Tailored interventions into broad attitude networks towards the COVID-19 pandemic

This study examines how broad attitude networks are affected by tailored interventions aimed at variables selected based on their connectiveness with other variables. We first computed a broad attitude network based on a large-scale cross-sectional COVID-19 survey (N = 6,093). Over a period of approximately 10 weeks, participants were invited five times to complete this survey, with the third and fifth wave including interventions aimed at manipulating specific variables in the broad COVID-19 attitude network. Results suggest that targeted interventions that yield relatively strong effects on variables central to a broad attitude network have downstream effects on connected variables, which can be partially explained by the variables the interventions were aimed at. We conclude that broad attitude network structures can reveal important relations between variables that can help to design new interventions.

emphasizes the importance of attitudes, behavioral control and the subjective norm on behavior. The HBM [3,4] focusses on perceived susceptibility and severity of the disease, benefits and barriers of behavior, self-efficacy and cues to action to explain health-related behavior. Studies into compliance with behavioral measures in the context of pandemics confirm the importance of the factors in these models, but also identify additional factors to play a role. More specifically, recent research into the COVID-19 pandemic associated multiple factors from the TPB and HMB with compliance, such as attitudes [5], social norms [6], self-efficacy [7], perceived controllability and severity [8], and risk perception [9].
Factors that contribute to explaining behavior during pandemics in addition to those in TPB and HMB are for instance trust in authorities and science [1, 9,10], knowledge [11,12] and demographics and personality variables [13,14]. This underlines the relevance of including a broad range of psychological factors extending beyond those in the prevailing models of (health) behavior to explain compliance with behavioral measures during this unprecedented and complex COVID-19 pandemic.
Taking into account the effects of the pandemic on the public's well-being is important. For instance, social isolation and loneliness, both potential consequences of the behavioral measures during the COVID-19 pandemic, are associated with adverse mental and physical health effects [15]. Previous research also found associations between anxiety and preventive behaviors during pandemics [1]. More specifically, fear as a result from the pandemic is related to compliance with behavioral measures during the COVID-19 pandemic [13,16], which further emphasizes the importance of taking into account (changes in) wellbeing during pandemics.
In conclusion, understanding the complex interplay of psychological factors underlying compliance with behavioral measures during the COVID-19 pandemic requires an integral and systemic approach, encompassing a broad range of variables relevant to behavioral compliance as well as factors relating to mental and physical health.

Economic Consequences
For me personally, I consider the economic consequences of the corona pandemic .. 1 (Extremely small) to 7 (Extremely severe) For my family and friends, I consider the economic consequences of the corona pandemic are..

Self-exempting Beliefs
I will not get infected with the coronavirus because I never get the seasonal flu (influenza) either.
1 (Strongly disagree) to 7 (Strongly agree) I think I am already immune (protected) against the coronavirus.

Negative Affect
The corona pandemic is making me angry. 1 (Strongly disagree) to 7 (Strongly agree) The corona pandemic is making me sad.

5
The corona pandemic is making me feel confused. The corona pandemic is making me feel uncertain. The corona pandemic is making me feel overwhelmed. The corona pandemic is making me feel frustrated. The corona pandemic is making me fearful. The corona pandemic is making me feel out of control.

Compassion
The corona pandemic is making me feel compassion.

Consideration of future consequences
Often I engage in particular behavior in order to achieve outcomes that may not result for many years. (R) 1 (Strongly disagree) to 7 (Strongly agree) My convenience is a big factor in the decisions I make or the actions I take. (R) I am willing to sacrifice my immediate happiness or well-being in order to achieve future outcomes. I think it is important to take warnings about negative outcomes seriously even if the negative outcome will not occur for many years. I generally ignore warnings about possible future problems because I think the problems will be resolved before they reach crisis level. To what extent did you experience feeling weak during the past two weeks?

Depressive Complaints
To what extent did you experience a loss of the will to live during the past two weeks? 1 (Not at all) to 5 (Very much) To what extent did you experience feeling lonely during the past two weeks? To what extent did you experience feeling blue during the past two weeks? To what extent did you experience a loss of interest in things during the past two weeks? To what extent did you experience feeling hopeless about the future during the past two weeks? To what extent did you experience a feeling of worthlessness during the past two weeks?

Anxiety Complaints
To what extent did you experience nervousness during the past two weeks?

Trust low condition
Before we continue to the third survey, we would like to share a couple of news articles with you.
Source: https://nieuws.nl/algemeen/20200512/peiling-minder-mensen-positief-over-corona-aanpak-overheid/ Trust in that the government can properly manage the corona outbreak has decreased. The number of people who have that confidence has fluctuated around 70 percent for a long time, but has now decreased to 63 percent. This is the conclusion of consultancy firm Citisens after a poll among more than 2,500 people from its own research panel.

The current study
The current study also shows that trust in the responsible authorities and experts has decreased. This could mean that the events and changes in policy of the past two weeks have a negative effect on people's trust in the authorities, healthcare and science.

Explanation
The decrease of trust is due to the fact that the responsible authorities and experts think differently about how to manage the pandemic and disagree on key points. The rules are also perceived as unclear and sometimes contradictory. It seems that the authorities have little understanding of how the virus behaves and how it can be effectively managed.
To know if you have seen the information above, we ask you a question about this: According to the information above, has the level of trust in the responsible authorities and experts increased, decreased or remained the same?

Increased
The cabinet's corona strategy is receiving increasing criticism from the House of Representatives. Although left and right want to make adjustments in a completely different manner.

Criticism of Prime Minister Rutte's corona policy comes from both left and right wing parties
Picture of Dutch politicians during a debate 2. Decreased

Trust high condition
Before we continue to the third survey, we would like to share a couple of news articles with you.
Source: https://www.ad.nl/binnenland/onderzoek-wijst-uit-dat-nederlanders-vertrouwen-hebben-in-aanpak-van-het-coronavirus~a29de95e/ (Photo adjusted) The Dutch have confidence in the approach to the coronavirus. This is shown in an interim report of the research into the impact of the virus, carried out by RIVM and Nivel, the Dutch institute for health research.

Research shows that the Dutch have confidence in the approach to the coronavirus
Picture of Jaap van Dissel, chairman of the Dutch Outbreak Management Team (photo adjusted) Source: https://nieuws.nl/algemeen/20200331/nederlanders-tevreden-met-corona-aanpak-overheid/

The current study
The current study also shows that trust in the responsible authorities and experts has increased. This could mean that the events and changes in policy of the past two weeks have a positive effect on people's trust in the authorities, healthcare and science.

Explanation
The increase in trust is due to the fact that the responsible authorities and experts share the same views on how to manage the pandemic and agree on key points. It is a fact that the rules can be complex, but the basis for the policy is reliable scientific research and the communication from the government about the policy is clear.
To know if you have seen the information above, we ask you a question about this: According to the information above, has the level of trust in the responsible authorities and experts increased, decreased or remained the same?

Social Norm low condition
Before we continue to the third survey, we would like to share a couple of news articles with you.
Source: https://nos.nl/artikel/2334600-drukte-op-hemelvaartsdag-parken-ontruimd.html In several regions in the Netherlands, the beautiful weather on Ascension Day led to crowds. Roads and areas are closed to prevent the influx of day trippers.

Busy on Ascension Day, parks vacated
Picture of an enforcer in a busy park Source: https://www.volkskrant.nl/nieuws-achtergrond/peiling-draagvlak-voor-maatregelen-neemt-af-onder-nederlanders~bd7f3285/ (Photo adjusted) The current study The current study also shows that people adhere less to the corona measures. The extent to which people find it important that others adhere to the corona measures has also decreased.

Explanation
The fact that people adhere less to the corona measures is because an increasing amount of people believes that the measures can be loosened, since the Intensive Care Unit statistics leave plenty of room for this and the economic damage will otherwise be irreparable.
To know if you have seen the information above, we ask you a question about this: According to the information above, has the level of compliance with the corona measures increased, decreased or remained the same?

Social Norm high condition
Before we continue to the third survey, we would like to share a couple of news articles with you.

Picture of sign with the text 'keep distance'
Almost all Dutch people adhere to the behavioral rules that must prevent infection with the coronavirus. 99 percent says they keep the requested 1.5 meter distance from others, 97 percent wash their hands more often and 93 percent stays at home as much as possible. More than half addresses others about violating these rules.

The current study
Despite the fact that some places are getting crowded, research shows that the vast majority of people still comply with the rules. The current study also shows that people increasingly adhere to the corona measures. The extent to which people find it important that others adhere to the corona measures has also increased.

Explanation
The fact that people adhere more to the corona measures is because an increasing amount of people believes that the measures are proportional to the current phase of the corona crisis. Slightly looser where possible, but careful, to avoid later regrets, as Prime Minister Rutte says.
To know if you have seen the information above, we ask you a question about this: According to the information above, has the level of compliance with the corona measures increased, decreased or remained the same?

The Dutch stay at home despite beautiful spring weather
Picture of two people walking outside in nature Despite the beautiful spring weather, the Netherlands seems to be following the advice to stay at home as much as possible. According to various regions and authorities, it has been quiet today in the nature reserves. To know if you have seen the information above, we ask you a question about this: According to the information above, how many new subscribers has Netflix welcomed in the first quarter of 2020?

Record number of new Netflix subscribers due to coronavirus
Picture of Netflix tv screen and smart phone with Netflix on its screen Netflix has never welcomed as many new subscribers in a quarter as in the first three months of 2020: 15.8 million worldwide, the streaming service announced on Wednesday. The record increase is the result of the corona measures, which means that many people have to stay at home.

Measures Support low condition
Before moving on to the final questionnaire, we would like to share some news items with you.

'Corona measures actually cause more deaths in the long term'
Picture of Ira Helsloot (professor Governance of Safety and Security at the Radboud University) The measures taken to combat the coronavirus cost more lives than the deaths they prevent. "Only you don't see the deaths that will fall in two years." Source: https://www.ad.nl/binnenland/wirwar-aan-winkelregels-funest-voor-draagvlak-coronamaatregelen~a141c30c/ (Photo adjusted) The current study The current study also shows that the corona measures are seen as exaggerated and unnecessary. Support for current measures remains low.

Explanation
The decreasing support for the corona measures is due to it becoming increasingly clear what To know if you have seen the information above, we ask you a question about this: According to the information above, are the effects of corona measures positive, negative or neutral?

Measures Support high condition
Before moving on to the final questionnaire, we would like to share some news items with you.

Picture of Eiffel tower in Paris and Big Ben in London
The strict lockdowns introduced in several European countries have saved millions of lives. British research shows that the number of deaths in Europe in early May, without lockdowns, would have risen to 3.1 million. In reality, 135,000 people had died at the time.
Source: https://www.ad.nl/politiek/br-rivm-90-procent-ic-opnames-voorkomen-door-coronamaatregelen~a2dd5dc5/ (photo adjusted) The current study The current study also shows that the corona measures are seen as effective and necessary. good fit with the phase of the corona crisis we are currently in. We have avoided even worse consequences together and now it is a matter of persistence.
To know if you have seen the information above, we ask you a question about this: According to the information above, are the effects of corona measures positive, negative or neutral?

Economic Consequences low condition
Before moving on to the final questionnaire, we would like to share some news items with you.
Source: https://www.telegraaf.nl/financieel/1582437355/economische-impact-coronavirus-valt-mogelijk-mee (photo adjusted) The economic impact of the new coronavirus may not be as serious as the rating agency S&P Global recently predicted. The International Monetary Fund (IMF) is a bit more cautious when it comes to estimating the damage, although Kristalina Georgieva (IMF) emphasized this weekend that the damage will be more serious if the virus continues to spread.

Economic impact of the coronavirus may be less serious than expected
Picture of empty terrace from the catering industry (photo adjusted) Source: https://www.rtlnieuws.nl/economie/artikel/5072101/recessie-corona-covid19-abn-amro-economie-daalt-krimpscenario-crisis (photo adjusted) The current study The current study also shows that people expect the economic damage to be less serious than expected. The pandemic will have adverse effects on our economy, but these will be temporary. People are optimistic about the economic impact of the pandemic on them personally. The economic consequences for family and friends are also estimated to be relatively small.

Explanation
The economic consequences of the Dutch policy during the corona pandemic are now estimated to be smaller than previously expected. The government's financial support is

ABN AMRO: corona recession becomes deep, but short
Picture of travelers at an airport (photo adjusted) The recession caused by the corona pandemic will be deep, but short-lived. That is what the economists of ABN Amro estimate. They assume that Dutch GDP will shrink by 3.5 percent this year. more than sufficient to deal with the consequences of the pandemic. The economy will also quickly recover due to the recent loosening of measures.
To find out whether you have seen the information above, we ask you a question about this: According to the information above, are the economic consequences more serious than expected, less serious than expected or neutral?
1. More serious than expected 2. Less serious than expected 3. Neutral

Economic Consequences high condition
Before moving on to the final questionnaire, we would like to share some news items with you.
Source: https://nos.nl/artikel/2337409-cpb-ongekende-krimp-van-6-procent-werkloosheid-verdubbelt.html The Dutch economy is shrinking by 6 percent this year due to the corona outbreak. Next year there will be limited recovery with a growth of 3 percent. Unemployment will rise and will double in 2021, the Central Planning Bureau (CPB) expects. The CPB speaks of an unprecedented shrinkage and warns of great uncertainty about the course of the corona crisis.

The current study
The current study also shows that people expect that the economic damage will be enormous. The economic consequences of the pandemic, both personally and for family and friends, are estimated to be very serious.

Explanation
The seriousness of the economic consequences of the Dutch policy during the corona pandemic is becoming increasingly clear. Our economy is shrinking and unemployment is rising. There is even talk of the worst economic crisis since World War II.
To know whether you have seen the information above, we ask you a question about this: According to the information above, are the economic consequences very serious, less serious than expected or neutral? "You can certainly say that this is the biggest economic crisis since the Second World War," says Peter Hein van Mulligen, chief economist at Statistics Netherlands. This morning Statistics Netherlands presented the results of the Dutch economy in the first quarter. And they did not look good, Van Mulligen sees. "I can't make it more fun." The first quarter of 2020 shows a shrinkage of 1.7 percent. After good results in January and February, a dramatic downturn in the last two weeks of March wiped out that prosperity.

Node construction and Principal Axis Factoring
Note that this section is identical to the section on node construction in Chambon, Dalege [17], with addition of variables that were measured once (i.e., Consideration of Future Consequences, Resilience and Coping).
The analysis for node construction was conducted with the largest and most diverse sample available: wave 1, including all participants that completed the first sample (i.e., no drop-out). The survey items were combined to form nodes: the combination of items was either predetermined by validated scales or a fixed operationalization or identified through Principal Axis Factoring (PAF; see Table S2). PAF was conducted with Oblimin rotation given the expected intercorrelation between items. Extraction of components was based on eigenvalues greater than one. The PAF results are discussed below. Items presented in italic were excluded from the node. The corona pandemic is making me angry. .729 2 The corona pandemic is making me sad. .475 3 The corona pandemic is making me feel confused. .678 4 The corona pandemic is making me feel uncertain. .649 5 The corona pandemic is making me feel overwhelmed. .472 6 The corona pandemic is making me feel frustrated. .890 7 The corona pandemic is making me fearful. .572 8 The corona pandemic is making me feel out of control. .628 9 The corona pandemic is making me feel compassion. .511

Detailed description per node
This section provides detailed description of each node in the COVID-19 broad attitude network, its interpretation and the scale reliability as observed in the current study.
Note that this section is similar to the node description in Chambon, Dalege [17], with addition of variables on individual differences that were measured once (i.e., Consideration of Future Consequences, Resilience and Coping).

Compliance
Compliance with behavioral measures was operationalized through the preventive behaviors as recommended by the Dutch national government [19] and National Institute for Public Health and the Environment [20]. We measured the (self-reported) extent to which participants adopted the preventive behaviors that were advised to the general public, regardless of symptoms (i.e., hygiene behaviors and physical distancing). Due to changes in the recommended behaviors and policy, the items underlying this node also changed over time. Specifically, the node consisted of five items in wave 1 (a = .75), of which one item was adapted in wave 2, and then a sixth item was added in wave 3 (resulting in six items from wave 3 onwards). Higher scores indicate that participants reported to display those behaviors much more throughout time and therefore reflect a higher degree of compliance with the behavioral measures.

Attitudes
The following nodes consisted of items that measured the different elements of attitudes relevant for the pandemic and the behavioral measures.
Based on prior research into psychological factors during the COVID-19 pandemic [21], we included items that formed the cognitive nodes Risk Perception, Health Risk (rsb = .59) and Economic Consequences (rsb = .75). The node Risk Perception (i.e., likelihood and severity of infection for oneself) was calculated by multiplying the scores on the two items [22]. Higher scores on the nodes Risk Perception and Health Risk indicate a higher perceived risk of infection and severe health consequences due to an infection, respectively. Higher scores on Economic Consequences indicate that participants expected the economic consequences of the pandemic to be more severe. Additionally, the items on Self-exempting Beliefs (rsb = .57) resulted in a node for which a higher score reflects stronger convictions about not being susceptible to the coronavirus.
The items on affect resulted in two nodes: Negative Affect (e.g., anger, anxiety and confusion; a = .89) and Compassion (single item). Higher scores indicate that the pandemic caused these emotions to be experienced more often. The construct worries resulted in the nodes Worries Virus (a = .73), with items encompassing worries about events during the pandemic that resulted directly from the virus (e.g., getting infected, losing someone they love), and Worries Measures (a = .67), with items concerning worries about events that resulted from the measures taken due to the virus (e.g., overloading the health care system, a recession). Higher scores reflect more worries.
Three behavioral attitude nodes were identified. The first node was formed by a single item that measured the intention to get vaccinated if a vaccine becomes available (Vaccination Intention), with a higher score indicating more intention. The items on attitudes toward the behavioral measures (general items and semantic differential scale items) resulted in two nodes: Measures Support (a = .90) and Measures Ease (rsb = .68), with items measuring the participants' support for the behavioral measures aimed at preventing the spread of the coronavirus, and the extent to which these measures were perceived as easy to comply to, respectively. Participants with a higher score on these nodes reported more support for and perceived ease of the behavioral measures.

Additional psychological factors
The following nodes were included in addition to attitudes.
The node Social Norm (rsb = .76) consisted of items measuring the prescriptive and descriptive social norm, combinedly forming one node for which a higher score implies participants perceived a stronger norm towards compliance with the behavioral measures.
Analysis of the items relating to perceived control resulted in two nodes: Control Infection (rsb = .61), with items measuring the extent to which people felt they could avoid an infection with the coronavirus, and Self-efficacy, a single item measuring whether people know how to protect themselves from the virus. Higher scores indicate more perceived control over an infection with the coronavirus and perceived self-efficacy, respectively.
The node Involvement (a = .84) consisted of items for which a higher score means participants perceived themselves to be more actively involved in the corona pandemic (e.g., allocated importance and watching the news), which is perceived as a dimension of attitude strength.
The node Perceived Knowledge was formed by a single item with higher scores meaning participants reported more (self-perceived) knowledge about the pandemic.
Trust (a = .86) was measured with items that formed a node by combining general trust in four actors that are crucial in the corona pandemic (adapted to the Netherlands): the authorities, the Dutch National Institute for Public Health and the Environment (RIVM), health care professionals and science. Higher scores reflect more trust in these actors.
Individual differences were included by adopting three available scales with items that measured personality aspects that were expected to be a relatively stable indication of people's response to the current pandemic, and therefore included only at wave 1. More specifically, we included the following scales as nodes:

Health and well-being nodes
Finally, several health-and well-being-related nodes were included into the broader network with attitudes and additional psychological factors. More specifically, we incorporated health-related nodes that could influence people's vulnerability (e.g., general health, illness) and thus fear of this pandemic, and thereby possibly affecting compliance. 2 Although the scale reliability of some nodes was lower than one would normally prefer, these were interpreted as sufficient given our objective of measuring evaluations instead of designing reliable measurement scales ( Furthermore, we included health-related nodes that could be influenced by the pandemic and associated behavioral measures (e.g., healthy lifestyle, psychological complaints).
Health was measured with the single item node General Health, consisting of an overall score participants assign to their health, with a higher score indicating better health.
The Jong Gierveld and van Tilburg [28]. We adjusted the answer scale and specified a timeframe of two weeks to meet the objectives of the current study. Higher scores indicate more loneliness.

Additional information on Network Analysis
The communities in the networks were determined through a community stability and detection analysis with the cluster walktrap algorithm. Stability analysis was conducted by repeating the community analysis (1000 iterations) and calculating how often different nodes belonged to the same community. This resulted in a score between 0 and 1 for each combination of nodes, in which 0 meant that these nodes never belonged to the same community, and 1 that these nodes belonged to the same community in every iteration.
Subsequently, a community detection analysis was conducted by identifying the communities with nodes that belonged to that community in over 90 percent of the iterations. The output of the community analyses is provided in section 3.1.2..
We used the package qgraph [33] to visualize the graphs, bootnet [34] for the stability and accuracy measures and igraph [35] for the community detection.

Additional information on results
3.1 Part 1 -Network structure 3.1.1 Detailed descriptive account of COVID-19 broad attitude network The COVID-19 broad attitude network, obtained through nodewise regression, is shown in Fig S1a (left). Nodes represent the measured psychological factors. In general terms, the right section of the network displays the (psychological) health nodes. The left half of the network is comprised of the cognitive and behavioral attitude nodes and the additional psychological nodes (e.g., social norm, perceived knowledge, trust and perceived control).
The colored groups represent communities (i.e., clusters with higher interconnectedness) that consist of nodes that were more connected to each other than to other nodes. The yellow cluster indicates that attitudes, the additional social psychological nodes and compliance with behavioral measures form a community in the COVID-19 broad attitude network as they were highly interconnected. This also applies to the orange community with nodes on mental health and well-being and the purple community with nodes on current and perceived risk of physical health.   Output on edge accuracy and centrality stability is also displayed on the next pages.

Edge weights
The edge accuracy plot is displayed to give a general impression of the width of the edge weight confidence intervals. Readers are referred to OSF for detailed information of specific edges.

Method
These analyses are conducted with bootstrap network estimation. In doing so, we specified the method to use as mgm (i.e., package bootnet, function bootnet, argument default = "mgm").

Interpretation guidance difference tests (see OSF)
Edge difference: a = .05, black boxes indicate significant differences between edges.
Centrality difference: a = .05, black boxes indicate significant differences between nodes, node strength is presented in the diagonal boxes.

Descriptives intervention conditions
Statistics of the nodes of wave 3 and 5, dived into the intervention conditions. First  * p <.01; indicating a significant difference between the low, high and/or control condition of the intervention. A specification of these differences is provided in the manuscript.  (14) 16 (14) 16 (14) 16 (14) 16 (14) 15 (15) 16 (15) 17 (1) 1 7-point Likert-scale, unless marked ~, indicating a 5-point Likert-scale. Please note that Risk Perception is the product of two items and Mental Well-being is the metric sum score of all items. * p <.01; ** p <.001; indicating a significant difference between the low, high and/or control condition of the intervention. A specification of these differences is provided in the manuscript. * p <.01; indicating a significant difference between the low, high and/or control condition of the intervention. A specification of these differences is provided in the manuscript.