusions: After the

country as of April 25, 2020. Using WHO (World Health Organization) data (up to March 1, 2020), a research article from the Lancet estimated that the mortality rate of COVID-19 would be 5.6% for China and 15.2% outside of China [1]. More recently, Johns Hopkins Coronavirus Resource Center estimated that the average mortality rate of COVID-19 was about 5.5% in China, but 7.6% for Wuhan [2]. Without a doubt, COVID-19 poses a major challenge to global public health.
To ght against COVID-19, prevention is of particular importance. Since the outbreak of virus, the Chinese authorities have taken various prevention and intervention measures, including school closure, cancelling public gathering, restricting movements of persons, isolation of infected individuals, and tracing contact.
The general recommendations issued by the WHO include: washing hands frequently with water and soap, wearing a face mask, social distancing, staying home, avoiding touching of the eyes, nose and mouth, practicing respiratory hygiene, and developing a healthy lifestyle with a proper diet, etc. [3]. In the early phase of the virus, Chinese people were found to be actively engaged in adopting various preventive measuring, especially wearing face masks and washing hands [4]. These measures were found to be effective. The number of con rmed cases of COVID-19 decreased substantially when control measures were implemented [5]. A number of scientists argued that personal preventive behaviors were the most effective ways of preventing the spread of the disease [6]. More importantly, the outbreak of the virus triggered adverse psychological impacts and psychiatric symptoms among the public [7]. And studies found that precautionary measures people adopted have protective psychological effects during the early stage of the epidemic in China [8].
Understanding public responses toward COVID-19 is vital for combating the virus. Therefore, it is important to explore the antecedents to desirable health behaviors during this global pandemic. Previous studies have proposed various behavioral models to explain preventive behaviors at the individual level: the Theory of Reasoned Action, the Health Belief Model, the Social Cognitive Model, the extended parallel process model (EPPM), etc. Based on these models, researchers identi ed a host of factors leading to preventive behavior during the COVID-19 pandemic: risk perception, self-e cacy, negative emotion such as fear and sadness, knowledge, belief in conspiracy, socioeconomic status, etc. [9][10][11][12][13]. At the early stage of COVID-19, a number of studies have been conducted to identify which factors predicted preventive behaviors in China. Knowledge and concerns about COVID-19, optimistic attitudes, perceived severity of the disease, political trust, and support for the government and health authorities were found to be signi cantly associated with a higher likelihood of taking preventive measures for COVID-19 [14][15][16]. However, few studies have juxtaposed cognitive and affective variables in one study.
On Feb 27, the National Health Commission of the People's Republic of China (NHC) announced that the number of newly con rmed COVID-19 cases in Hubei outside Wuhan and outside Hubei dropped to single-digit numbers for the rst time. During the initial stage of COVID-19 outbreak, intensive media coverage on the topic in China not only increased people's knowledge about the virus, but also triggered fears among the public. After the rising phase of the virus outbreak, the decreasing number of con rmed cases and mortality may in uence people's estimation about the severity and controllability of COVID-19.
In addition, there was a reduction in the levels of stress, anxiety and depression before and after the rising period of COVID-19 in China [17]. Therefore, we ask the following two questions: (1) What are the roles of cognitive and affective variables in predicting behavioral responses toward COVID-19 after the rising phase of COVID-19 in China? (2) Which factor (cognition vs. affect) possesses greater explanatory power in predicting people's behavioral responses? In this study, we used data from a cross-sectional survey elded after the rising phase of the virus outbreak in China to further our understanding of preventive measures in response to COVID-19 in China.

Study design, procedure and participants
This study started data collection in China on March 2, 2020, when the spread of COVID-19 had already lasted more than three months since the rst 41 infected cases of COVID-19 were publicized on December 31, 2019. The data collection ended on March 23, when the virus reached more than 190 countries and infected nearly three million people globally. At the time of writing, the virus has led to 83,325 infected individuals and 4,634 deaths in China [18].
Data for this study were obtained from an online survey of nationally representative internet users in China. The data collection was contracted out to a commercial marketing research company in China. The company maintains a large panel pool of 4 million participants covering all Chinese provinces. In order to achieve a representative sample, the study employs a strati ed quota sampling technique to recruit respondents. The quotas for subcategories of gender, age and education groups are according to the most recent CNNIC (China Internet Network Information Center) report [19]. A total of 3,000 respondents of 18 years' old or above lled out the survey with a response rate of 24.56%. This study was approved by the Institutional Review Board of Fudan University. All the methods in the study were conducted according to the criteria set by the ethics committee of Fudan University and each subject signed an informed consent before participating to the study.

Knowledge of COVID-19
Four questions were designed from popular misinformation regarding COVID-19 in China. Previous studies measured COVID-19 knowledge using questions related to the clinical presentations, transmission routs, prevention and control, etc. Owing to the overwhelming number of news reports on this public health emergency, prior studies found that most people have known basic information regarding COVID-19 [20][21]. Since knowledgeable individuals are less likely to be misguided by misinformation, a better way of distinguishing people with high levels of knowledge from the rest is to measure their ability to identify misinformation. We asked respondents to indicate whether they believed the statement we presented regarding COVID-19 on a 4-point scale, ranging from 'de nitely false' to 'de nitely true'. When a respondent chose 'de nitely true' or 'true', they received 1 point if the statement is correct, and 0 point if the statement is wrong. We summed the points of all four questions to formulate an index for knowledge of COVID-19, ranging from 0 to 4.

Risk perception
Respondents were asked to estimate their own risk of COVID-19 infection, other people's risk of infection, as well as risk of dying from the virus on a 0-100 scale, with 100 being the highest perceived risk. We used the average score of the three indicators to operationalize risk perception.

Negative emotion
Negative emotion was measured by asking respondents to report their levels of 'sadness', 'fear', 'anger' and 'shock' in response to COVID-19 on a 5-point scale (1 = strongly disagree, 5 = strongly agree). The items were averaged for each respondent to formulate an indicator of negative emotion.

Going out for activities
Respondents were asked to report the frequency of going out for work, entertainment, for daily life necessities, and for social functions since the Wuhan lockdown on January 23, 2020 on a 4-point scale ranging from 1 to 4 where 1 means never and 4 means frequently. Respondents' answers were averaged to form a score where higher score indicates higher frequency.

Preventive behaviors
Using the same 4-point scale, respondents were asked to report how frequently they had taken the following measures recommended by the WHO to protect themselves from COVID-19: wearing masks, washing hands with water and soap, avoiding hand shaking or hugging, avoiding eating in restaurants. Their answers were averaged to form an index where higher scores represent higher levels of preventive behavior.

Demographics
Respondents were also asked about their age, gender, education attainment, marital status, living area, and household monthly income.

Data analysis
All statistical analyses were performed using SPSS statistical software, version 26. Foremost, we performed the descriptive analyses to present the means and the Cronbach's Alphas of the variables. Then, multiple regressions were conducted to examine and to compare the predictors of behavioral responses toward COVID-19. First, we only entered demographic factors into the model to examine how much variance in behavioral responses can be explained by the control variables. Then we added cognitive and affective variables separately into the model to compare the explanatory power of these variables. Last, we entered all the predictors simultaneously into the model.

Sample characteristics
Of the total respondents (N = 3,000), 47.6 percent were female, and more than 60 percent of them aged below 40 years. About 38.1 percent of the respondents nished secondary school and 75.3 percent of the respondents were married. In addition, nearly half had a monthly average household income between 10,001 and 30,000 yuan (USD 1,414 to 4,240). The majority of the respondents lived in urban areas (60.8%). The distribution of demographic variables (age, gender, education and income) of our sample are very similar to those reported in the latest CNNIC report (Table 1). Descriptive analysis According to Table 2

Predicting behavioral responses towards COVID-19
Multiple regression analyses were performed to examine factors predicting behavioral response towards COVID-19. For going out for activities (Table 3), the explanatory power of the demographic factors was fairly small (3.70%). Introducing cognitive factor into the model explained only an additional 1.00% of the variance in frequency of going out for activities (F (7,2992) = 21.47, p < .001). As Model 2 shows, knowledge was negatively related to the frequency of going out for activities (β=-0.10, p < .001). Model 3 demonstrates that negative emotion (β = 0.34, p < .001) and risk perception (β = 0.04, p = .01) were also signi cant predictors of frequency of going out for activities. After adding the affective variables, the R 2 increased from 3.70 to 15.90 (F (8, 2991) = 70.50, p < .001).
These results showed that cognitive and affective variables were important explanatory variables, but affective variables had stronger explanatory power in predicting frequency of going out for activities.
We then entered all cognitive and affective variables simultaneously into the model (Model 4). Results from Model 4 were consistent with those from Model 2 and 3. We performed the same set of analyses for predicting preventive behaviors. As was shown in Table 4, demographic variables only explained about 3.70% of the variance in preventive behavior adoption (Model 5). Cognitive predictors explained 4.40% of the variance (F (7, 2992) = 37.89, p < .001) and the two affective variables explained 8% of the variance (F (8, 2991) = 49.70, p < .001). The explanatory power of the affective variables was greater than that of the cognitive variable.

Principal Results
The aim of this study was to investigate the role of cognitive and affective factors in predicting behavioral responses among adult Chinese after the initial rising phase of the virus outbreak. Overall, our study found that both cognitive and affective factors were both signi cant predictors of going out activities and preventive behaviors during the COVID-19 pandemic.
Similar to previous studies, this study demonstrated that knowledge was a critical explanatory factor of behavioral responses toward the epidemic disease. The more knowledgeable the respondent is, he or she will be more likely to adopt preventive measures and less likely to go out for activities. During the very early stage of the epidemic, studies showed that knowledge was signi cantly associated with a lower likelihood of potentially dangerous practices in response to COVID-19 in China [22]. This is consistent with our ndings, which highlight the importance of knowledge. It is worth mentioning that the overall correct response rate of the knowledge questions in our study is only 7.5 percent, which is much lower than those in previous studies conducted at the early phase of COVID-19. Such a gap is primarily because of the design of our knowledge measures. Adapted from widespread misinformation about the virus, these questions are a little tricky for people to judge. In addition, most of the previous studies conducted online collected data using non-probability sampling method, which might not re ect the entire Chinese online population. For instance, an overall correct rate of 90% on the COVID-19 knowledge was based on a sample with highly educational attainment (82.4% of the respondents held an associate's degree or higher) [23]. In our study, the sample characteristics of the young, low education, and low income could be another possible explanation.
This study also found that affective factors played important roles in predicting behavioral outcomes.
The more negative emotions an individual experienced, the more likely they were to go out for different activities and the less likely they were to take preventive measures. This is inconsistent with the ndings from recent studies showing that negative emotion, exaggerated levels of anxiety and fear of COVID-19 in particular, was associated with positive behavioral change such as social distancing and improved hand hygiene [24]. Previous studies found that different patterns of emotional responses resulted in different coping strategies toward problems. Fear was linked with desire for protection but sadness was more likely to be associated with non-active coping methods [25]. By this logic, it is possible that our negative emotion measure was dominated by sadness more than fear. According to our data, the most common emotion respondents reported that they experienced were shock (3.70 out of 5) and sadness (3.47 out of 5). Another explanation could be the reverse causal relationship between emotion and preventive behaviors. During the rapid rise period of the COVID-19 outbreak, personal prevention measures were found to be associated with lower levels of stress, anxiety and depression in China [26]. In our study, the pattern is the same. The more adoption of preventive measures, the lower levels of negative emotion one would experience (b=-0.09, p = .006). In turn, after the rising phase of the virus outbreak, low levels of negative emotion could lead to more preventive behaviors.
Perceived risk was also found to be negatively associated with the frequency of adopting preventive behaviors. Previous studies identi ed three hypotheses between risk perception and health behaviors.
First, the accuracy hypothesis asserts that perception of risk at any time re ects one's risk behaviors at that time. Second, the behavior motivation hypothesis argues that risk perception at one time leads to increased preventive behaviors at a later time. Third, the risk reappraisal hypothesis indicates that if an action is believed to reduce risk, people who take the action will lower their personal perception of risk [27]. Findings from a recent study supported the behavioral motivation hypothesis at the early stage of COVID-19. Higher levels of perceived risk of infecting coronavirus seem to be associated with greater

Limitations And Conclusions
Like other empirical studies, our study has several limitations. First, we cannot ascertain the causal relationships between cognition, affect, and behavior using cross-sectional data. Second, although we have operationalized our variables carefully, we admit that the measures for emotion and knowledge were not perfect. For instance, worry, one of the important negative emotions when facing an epidemic, was not assessed in our study. Third, our study merely focused on internet users in China, and extra caution is needed when generalizing these ndings to the whole Chinese population. For people who propensity of respondents to engage in preventive behaviors such as hand washing and social distancing [28]. In our study, the ndings supported the reappraisal hypothesis. More preventive behaviors (b=-3.08, p < .001) and less going out for activities (b = 3.30, p < .001) were associated with lower levels of risk perception.
Taken together, our study found that cognitive and affective variables both played important roles in predicting behavioral responses towards COVID-19. More importantly, affective variables demonstrated stronger explanatory power in predicting both going out and preventive behaviors in our case. Based upon cognitive appraisal theory, there are two stages for people to evaluate how a particular event will affect their well-being: the primary appraisal refers to a person's evaluation of whether he or she is at risk when facing the event; the secondary appraisal refers to a person's evaluation of what can be done to cope with the situation or to prevent potential harm [29]. Previous studies considered risk perception as the primary appraisal while people's knowledge as the secondary appraisal. At the early stage of COVID-19, the primary appraisal was found to be more important in predicting preventive behavior [30]. Findings in our study shows that even after the rising phase of the virus outbreak, the rst appraisal still played a more dominant role. Another explanation could be that the strict control measures implemented by Chinese government, as well as the effective preventive behaviors people adopted decreased Chinese people's perceived risk of the incident, which might in uence their behavioral responses.

Implication
Findings of this study provide some practical implications for ghting COVID-19. First, since affect seems to be a crucial determinant of preventive behavior adoption, it is important to decrease people's negative emotion and risk perception. Therefore, the media should be careful when covering the pandemic. Constant reporting of rising numbers of victims, dissatisfaction toward government policies, tear-jerking stories, and nger-pointing sound bites could all possibly drive up high negative emotion among the public. Delivering timely and factual information regarding virus prevention and cautioning against the over ow of negative emotion seem to be important strategies for news coverage during the COVID-19 pandemic.
Second, knowledge is also an important predictor of adopting appropriate behavioral, it is therefore important to improve people's coronavirus knowledge through public health information campaigns. Such information should target people who are less likely to adopt preventive measures.
have no access to the internet, their behavioral pattern could be different because of their lower socioeconomic status and lack of access to information. Third, our study was conducted in China, a society with a set of special cultural and political features. Despite the fact that our ndings cannot be blindly generalized to the other societies, we provide empirical evidence of behavioral responses towards All the methods in the study were conducted according to the criteria set by the ethics committee of Fudan University.
Each subject signed an informed consent before participating to the study.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Authors' contributions FS conceptualized, analyzed and interpreted the data. He also revised the edited the manuscript. CM was a major contributor in writing the manuscript. YL conceptualized the data. YJ C collected the data. All authors read and approved the nal manuscript.