This study developed and validated a novel scale to assess individuals’ fear perception and adaptability in pandemic. It was approved by Ethics Committee of Institute of Intelligent Machines, Chinese Academy of Sciences, Hefei. The questionnaire (including the scale) was published online. All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The online survey was restricted to adults aged 18 years or older. Before beginning the questionnaire, participants were presented with an information page describing the study purpose, procedures, and data protection. Proceeding to complete the survey indicated their consent. No identifying information (e.g., names, phone numbers, ID numbers) was collected, ensuring complete anonymity.

Firstly, we conducted a comprehensive review of fear-related scales and behaviors that fear may induce during epidemics. We organized an initial pool of 18 items based on existing fear scales and other psychological measures, focusing on three aspects: self-efficacy, perceived susceptibility, and perceived threat. Additionally, we identified four of the most common behavioral changes from academic articles, social media, official news, and field observations as external validators. To evaluate content validity, we invited a panel of eight experts, including two behavioral scientists, two clinical psychologists, two nurses specializing in infectious disease care, and two clinical physicians. The experts were asked to independently assess each item on several criteria: (a) relevance to the construct, (b) clarity of wording, (c) theoretical representativeness, and (d) practical comprehensibility for the general population. Ratings were made on a 4-point scale (1 = not relevant, 4 = highly relevant). There were three iterative rounds of review, each including at least six of the experts, with partial changes in panel composition to avoid groupthink. Finally, 11 of the original 18 items were removed due to low relevance or redundancy, and the four behavioral external validators were refined into three. The final version of the Pandemic Fear Perception and Adaptability Scale (PFPA) contained seven items: two items assessing self-efficacy, three assessing perceived susceptibility, and two assessing perceived threat. All items are rated on a 7-point Likert scale (1 = totally disagree, 7 = strongly agree).

The seven items of PFPA are shown below:

Subscale ‘Self-efficacy’ included:

Subscale “Perceived susceptibility” included:

Subscale ‘Perceived Threat’ included:

The two self-efficacy items were explicitly anchored to the three external validators. Specifically, participants were first asked whether they had engaged in these behaviors (binary yes = 1/no = 2). Then, the self-efficacy subscale assessed their perceived ability to perform these same behaviors and the belief that performing them would be effective. This design ensured that self-efficacy ratings were grounded in concrete, context-specific behaviors rather than abstract general beliefs. These questions are not in the PFPA, but as the external behavioral validators.

Fear-induced behavior-change questions included:

At last, 8 individuals (5 men and 3 women, mean age = 27.925 years, SD = 5.233) were asked to answer the initial scale and a four-point Likert scale (do not understand, partial understand, understand and totally understand) that was utilized to assess whether they understood the meaning of each item. Seven individuals indicated ‘totally understand’ and one expressed ‘understand’. The total score of PFPA ranges from 7 to 49. Higher score of overall PFPA represents more severe fear of the pandemic. The higher score in subscales indicates higher self-efficacy, more perceived susceptibility and more threat individuals felt.

In addition to the PFPA, we included six affect-related items adapted from the State–Trait Anxiety Inventory (STAI). These items assessed feelings of calmness, security, peacefulness, happiness, and relaxation, as well as one item reflecting fearfulness. Most items were positively worded, representing the conceptual opposite of fear. Responses were recorded on a 4-point Likert scale (1 = totally disagree, 4 = strongly agree). Although these items do not constitute a complete STAI subscale, they were included to provide preliminary convergent validity evidence. The detailed information for scales of PFPA, external validators and STAID items can be checked in Supplementary Material.

The questionnaire was distributed online via a survey platform called ‘Questionnaire Star’, which is a professional online platform for surveys, exams, assessments, and voting, widely utilized in commercial, research, and personal interest fields in China. The questionnaires were distributed and collected at the end of December 2022 and the beginning of January 2023, during which period of time the local government lifted most of the COVID-19 prevention and control measures. Participation in this study was entirely voluntary.

A total of 1,068 individuals completed the online survey. To ensure data quality and ethical compliance, a multi-step screening procedure was applied to the raw responses. First, an attention check item was embedded in the questionnaire (“Please select ‘Moderate’ for this question”). Participants who failed this check were excluded, resulting in 999 valid participants. Second, individuals younger than 18 years old, as well as respondents whose answers were clearly nonsensical (e.g.: not a number), were removed, leaving 959 participants. Third, two pairs of questions with contradictory meanings were used to further detect random or careless responses, ‘I feel very secure’ and ‘I feel very fearful’. The answer of ‘I feel very fearful’ did reverse-scoring first. Participants who endorsed mutually exclusive extreme values (e.g., responding “1” to one item and “4” to its opposite) were excluded. After this final screening step, the analytic sample comprised 923 participants. The flowchart of participants screening is shown in Figure 1. Thirty participants were asked to answer the scale again the day after first participating, for the purpose of test–retest. All participants were Chinese and speaking Mandarin. The background information of participants is shown in Table 1.

The psychometric evaluation of the PFPA was conducted using both Classical Test Theory (CTT) and the Rasch model. CTT analyses included internal consistency (Cronbach’s α), test–retest reliability, intraclass correlation coefficients, corrected item–total correlations, mean inter-item correlations, and the standard error of measurement (SEM).

Construct validity was assessed within a Confirmatory Factor Analysis (CFA) framework using Structural Equation Modeling (SEM). The CFA was performed on the seven Likert items, testing the hypothesized three-factor model (self-efficacy, perceived susceptibility, perceived threat). Model fit was evaluated using χ²/df, Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Goodness of Fit Index (GFI), and Root Mean Square Error of Approximation (RMSEA). In addition, Composite Reliability (CR) and Average Variance Extracted (AVE) were computed to assess convergent validity. Standardized factor loadings were also reported. CFA was implemented, with the Maximum Likelihood estimator, chosen given the sample size and approximate multivariate normality assumptions, which were tested prior to analysis.

Rasch analysis included item and person separation reliability, item and person separation indices, point–measure correlations (PT-Measure), differential item functioning (DIF) across sex and age (median split at 25 years), and infit and outfit mean-square statistics (MNSQ).

For external validity, Pearson or Spearman correlations (as appropriate) were computed between PFPA scores and (a) six STAI-derived items (positively worded affective states) and (b) three binary behavioral validators.

All tests were two-tailed with a significance threshold of p < 0.05, and 95% confidence intervals were reported where applicable (90% intervals were used for RMSEA as per standard SEM practice). Results are presented to three decimal places (except Rasch outputs, which were limited by WINSTEPS formatting). Analyses were conducted in SPSS 25.0, including function AMOS, and WINSTEPS 3.75.0.

Classical Test Theory (CTT) analysis was conducted to evaluate the reliability of the PFPA scale with a sample of 923 participants. As shown in Table 2, the overall scale demonstrated acceptable internal consistency (Cronbach’s α = 0.700), with subscale alphas: 0.703 (Self-efficacy), 0.688 (perceived susceptibility), and 0.833 (perceived threat). Test–retest reliability was evaluated using the Intraclass Correlation Coefficient (ICC(2,1), two-way random effects, single measure, absolute agreement) for a subsample of 30 participants, with the 95% confidence interval calculated using Fisher’s Z transformation [0.590, 0.891], p < 0.001. The common inter-item correlation was 0.255, indicating moderate item interrelatedness, and the SEM was 3.936, reflecting adequate precision for a new scale. Item-level statistics, including corrected item-total correlations (0.205–0.528) and alpha if item deleted (0.638–0.714), confirmed no item substantially reduced reliability, as shown in Table 3.

As shown in Figure 2, inter-item and subscale correlations were examined to assess the internal consistency of the PFPA scale (N = 923). The average inter-item correlation was 0.255, indicating moderate item interrelatedness suitable for a multidimensional scale. Item correlations ranged from 0.020 (Item2-Item4) to 0.713 (Item6-Item7), with stronger correlations within subscales (e.g., Item3-Item4, r = 0.525; Item6-Item7, r = 0.713) supporting their coherence. Subscale correlations ranged from 0.124 (Self-efficacy and Perceived Susceptibility) to 0.516 (Perceived Susceptibility and Perceived Threat), confirming related but distinct constructs. These findings, visualized in heatmaps (Figure 2), reinforce the scale’s reliability while supporting discriminant validity, as subscale correlations remained below 0.85.

As shown in Table 2, Confirmatory Factor Analysis (CFA) was conducted to evaluate the construct validity of the Pandemic Fear Perception and Adaptability Scale (PFPA) with a sample of 923 participants. As shown in Table 2, the overall scale CR was 0.898, with subscale CRs of 0.730 (self-efficacy), 0.717 (perceived susceptibility), and 0.835 (perceived threat), all exceeding the 0.7 threshold for acceptable reliability.

Moreover, the hypothesized three-factor model (self-efficacy, perceived susceptibility, perceived threat) demonstrated excellent fit with the data: χ²(11) = 15.123, p = 0.177; χ²/df = 1.375; CFI = 0.998; GFI = 0.995; TLI = 0.995; RMSEA = 0.020 (90% CI: 0.000–0.043, PCLOSE = 0.989); RMR = 0.046; AGFI = 0.988; AVE = 0.563. The non-significant χ² value indicates no substantial deviation between the observed and hypothesized covariance matrices, supporting model fit. The χ²/df ratio of 1.375, well below the threshold of 2.0, suggests a parsimonious model. CFI, GFI, and TLI values exceeding 0.95, along with an RMSEA of 0.020 and low RMR of 0.046, further confirm excellent fit.

Rasch analysis was performed on 923 participants. As shown in Table 2, the item separation reliability was 0.97, and the item separation index was 5.83, indicating excellent item differentiation. The person separation reliability was 0.61, with a person separation index of 1.25, acceptable for early-stage development. As shown in Table 3, Infit and Outfit Mean Square (MNSQ) values ranged from 0.82 to 1.16 (Infit) and 0.82 to 1.36 (Outfit), within the acceptable range of 0.5–1.5. ICC plots (Figure 3) showed good alignment for most items, with Items 2 and 5 exhibiting empirical points exceeding 95% confidence intervals in several categories, indicating slight misfit possibly due to response variability. Differential item functioning (DIF) analysis revealed minimal bias across gender (|DIF| ≤ 0.06) and age (|DIF| ≤ 0.21), supporting measurement invariance.

Convergent validity of the PFPA scale was assessed by correlating its total and subscale scores with six affect-related items adapted from the STAID and a behavioral total score using Pearson correlations (N = 923). The Cronbach’s Alpha of STAID itself is 0.866. The PFPA total score showed a significant negative correlation with the STAI-derived items (r = −0.204, 95% CI [−0.265, −0.141], p < 0.001), as well as its subscales: self-efficacy (r = 0.227, 95% CI [0.165, 0.287], p < 0.001), perceived susceptibility (r = −0.305, 95% CI [−0.362, −0.245], p < 0.001), and perceived threat (r = −0.215, 95% CI [−0.276, −0.153], p < 0.001).

For the external behavioral validators, the PFPA total score was again significantly correlated (r = −0.219, 95% CI [−0.280, −0.157], p < 0.001), as were its subscales: self-efficacy (r = −0.155, 95% CI [−0.217, −0.091], p < 0.001), perceived susceptibility (r = −0.148, 95% CI [−0.211, −0.084], p < 0.001), and perceived threat (r = −0.173, 95% CI [−0.235, −0.110], p < 0.001).

The PFPA showed a relatively low Person Separation Index (PSI = 1.25), suggesting limited sensitivity in distinguishing individuals with different levels of fear. Such findings are common in short scales during early validation stages. In our study, in the context of China’s abrupt relaxation of COVID-19 restrictions in December 2022, participants exhibited high homogeneity in fear responses, potentially reducing response variance. This situational specificity suggests that the PFPA effectively captured collective fear reactions during a crisis, rather than indicating a design flaw. Moreover, internal consistency concerns emerged in the self-efficacy and Perceived susceptibility, which showed weaker fit in Rasch and ICC analyses. It should also be noted that short item subscales typically depress Cronbach’s α because reliability coefficients are sensitive to the number of items, even when the items are conceptually appropriate. At last, although the overall internal consistency of the PFPA reached the commonly recommended threshold (α = 0.700) in the final sample, one subscale (perceived susceptibility) showed a Cronbach’s α slightly below 0.70 (α = 0.688). This is consistent with prior psychometric guidance that values between 0.60 and 0.70 are acceptable in early-stage validation, particularly for short scales with few items. Nevertheless, future studies should consider adding or refining items to enhance the reliability of this dimension.

The PFPA was not directly compared against widely validated pandemic-related fear scales such as the FCV-19S or the CAS. This limits the ability to situate PFPA within the broader measurement landscape. To partially address this, we correlated PFPA with affect-related items adapted from the STAID, which showed expected negative associations, as well as with behavioral items reflecting protective actions, which showed positive associations. While these analyses provided preliminary convergent and external validity evidence, future research should directly examine correlations with FCV-19S, CAS, and related measures in different populations.

The study sample was skewed toward younger adults, reflecting the online convenience recruitment strategy. This demographic imbalance limits generalizability, particularly to older populations who may perceive and respond to pandemic fear differently. Although this sampling strategy was chosen to capture real-time data immediately following the sudden lifting of COVID-19 restrictions, future work should test the PFPA in more representative and age-diverse samples.

The PFPA was developed in the unique sociocultural context of China during the immediate post-lockdown period in December 2022. While this setting provides rare and valuable insights, the findings may not generalize across different cultural contexts. Future studies should adapt and validate the PFPA cross-culturally, testing measurement invariance across diverse populations (e.g., Western countries, Southeast Asia, Africa), where health literacy, risk perception, and pandemic experiences differ.

Certain methodological choices—such as limited cognitive pre-testing and reliance on convenience sampling—were constrained by the urgency of the situation. Nevertheless, these data offer an invaluable snapshot of fear and adaptability during a critical transition period. Future studies under more controlled conditions can refine the PFPA psychometric robustness and broaden its practical applications.