Anxiety in the face of the first wave of the spread of COVID-19 in Israel: Psychosocial determinants of a “Panic-to-complacency-continuum”

Abstract

Background and Methods

Based on an established ongoing prospective-longitudinal study examining anxiety in response to COVID-19, a representative sample of 1018 Jewish-Israeli adults were recruited online. A baseline assessment was employed two days prior to the first spread of COVID-19, followed by six weekly assessments. Three classes of general anxiety and virus-specific anxiety were identified: (1) “Panic” (a very high and stable anxiety throughout the spread), (2) “Complacency” (a very low and stable anxiety throughout the spread), and (3) “Threat-Sensitivity” (a linear increase, plateauing at the 5th wave). For general-anxiety only, a fourth, “Balanced,” class was identified, exhibiting a stable, middle-level of anxiety. We tested theory-based, baseline, social-cognitive predictors of these classes: self-criticism, perceived social support, and perceptions/attitudes towards the Israeli Ministry of Health. We also controlled for trait anxiety. Multinomial regression analyses in the context of General Mixture Modeling were utilized.

Results

Baseline virus-specific anxiety linearly predicted emerging virus-specific anxiety classes. Virus-specific panic has higher trait anxiety than the other two classes. The general anxiety panic class was over-represented by women and exhibited higher baseline general anxiety and self-criticism than all other classes, and higher baseline virus-specific anxiety along with lower perceived support and less positive perceptions of the ministry of health than two of the three other classes.

Conclusions

Preexisting anxiety shapes subsequent anxious responses to the spread of COVID-19. The general-anxiety panic class may be markedly demoralized, requiring targeted public-health interventions.

1. Introduction

Since its eruption in December 2019, the coronavirus (COVID-19) pandemic has delivered devastating blows to societies worldwide. This predicament is still operative despite the development of effective vaccines and medications. The ever-oscillating course of the pandemic yields one stable insight: Human behavior is paramount, both in term of spreading the virus and of containing it. Specifically, hygiene behavior, physical distancing, vaccine compliance, and use/misuse of social media are key behaviors impacting the spread of COVID-19. To understand human behavior in this crisis, we designed and implemented the COVID-19-Israeli Public Behavior Project (COVID-19-IPBP; Shahar et al., 2022), an interdisciplinary, multi-wave investigation of a large sample of Israelis, targeting participants' emotional distress, social relations, and compliance with guidelines during the COVID-19 crisis. This report is the second from the COVID-19-IPBP. Like the first report, it focuses on the first wave of the spread and on participants' anxiety. Still, the second report extends the first one on theoretical and empirical counts. Below we account for our focus on anxiety in mass medical crisis, and this also necessitates reviewing findings of the first report. Next, we present the rationale for the present investigation.

1.1. Anxiety during mass medical crises

While in virtually all epidemics and pandemics at least some level of anxiety is expected, no answer exists to the question of how much anxiety is warranted. Most studies focused on two extreme scenarios: a very high anxiety, also titled “panic,” and a very low anxiety, also labeled “complacency.” Evidence as to the presence of panic in the face of epidemics and pandemics is mixed, with some studies documenting its presence (e.g., Schultz et al., 2016), and others documenting its absence (e.g., Sherlaw and Raude, 2013). Conversely, there is evidence for the presence of complacency (Jones and Salathé, 2009; Wang and Kapucu, 2006). Complacency might hinder preparedness (Wang and Kapucu, 2006) and compliance with governments’ instructions, such as those concerning social distancing (Jones and Salathé, 2009).

Notably, however, most extant research on population anxiety in the face of pandemics is cross-sectional, rendering it uninformative with respect to trajectories of anxiety over time. Moreover, such research is predicated upon the assumption that the population in question is characterized by a unified reaction to the crisis. This stands in contrast to other strands of trauma research that identify heterogeneous responses to traumatic events (Donoho et al., 2017).

Both limitations are addressed in the first report from COVID-19-IPBP, targeting the first wave of the spread of COVID-19 in Israel (Shahar et al., 2022). Israeli residents have been occasionally exposed to military threats; hence they are experienced in managing life-threatening situations. Indeed, owing to ongoing threats, the government and army as well as municipal and law-enforcement agencies have considerable power over civilians, particularly at times of crisis. Similar to other (but not all) countries, Israel hermetically closed its borders, thereby controlling the level of the spread.

We examined two types of anxiety: general anxiety and virus-specific anxiety. The data used for modeling trajectories of both outcomes were six weekly assessments post-exposure, with baseline, pre-exposure data used to predict the trajectories. We hypothesized both panic and complacency trajectories, as well as a third class titled Threat Sensitivity: a moderate linear increase in both general anxiety and virus-specific anxiety throughout the spread. The rationale for proposing threat-sensitivity was that an increasing threat should result in an increasing anxious response.

Findings from General Mixture Modeling were largely consistent with our hypotheses. For both general anxiety and virus-specific anxiety, the hypothesized panic, complacency, and threat sensitivity trajectories/classes were identified. Panic was represented via elevated and stable levels of throughout the study period (12% and 25% for each outcome, respectively), whereas complacency was manifested through very low and stable levels (29% and 9%). Threat sensitivity was expressed via a linear increase in anxiety from Wave 1 (post-exposure) to Wave 4. Unexpectedly, a plateau—arguably manifesting habituation—was observed at the fourth wave, continuing until the sixth (29% and 66% for general anxiety and virus-specific anxiety). In addition, for general anxiety only, an unexpected, fourth class was evinced, exhibiting stable mid-levels (30%). Our interpretation of this class, again, post hoc, was that it reflects a balanced response, escaping the extreme high and low responses exhibited by panic and complacency. We therefore titled this class “Balanced."

Another important objective of our analyses was to predict membership in the identified classes via baseline, pre-exposure variables. As for demographics, gender, and age were revealed as important, but only for general anxiety: Women were more likely to belong to the Panic group than to each of the other three groups. Additionally, older respondents were more likely to be in the general anxiety complacency class than the balanced class. More importantly, we sought to establish directionality in the link between general anxiety – representing a psychopathological outcome – and virus-specific anxiety – deemed tantamount to “perceived stress”, a known predictor of psychopathology. Indeed, a clear directionality was evinced. When general anxiety classes were considered as outcomes, baseline levels of both general anxiety and virus-specific anxiety predicted these classes in expected ways. Baseline general anxiety was also associated with higher likelihood of membership in the panic class than any of the other three and lower likelihood of membership in the complacency class than any other class, although not discriminating between the balanced and threat-sensitive classes. Baseline virus-specific anxiety predicted a lower likelihood of membership in the complacency class compared with the panic or threat-sensitive classes and greater likelihood of membership in the panic class than the balanced class. In contrast, when virus-specific anxiety classes were considered as outcomes, baseline virus-specific anxiety – but not baseline general anxiety, predicted the three classes in an expected, linear fashion: Panic > threat sensitivity > Complacency. Thus, the directionality appears to emanate from virus-specific anxiety to general anxiety, rather than vice versa.

The first report thus identified a continuum of anxiety responses for both general anxiety and virus-specific anxiety. The continuum commences with panic as the most extremely anxious response, culminates with complacency as the least extreme anxious response, with threat sensitivity—and, for general anxiety, the balanced class —in the middle. Furthermore, findings of this first report suggest that baseline virus-specific anxiety may lead to the development of subsequent, post-exposure, general anxiety classes.

1.2. The present investigation: a social-cognitive (self-in-relations) theoretical perspective

The underlying theoretical umbrella for this investigation is a group of psychological theories focusing on representations of self-in-relations, namely, theories that capture the Subjective-Agentic Personality Sectors (SAPS), as articulated by Shahar (2020). Chief among these theories are social-cognitive theory (e.g., Bandura, 1997; Mischel and Shoda, 1995) and empirically supported versions of object relations theory (Blatt et al., 1997; Westen, 1991). The common denominator of these theories is the assumption whereby mental representations of the self, primarily in an interpersonal context, constitute the cornerstones of subjective processes, regulating behavior and response to stress (Shahar, 2020).

From the wide array of SAPS variables, we focused on three concepts: self-criticism, perceived social support, and positive and negative perceptions of the Israeli ministry of health. Self-criticism is defined as the tendency to set increasingly high self-standards and to adopt a punitive stance toward the self once these standards are not met. This dimension is considered one of the most robust dimensions of personality vulnerability to both depression and anxiety (Shahar, 2015). Mounting evidence indicate that self-criticism is distinct from low self-esteem, as the former dimension pertains to an overarching stance taken by the self toward the self, whereas the latter dimension refers to a primarily cognitive process of taking a stock of self-strength (Shahar, 2015). In fact, self-esteem appears to be an outcome of self-criticism and a moderator of self-criticism's effect on emotional distress (Abela et al., 2006). Self-criticism is shown to be associated with maladaptive coping strategies, including avoidant coping (Dunkley et al., 2006), a maladaptive strategy that is shown to be linked with COVID-19 related emotional distress (Carnahan et al., 2022). As such, self-criticism could be expected to predict membership in the more extreme anxiety classes concerning both general anxiety and virus-specific anxiety. Yet, because anxiety is strongly tied to temperamental, trait-based differences (Fox and Pine, 2012; Spielberger, 1989), it was necessary to control for temperamental, trait-based anxiety in evaluating the impact on self-criticism on symptomatic anxiety.

Self-criticism is a risk-related factor focused on the self-concept. In contrast, perceived social support is a resilience-related factor focused on other people. The literature distinguishes between received and perceived support, with the former type of support often exhibiting both protective and adverse effects whereby the later type results in mostly protective ones. In particular, the stress-buffering hypothesis of social support, according to which support ameliorates the effect of life stress on distress (e.g., Bowen et al., 2014; Cohen, 2004), is shown almost exclusively for perceived support (Cohen, 2004; e.g., (e.g., Shahar et al., 2009).

Another important social-cognitive variable is perception/attitude toward the governmental agency that oversees the COVID-19 crisis, which, in Israel, is the ministry of health. Attitudes toward agencies overseeing population-level crises have been shown to be predictive of populations' emotional responses in expected ways: positive perceptions/attitudes such as trust and respect are associated with positive emotional responses whereas negative perceptions/attitudes such as mistrust and hostility are associated with negative emotional responses (e.g., Taylor-Clark et al., 2005). Epidemics/pandemics are no exception in this respect. A strong illustration is our own program of research examining perceptions/attitudes toward the Israeli Ministry of Health during the mass-vaccination campaign launched in Israel in 2013, targeting newly-erupted poliovirus. Participants who reported experiencing the ministry of health as “caring” (a positive attitude) prior to the beginning of the campaign were particularly likely to get vaccinated, (Shahar et al., 2017), whereas participants who reported experiencing the Israeli Ministry of Health as “hysteric” (negative attitude) showed higher levels of vaccine hesitancy (Veksler-Noyman et al., 2021).

From a theoretical point of view, a focus on attitudes toward the Israeli ministry of health complements the above focus on self-and-other representation: Attitudes toward the ministry of health can be seen as an indication of individuals' perception of their higher-order social ecology, in other words, perception of their societal leadership. According to Bronfenbrenner's (1977) social-ecological theory, governmental actions represent the higher-order “exosystem”, pertaining to features of the social structure and/or large-scale societal processes. Effects of the exosystem trickle down on the micro-system (i.e., those systems directly impacting the individual) and the mesosystem (interactions between the systems directly influencing the individuals without the latter's active involvement), ultimately impacting the individual. While there are myriad ways to examine exosystemic effects, one important way may be the assessment of individuals' subjective perception of exosystemic agencies (e.g., Rizzo et al., 2021).

Therefore, we hypothesized that, H1: Even after controlling for baseline levels of general anxiety and virus-specific anxiety, demographics, and baseline trait anxiety—baseline levels of self-criticism, perceived social support, and perceptions/attitudes toward the Israeli ministry of health will predict the various classes in the panic-complacency continuum.

In terms of specific patterns, it was easiest to make predictions concerning the Panic group, because our assumption was that its members are the most impaired. Thus, we hypothesized that, H2: Over and above baseline general anxiety and virus-specific anxiety and the aforementioned role of female gender (Shahar et al., 2022), members of the general anxiety panic group will be characterized by the highest levels of trait anxiety, self-criticism, and negative perceptions/attitudes toward the Israeli ministry of health, as well as by the lowest levels of perceived social support and positive attitudes toward the Israeli ministry of health. Similar predictions were made for virus-specific anxiety panic as compared to threat sensitivity and complacency.

It was more difficult to predict specific patterns for differences between the remaining general anxiety and virus-specific anxiety classes, although it made sense to expect that the complacency group for both outcomes will differ from the other classes in terms of low trait anxiety. Other, more nuanced class differences were expected, although they were examined on an exploratory basis.

We also hypothesized that, H2: The two protective factors – perceived social support and positive perceptions/attitudes toward the Israeli ministry of health – would buffer against the adverse effect of baseline virus-specific anxiety on the pain-complacency class continuum. Recalling that Shahar et al. (2022) found that bassline virus-specific anxiety predicts subsequent general anxiety classes, but not vice versa, we see that this directional relationship is consistent with the depiction of virus-specific anxiety as the (perceived) stress leading to a distress response (general anxiety). From a risk/resilience perspective, it would be natural to expect that pre-exposure levels the two social-cognitive protective factors assessed in this study may ameliorate this adverse effect of baseline virus-specific anxiety.

2. Methods

2.1. Procedure and participants

This study received approval from the Ethics Committee of the Department of Psychology of Ben-Gurion University of the Negev in Beer-Sheva, Israel. Extensive details as to the sampling and design, participants' recruitment, and measurement are provided in (Shahar et al., 2022) and are also publicly available online (https://osf.io/gu4st/). For clarity's sake, we highlight the project's most pertinent features. We already noted that our analyses are based on seven measurement occasions: a baseline assessment and six weekly waves. In Table 1 we outline each wave's dates, related COVID-19 effects (i.e., number of cases, deceased), and related government restrictions (See also the flowchart in Fig. 1 of Shahar et al., 2022).

Table 1.

Dates of assessment waves and related COVID-19 status and restrictions.

Assessment Wave	Date	COVID-19 Status	Restrictions
Wave 0	February 19th, 2020	No known carrier	None
Wave 1	February 25th, 2020	Three adults infected	Warning against traveling to Hubei, China.
Wave 2	March 4th, 2020	16 adults infected	Mandatory quarantine to Israelis incoming from Italy, France, Spain, Austria, Germany, and Switzerland.
Wave 3	March 11th, 2020	99 adults infected; Two adults hospitalized.	Gatherings of > 100 people were prohibited, and people younger than 65 were instructed to refrain from visiting the elderly.
Wave 4	March 18th, 2020	524 adults infected	All Israelis arriving from other countries were instructed to self-quarantine. The education system was inoperative. Gathering of > 10 people was prohibited. Lockdowns were placed on targeted areas were spread was high.
Wave 5	March 25th, 2020	2436 adults infected. One hospitalized adult died from the virus.	People were instructed to Minimize outings to strictly crucial activities, and not to travel more than 100 m from home otherwise. Group praying (“minyan”) was prohibited.
Wave 6	April 1st, 2020	6168 adults infected.	Prohibitions and guidelines were dramatically increased and were enforced thereafter (e.g., fines for not wearing masks).

Data were collected online via a commercial web panel specializing in research. Because this panel only sampled Israeli-Jews, we did not have access to ethnic minorities (see limitations at the end of this article). Nonetheless, the sample was representative of the adult Israeli-Jewish population in terms of gender and age, and very closely representative in terms of residence, religiousness, education and social-economic status. The original N was 1018. Using Week 0 (the “strong baseline”) as an anchor, attrition during subsequent weeks was 14%, 19%, 20%, 23%, 23%, and 26%, respectively, with an average of 21%. Stringent data cleaning (Shahar et al., 2022) reduced N to 991, although specific analyses had to rely on listwise deletion of missing values, yielding an N of 958.

2.2. Measures

Our assessment strategy was based on numerous, very brief, self-report measures. This strategy was developed because of the rapidly changing nature of the COVID-19 crisis, frequent quarantine, and curfew, coupled with our interest in a wide array of psychosocial constructs.

General anxiety was assessed by averaging two items from the State Anxiety Inventory (SAI; Spielberger, 1989): “I am tense” and “I am anxious” (a 5-point Likert scale; Cronbach's αs > 0.87 for all waves). As indicated by (Shahar et al., 2022), mean levels of general anxiety in this sample were nearly identical to those reported by Israeli et al. (2018) following a military escalation involving missile attacks on Israeli civilians, thus further corroborating the measure's validity.

Virus-specific anxiety was assessed via a single item worded as, “To what extent are you worried/stressed from the spread of the coronavirus?” with responses on a 7-point scale. Single-item measures were shown to be successful in tapping perceived stress during mass traumas (e.g., Shahar et al., 2009).

Self-criticism was assessed via a single item taken from the Depressive Experiences Questionnaire (DEQ; Blatt et al., 1976), arguably the most extensive used measure of personality vulnerability to depression. Its 66 items tap self-criticism, dependency, and personal efficacy (construed as an index of resilience). Over the years, briefer versions of the DEQ were tested, particularly those tapping self-criticism. The briefest DEQ-self-criticism measure, developed by Shahar and colleagues (see Shahar, 2015), includes six items with a clear content validity. The item with the clearest content validity is worded: “I have a tendency to be very critical toward myself.” This item was used herein. Per the DEQ's format, a 7-point scale was used, anchored at 1 = “strongly agree” up to 7 = “strongly disagree".

Trait anxiety was assessed via a straightforward, single item, worded as follows: “I am an anxious type of person.” Because this construct was measured as a temperamental counterpart of self-criticism, the same 7-point scale used for self-criticism was applied to trait anxiety. Not only does this item appear to tap the gist of the trait anxiety construct, it is also likely to assess past anxiety disorders, albeit indirectly. The rationale here is that, while many persons who are “of the anxious type” will not develop anxiety disorders, it is hard to imagine people with anxiety disorders who will not endorse this item.

Perceived social support was measured via a single item: “To what extent to you feel that you have available emotional support from close people?” A 7-point scale was used, anchored at 1 = “Not at all” to 7 = “Very much.” Single-item measures of perceived social support have been successfully used in previous research (Knapstad et al., 2014).

Finally, we assessed negative and positive attitudes toward the Israeli Ministry of Health using a modified measure utilized by Shahar et al. (2017) and Noyman-Veksler et al. (2021). Both reports were based on a two-item measure inquiring about the extent to which respondents believe that the ministry of health is “caring” (first item; positive attitudes) and whether the ministry of health is “hysteric” (second item; negative attitudes). Although both items were shown to be predictive of compliance and hesitancy concerning the mass vaccination launched by the IMOH against poliovirus, we sought to improve the sensitivity and accuracy of this measure in the following ways. First, realizing that the word “hysteric” is colloquial, we replaced it with “anxious.” Second, we added four additions to the caring item that taps positive attitudes. Thus, we inquired about the extent to which respondents feel that the ministry of health (1) is handling the COVID-19 crisis well; (2) is caring; and (3) conducts itself confidently, and the extent to which the respondent (4) trusts the ministry of health. All five items (i.e., manages well, caring, anxiety, confidence, trust) were based on a 5-point scale, whereby 1 = “Not at all,” 2 = “A little,” 3 = “Don't know,” 4 = “To some extent,” and 5 = “Very much."

Next, we subjected these four items to a principal component analysis with varimax rotations using the STATISTICA software (StatSoft Inc, 2012). A single component was extracted with an eigenvalue of 3.00, explaining 60.10% of the variance of the items. Very high loadings of the positive attitudes items on the extracted component appeared: −0.84, -0.87,−0.84, and −0.87 for manages well, caring, confidence, and trust, respectively. The loading of the “anxiety” item on this component was small: 0.19. Thus, we created a composite score of positive attitudes by averaging the four pertinent variables (α = 0.88) and used the “anxiety” item as a single-item measure of negative attitudes.

2.3. Data analysis

Analyses proceeded in four stages.

2.3.1. Stage 1: descriptive statistics and cross-sectional regressions

Shahar et al. (2022) presented means and standard deviations of the study variables, with the exception of trait anxiety, self-criticism, perceived social support, and positive/negative perceptions of the IMOH. Hence, we calculated means and standard deviations of these additional variables, and the correlations among them. Note that these are correlations pertaining to the baseline (Week 0) assessment. Next, we examined associations between these study variables and the demographic variables: gender, religiousness binarized, education binarized, and age groups. For the first, binary variables, independent-sample t-tests were calculated, and statistically significant t-values were also accompanied by calculating effect size using Cohen's d. For the age groups, a 5-level variable, we conducted an Analysis of Variance (ANOVA) separately for each of the study variables. Statistically significant F-tests were followed by post-hoc analyses identifying group differences, using Tukey's Honest Significant Difference (HSD) with unequal Ns. Finally, we computed correlations between these variables and the variables used for Shahar et al. (2022).

2.3.2. Stage 2: cross-sectional multiple regression analyses

To examine unique associations with general and virus-specific anxiety, we regressed each of these outcomes onto demographics, trait anxiety, self-criticism, social support, and positive and negative (anxious) perceptions of the ministry of health. When general anxiety was considered as an outcome, virus-specific anxiety was added to the list of predictors, and vice versa.

2.3.3. Stage 3: tests of hypothesized main effects

As reported in Shahar et al., 2022, the N of 991 was subjected to Latent Class Growth Models and General Mixture Models using the software Mplus version 8.5 (Muthen & Muthen, Los Angeles, California). The classes/trajectories identified for general anxiety and virus-specific anxiety were then predicted using multinomial regression conducted in the context of the General Mixture Modeling analyses. Separate analyses were conducted for general anxiety and virus-specific anxiety classes. The demographic predictors were: gender, age group (see above), binarized education (academic degree completed), and binarized religiousness (non-religious [secular and traditional] vs. religious [national religious and ultraorthodox]). Employment was not entered into the analysis because 92% were employed. Similarly, income was not entered into the analysis because 109 (>10%) did not respond to the income item. In addition to gender, age group, and binarized education and religious identification, the list of predictors also included Wave 0 levels of general anxiety and virus-specific anxiety.

For the present investigation, these multinomial regressions were conducted again, except that the focal predictors were added: trait anxiety, self-criticism, perceived social support, perception of the ministry of health as anxious (negative attitudes), and perception of the ministry of health as positive/benevolent. Consistent with Shahar et al. (2022), in order to account for familywise statistical error (which may be increased by the semi-exploratory nature of some of our group comparisons), we applied Holm's (1979) correction for each predictor separately. For these effect sizes that are negatively directed (i.e., O.R.s < 1), we performed the 1/OR transformation (see Chen et al., 2010) o be able to compare them with positively directed effects (O.R.s > 1). Chen et al. (2010) recommend that ORs = 1.68, 3.47, and 6.71 are equivalent to Cohen's d 0.2 (small), 0.5 (medium), and 0.8 (large).

2.3.4. Stage 4: tests of hypothesized stress-buffering effects

To test the hypothesized stress-buffering effects, we added two 2-way multiplicative interaction terms to the models described in Stage 3: one involving baseline virus-specific anxiety by baseline perceived social support, and the other involving baseline virus-specific anxiety by the positive attitudes towards the ministry of health index. As in Stage 3, analyses were run separately for the general anxiety and virus-specific anxiety classes.

3. Results

3.1. Stage 1: descriptive statistics

Means, standard deviations, and intercorrelations among this study's variables are presented in Online Supplementary Material #1. The correlations were largely low-to-moderate. As expected, self-criticism was moderately correlated with trait anxiety.

Online Supplementary Material #2 contains t-tests examining differences in gender, religious identification, and education in terms of the study variables. Women had lower levels of positive attitudes toward the ministry of health and higher self-criticism and trait anxiety than men. Religious participants had lower levels of anxious attitudes toward the ministry of health, lower self-criticism and trait anxiety, and higher positive attitudes toward the ministry of health than non-religious participants. Participants with an academic education had lower levels of positive attitudes toward the ministry of health, but also lower trait anxiety and higher perceived support, than less educated participants. Note, however, that for almost all statistically significant t-tests, the effect sizes were small (<0.20). The single exception was a moderate effect size for religiousness and positive attitudes toward the ministry of health: Religious participants had higher values than non-religious ones. Online Supplementary Material #2 also describes an Analysis of Variance (ANOVA) with the 5-level age group variable as the predictor and the five study variables as the outcomes.

Finally, we computed correlations among the study variables and the anxiety variables used for Shahar et al. (2022). These appear in Online Supplementary Material #3. Because of the large sample size, even small correlations may reach statistical significance. Hence, correlations above 0.20 are bolded and underlined. The clear pattern emerging from Online Supplementary Material # 3 is that only trait anxiety exhibits consistent correlations greater than 0.20 magnitude with the anxiety variables across the assessment waves.

3.2. Stage 2: cross-sectional multiple regression analyses

3.2.1. General anxiety

The model accounted for 47% of the variance of general anxiety (Adjusted R ²; F _[10,944] = 87.82, p < .001; N = 955). Statistically significant predictors were: gender (women are more anxious at baseline; b = 0.05, SE = 0.02, t = 2.39, β = 0.12, p < .05), age group (the younger the participant, the less anxious he was at baseline; b = −.04, SE = 0.01, t = −2.36, β = −0.05, p < .05), religious identification (non-religious participants were more anxious at baseline; b = −0.14, SE = 0.01, t = −2.31, β = −0.05, p < .05), baseline VSA (b = 0.32, SE = 0.01, t = 22.88, β = 0.57, p < .001), trait anxiety (b = 0.09, SE = 0.01, t = 5.92, β = 0.15, p < .001), and perceived social support (b = −0.05, SE = 0.01, t = −4.18, β = −0.10, p < .05).

3.2.2. Virus-specific anxiety

The model accounted for 43% of the variance of this outcome (Adjusted R ²; F _[10,944] = 73.62, p < .001; N = 955). Statistically significant predictors were: religious identification (non-religious participants were more anxious at baseline; b = −0.32, SE = 0.11, t = −2.82, β = −0.07, p < .01), education (less educated were more anxious at baseline; b = −0.29, SE = 0.09, t = −3.13, β = −0.07, p < .001), and general anxiety (b = 1.08, SE = 0.04, t = 22.88, β = 0.62, p < .001).

3.3. Stage 3: tests of hypothesized main effects

3.3.1. General anxiety

In Online Supplementary Material #4, we present the means and standard deviations of the continuous predictors (i.e., Wave 0 levels of general anxiety, virus-specific anxiety, trait anxiety, self-criticism, perceived social support, perception of the ministry of health as anxious, and positive/benevolent perceptions of the ministry of health), which are implied by the multinomial regression analyses targeting the general anxiety classes. These implied means and standard deviations were derived from the “BCH” Command of MPlus. In Fig. 1 we visualize mean differences in these predictors as a function of the four general anxiety classes.

Fig. 1.

Implied means of the continuous study variables as a function of the General Anxiety classes. Notes: GA = General Anxiety; W0 = Wave 0; VSA = Virus-Specific Aniety; Trait Anx = Trait Anxisty; Self-Crit = Self-criticism; Perceived Sup = Perceived Social Support; IMOH Anx = Israeli Ministry of Health, perception as anxious; IMOH Ben = Israeli ministry of health, benevolent perceptions. Importantly, there are different scales for different variables. GA, IMOH-Anx and IMOH-Ben are on a 1-5-point scale, whereas VSA, Trait Anxiety, Self-Criticism, and Perceived Support are on a 1-7-point scale.

In Table 2 a and Table 2b we present results of the multinomial regression analyses targeting general anxiety classes. The columns in the tables pertain to the six comparisons emanating from the four general anxiety classes. The first column presents the putative predictors: demographics (gender, age groups, binary religiosity and education), state anxiety (Wave 0 general anxiety and virus-specific anxiety), social variables (ministry of health as anxious, positive attitudes toward the ministry, and perceived social support), and personality/self variables (trait anxiety and self-criticism). The next columns are split in half, one pertaining to the multinomial regression's estimates (b) and related standard error, and the other pertaining to odds ratios and related a 95% confidence interval. Although herein we treat only the effects suriving the Holm (1979) correction method as meaningful, we also present the second for the benefit of future meta-analyses and replications. Accordingly, p values are presented for all statistically significant effects.

Table 2.

Results of the multinomial regression analyses predicting the general anxiety classes.

2a.
	Panic vs. TS		Complacency vs. TS		Balanced vs. TS
	b/SE	O.R.(C.I.)	b/SE	O.R.(C.I.)	b/SE	O.R.(C.I.)
DEMOGRAPHICS
Gender	1.13 (p < .01)	3.10 (1.41/6.80)	−.34/.24	.70 (.43/1.14)	−.17/.28	.83 (.48/1.45)
Age Groups	−.13/.13	.87 (.67/1.13)	.13/.08	1.14 (.96/1.36)	−.08/.10	.91(.74/1.21)
Religiosity Binary	−.44/.51	.64 (.23/1.76)	.21/.28	1.23 (.70/1.66)	−.04/.34	.95 (.48/1.88)
Education Binary	.04/.37	1.05 (.50/2.16)	.25/.25	1.28 (.78/2.11)	.59/.28 (p < .05)	1.80 (1.04/3.12)
STATE ANXEITY
GA Wave 0	.98/.18 (p < .001)	2.68 (1.86/3.85)	−.65/.19 (p < .01)	.52 (.35/.77) 1/OR = 1.92	−.04/.17	.95 (.67/1.36)
VSA Wave 0	.23/.14	1.26 (.95/1.66)	−.42/.08 (p < .01)	.65 (.55/.78) 1/OR = 1.53	−.31/.09 (p < .01)	.73 (.60/.88) 1/OR = 1.36
SOCIAL VARIABLES
IMOH-ANX	.29/.13	1.22 .94/1.59	−.04/.10	.96 (.78/1.17)	.05/.10	1.05(.85/1.29)
IMOH-POS	−.35/.16 (p < .05)	.70 .51/.97)	.21/.11	1.23 (.99/1.53)	.10/.12	1.11 (.87/1.42)
Perceived Support	−.28/.09 (p < .01)	.75 (.62/.90) 1/OR = 1.33	.02/.07	1.02 (.88/1.18)	−.08/.08	.91 (.77/1.08)
PERSONALITY AND SELF
Trait Anxiety	.08/.10	1.08 (.88/1.34)	−.29/.07 (p < .001)	.74 (.64/.87) I/OR = 1.35	−.10/.08	.89 (.76/1.05)
Self-criticism	.34/.12 (p < .01)	1.41 (1.11/1.79l)	.02/.08	1.02 (.87/1.20)	−.05/.10	.94 (.77/1.14)

2b.
	Balanced vs. Panic		Balanced vs. Complacency		Complacency vs. Panic
	b/SE	O.R.(C.I.)	b/SE	O.R.(C.I.)	b/SE	O.R.(C.I.)
DEMOGRAPHICS
Gender	−1.30/.41 (p < .01)	.27 (.12/.60)| 1/OR = 3.70	.17/.24	1.18 (.74/1.90)	−1.48/.41 (p < .01)	.22 (.10/.51) 1/OR = 4.54
Age Groups	.04/.13	1.04 (.80/1.36)	−.22/.08 (p < .01)	.79 (.67/.94) 1/OR = 1.26	.27/.13 (p < .05)	1.31 (1.00/1.71)
Religiosity Binary	.39/.51	1.48 (.54/4.06)	−.26/.29	.77 (.43/1.36)	.65/.51	1.92 (.70/5.31)
Education Binary	.54/.37	1.71 (.83/3.55)	.33/.24	1.40 (.87/2.25)	.20/.38	1.22 (.57/2.61)
STATE ANXEITY
GA Wave 0	−1.02/.20 (p < .001)	.35 (.24/.53) 1/OR = 2.85	.60/.21 (p < .01)	1.83 (1.20/2.81)	−1.63/.24 (p < .001)	.19 (.12/.31) 1/OR = 5.26
VSA Wave 0	−.54/.14 (p < .0001)	.58 (.43/.77) 1/OR = 1.72	.11/.08	1.11 (.95/1.31)	−.65/.15 (p < .001)	.51 (.38/.69) 1/OR = 1.96
SOCIAL VARIABLES
IMOH-ANX	−.15/.14	.85 (.65/1.12)	.09/.09	1.09 (.90/1.32)	−.24/.14	.78 (.58/1.04)
IMOH-POS	.45/.17 (p < .01)	1.51 (1.13/2.21)	−.10/.11	.90 (.72/1.12)	1.24/	1.75 (1.24/2.47)
Perceived Support	.19/.10 (p < .06)	1.21 (.99/1.48)	−.11/.06	.89 (.78/1.02)	.30/.10 (p < .01)	1.36 (1.11/1.66)
PERSONALITY AND SELF
Trait Anxiety	−.19/.11	.82 (.66/1.02)	.18/.08 (p < .05)	1.20 (1.02/1.41)	−.37/.11 (p < .01)	.68 (.54/.86) 1/OR = 1.47
Self-criticism	.40/.13 (p < .01)	.66 (.51/.85) 1/OR = 1.51	−.08/.07	.91 (.79/1.07)	−.32/.12 (p < .05)	.72 (.56/.93) 1/OR = 1.38

Notes.

SE = Standard Errors; OR = Odd Ratios; TS = Threat Sensitivity Class; CI = Confidence Interval; GA = General Anxiety; VSA = Virus Specific Anxiety; IMOH-ANX = Israeli Ministry of Health – Anxiety Item; IMOH-PO = Israeli Ministry of Health, Positive Attitudes Index.

As shown in Tables 2a and 2b, the general anxiety panic class was characterized by a higher prevalence of women and higher levels of Wave 0 general anxiety and self-criticism, compared to each of the other classes. Compared to general anxiety threat sensitivity and complacency, but not Balanced, the general anxiety panic class had lower levels of perceived social support. General anxiety panic also had higher levels of Wave 0 virus-specific anxiety and lower levels of positive attitudes toward the ministry of health than the balanced and complacency classes (but not threat sensitivity).

General anxiety complacency differed from each of the other classes in terms of lowest levels of Wave 0 general anxiety. Additionally, general anxiety complacency had lower levels of Wave 0 VSA and trait anxiety compared to general anxiety panic and threat sensitivity (but not balanced), and older age compared to Balanced. Finally, threat sensitivity differed from balanced only in terms of Wave 0 virus-specific anxiety (higher in the former class).

Most of the effects surviving the Holm correction were small in size. Three exceptions were evinced: a moderate effect size for the higher prevalence of women in general anxiety panic compared to balanced; the large effect size for the higher prevalence of women in general anxiety panic compared to complacency; and the large effect size for the higher level of Wave 0 general anxiety in general anxiety panic compared to complacency.

3.3.2. Virus-specific anxiety

In Online Supplementary Material #5, we present results of the multinomial regression analyses targeting virus-specific anxiety classes. Here, the pattern is much simpler than the one evinced for general anxiety. As expected, the three classes were ordered linearly in terms of Wave 0 virus-specific anxiety: Panic > threat sensitivity > complacency. The only additional effect surviving Holm's correction pertained to trait anxiety differentiating between panic and threat sensitivity (higher levels in the former class). The size of the effect was mostly small, with the single exception of a large effect size evinced for Wave 0 virus-specific anxiety in differentiating panic and complacency.

3.4. Stage 4: tests of hypothesized stress-buffering effects

No statistically significant interactions were found.

4. Discussion

Building on Shahar et al. (2022), we examined theory-based psychosocial predictors of anxiety trajectories during the first wave of the spread of COVID-19 in Israel. Using a group of theories that focus on mental representations of self and others (see integration in Shahar, 2020), as well as Bronfenbrenner's (1977) social-ecological theory, we focused on four social-cognitive predictors: self-criticism, perceived social support, and negative and positive attitudes towards the Israeli Ministry of Health. The predictive effect of these variables was examined while controlling for the potentially confounding effect of trait anxiety. Multinomial regression analyses were run with these novel predictors, over and above those examined by Shahar et al. (2022).

Our findings reaffirm the strong predictive effect of baseline levels of virus-specific anxiety. This variable was the only state-anxiety predictor differentiating all virus-specific anxiety classes. Moreover, baseline virus-specific anxiety consistently predicted general anxiety classes, on top of the expected predictive effect of baseline general anxiety. Also noteworthy is the failure of perceived social support and positive attitudes toward the ministry of health to buffer against this predictor's effect on both general anxiety and virus-specific anxiety classes. While this finding is highly consistent with global reports about the devastating mental health effects of COVID-19. It should be mentioned that two studies in Israel reported an actual reduction of the incidence of attempted suicide (Travis-Lumer et al., 2021a) and incident schizophrenia (Travis-Lumer et al., 2022) despite an increase in the incident rate of antidepressant use (Frangou et al., 2022). Notably, however, authors of the two studies acknowledge the possibility of an increased risk for attempted suicide in the more distant future, and they also forecasted – mathematically – an increased risk for incident schizophrenia once restrictions are lifted.

Social-cognitive predictors were only relevant to general anxiety classes. Members of the general anxiety panic class were more self-critical than all other counterparts, had lower levels of perceived social support than threat sensitivity and complacency, and had lower levels of positive attitudes toward the ministry of health than balanced and complacency. Thus, in addition to panic members' being more anxious pre-exposure, they also appear to be highly demoralized, having little faith in themselves, close others, and social institutions overseeing the crisis. That members of this class were also over-represented by females is consistent with long-standing findings as to the higher vulnerability of women – compared with men – to emotional disorders, both prior to, and during, COVID-19 (Liu et al., 2021).

Our results are consistent with the existence of a complex, panic-to-complacency continuum concerning the population's anxious response to the COVID-19 crisis. Thus, while at the fifth assessment wave (Wave 4), panic and threat sensitivity evinced similar levels of general anxiety, we found that the two classes differed markedly in terms of demographics and social cognition: The general anxiety panic class consisted of more women, and exhibited higher of self-criticism and lower perceived support. Similarly, complacency and balanced—the two classes with the lowest levels of general anxiety— differed in terms of age and baseline general anxiety. Additionally, threat sensitiivty and balanced, constituting the two mid-level general anxiety classes, nevertheless differed in terms of baseline virus-specific anxiety. Even in the case of virus-specific anxiety classes, where the pattern was much more parsimonious, baseline virus-specific anxiety differentiated the middle group, threat sensitivity, from both the highest (panic) and the lowest (complacency). Finally, threat sensitivity had lower levels of trait anxiety than panic. These effects encourage a nuanced appraisal of the unfolding of anxiety during medical crises.

4.1. Limitations and strengths

While the representativeness of our sample in terms of Israel's Jewish adult population appears to be very high, ethnic minorities (e.g., Arab, Bedouin) were left out because of the online sampling procedure (see Shahar et al., 2022). It is thus incumbent on us to reach out to all Israeli ethnic minorities. Another limitation pertains to our reliance on very brief, solely self-report measures. Although the results attained constitute evidence for the psychometric quality of our abbreviated measures, long measures are psychometrically superior and may have yielded additional findings. In particular, the single item assessing negative attitudes toward the Israeli Ministry of Health – “anxious” – did not yield statistically significant findings, and is thus flagged for future replication and upgrades. Finally, generalizability to countries outside of Israel should be made with great caution.

The study's strengths include the utilization of a strong, pre-exposure baseline, the employment of high-resolution assessments, high retention of participants across assessment waves, and the implementation of General Mixture Modeling analyses and multinomial regression analyses with Holm Correction. These strengths yield a fine-grained understanding of the unfolding of anxiety in the Jewish-Israeli population, while also enabling causal inference (within the confines of non-experimental studies).

4.2. Public health implications

That people who are generally anxious before the crisis (whether because of general state, crisis-specific, or trait anxiety) become even more anxious throughout the crisis underscores the utility of mass screening of anxiety by government officials and the employment of brief, digital interventions to address the fears of anxious individuals (Dorison et al., 2020).

Members of the general anxiety panic class were highly demoralized. Demoralization is identified as a barrier to evidence-based psychological treatment (e.g., Tarescavage et al., 2015), and may also derail compliance with official guidelines. Identifying and intervening with demoralized individuals may thus increase compliance and improve the management of the medical crisis.

The third implication concerns leadership. Attitudes toward the acting agency (e.g., the ministry of health) may be expressed in terms of personification, that is, by individuals experiencing the agency as a protagonist in their lives (Shahar et al., 2017; Noyman-Veksler et al., 2021). Leaders of the acting agency are likely to symbolize this protagonist, and thus have the power to shape the nature of the aforementioned personification. Calm, coherent, and realistically optimistic leadership has the potential to tilt the personification in a positive direction, thus increasing individuals' alliance with the agency and compliance with its guidelines (Kahn, 2009).

5. Conclusions

As of the time of writing of this paper, the entire world has been hit by six waves of the pandemic since its declaration as a pandemic at the beginning of 2020. Despite the presence of effective vaccinations, the world still braces itself to collaborate globally in the fight against COVID-19, and the public's behavior is still considered the paramount factor dictating the level of the spreads. Emerging public health threats from monkeypox, re-emergence of polio in several countries and the effects of climate change show the importance of developing broad perspectives on public health responses. Understanding public anxiety and its unfolding during the crisis is a crucial element in promoting effective public health interventions, demanding multi-disciplinary collaborations. Preexisting anxiety shapes subsequent anxious responses to the spread of COVID-19. The general anxiety panic group may be markedly demoralized, requiring public-health attention. Because self-criticism has been shown, over almost five decades of research, to constitute a formidable marker of vulnerability to psychopathology and/distress, we believe that the panic group could be identified – both in the general population and in clinics – as evincing elevated levels of general anxiety coupled with high levels of self-criticism.

The COVID-19-Israeli Public Behavior Project can serve as an example for an interdisciplinary collaboration. Nevertheless, it remains to be seen whether these kinds of projects are embedded within existing public health structures sustainably. As mentioned above, leadership is crucial during emergencies, but not less important during “normal” preparation and capacity-building times. |Collaboration between Israeli and European universities and public health institutes are currently underway, focusing on capacity building among future public health leaders (Bashkin et al., 2021). We hope that our findings would be useful to policymakers and public health experts – inside and beyond Israel – not just in stirring public behavior toward responsible routes, but on broader terms how to create the structures to study and engage in understanding better how best to engage with different parts of society.

Credit author statement

Golan Shahar,Ph.D: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Roles/Writing - original draft.; Limor Ahronson-Daniel,Ph.D: Conceptualization; Funding acquisition; Investigation; Methodology; Roles/Writing – Review & Editing.; David Greenberg, MD: Conceptualization; Funding acquisition; Investigation; Methodology; Roles/Writing – Review & Editing.; Hadar Shalev, MD: Conceptualization; Funding acquisition; Investigation; Methodology; Roles/Writing – Review & Editing.; Avichai Tendler, Ph.D: Conceptualization; Formal analysis; Investigation; Methodology; Roles/Writing – Review & Editing.; Itamar Grotto, M.D., Ph.D: Conceptualization; Investigation; Methodology; Roles/Writing – Review & Editing.; Patrick Malone, Ph.D.: Formal analysis; Methodology; Roles/Writing - original draft.; Nadav Davidovitch, MD, Ph.D.: Conceptualization; Funding acquisition; Investigation; Methodology; Roles/Writing – Review & Editing.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.